1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011,
16 Free Software Foundation, Inc.
18 Permission is granted to copy, distribute and/or modify this document
19 under the terms of the GNU Free Documentation License, Version 1.3 or
20 any later version published by the Free Software Foundation; with the
21 Invariant Sections being ``GNU General Public License'' and ``Funding
22 Free Software'', the Front-Cover texts being (a) (see below), and with
23 the Back-Cover Texts being (b) (see below). A copy of the license is
24 included in the gfdl(7) man page.
26 (a) The FSF's Front-Cover Text is:
30 (b) The FSF's Back-Cover Text is:
32 You have freedom to copy and modify this GNU Manual, like GNU
33 software. Copies published by the Free Software Foundation raise
34 funds for GNU development.
36 @c Set file name and title for the man page.
38 @settitle GNU project C and C++ compiler
40 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
41 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
42 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
43 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
44 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
45 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
46 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
48 Only the most useful options are listed here; see below for the
49 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
52 gpl(7), gfdl(7), fsf-funding(7),
53 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
54 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
55 @file{ld}, @file{binutils} and @file{gdb}.
58 For instructions on reporting bugs, see
62 See the Info entry for @command{gcc}, or
63 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
64 for contributors to GCC@.
69 @chapter GCC Command Options
70 @cindex GCC command options
71 @cindex command options
72 @cindex options, GCC command
74 @c man begin DESCRIPTION
75 When you invoke GCC, it normally does preprocessing, compilation,
76 assembly and linking. The ``overall options'' allow you to stop this
77 process at an intermediate stage. For example, the @option{-c} option
78 says not to run the linker. Then the output consists of object files
79 output by the assembler.
81 Other options are passed on to one stage of processing. Some options
82 control the preprocessor and others the compiler itself. Yet other
83 options control the assembler and linker; most of these are not
84 documented here, since you rarely need to use any of them.
86 @cindex C compilation options
87 Most of the command-line options that you can use with GCC are useful
88 for C programs; when an option is only useful with another language
89 (usually C++), the explanation says so explicitly. If the description
90 for a particular option does not mention a source language, you can use
91 that option with all supported languages.
93 @cindex C++ compilation options
94 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
95 options for compiling C++ programs.
97 @cindex grouping options
98 @cindex options, grouping
99 The @command{gcc} program accepts options and file names as operands. Many
100 options have multi-letter names; therefore multiple single-letter options
101 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
104 @cindex order of options
105 @cindex options, order
106 You can mix options and other arguments. For the most part, the order
107 you use doesn't matter. Order does matter when you use several
108 options of the same kind; for example, if you specify @option{-L} more
109 than once, the directories are searched in the order specified. Also,
110 the placement of the @option{-l} option is significant.
112 Many options have long names starting with @samp{-f} or with
113 @samp{-W}---for example,
114 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
115 these have both positive and negative forms; the negative form of
116 @option{-ffoo} is @option{-fno-foo}. This manual documents
117 only one of these two forms, whichever one is not the default.
121 @xref{Option Index}, for an index to GCC's options.
124 * Option Summary:: Brief list of all options, without explanations.
125 * Overall Options:: Controlling the kind of output:
126 an executable, object files, assembler files,
127 or preprocessed source.
128 * Invoking G++:: Compiling C++ programs.
129 * C Dialect Options:: Controlling the variant of C language compiled.
130 * C++ Dialect Options:: Variations on C++.
131 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
133 * Language Independent Options:: Controlling how diagnostics should be
135 * Warning Options:: How picky should the compiler be?
136 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
137 * Optimize Options:: How much optimization?
138 * Preprocessor Options:: Controlling header files and macro definitions.
139 Also, getting dependency information for Make.
140 * Assembler Options:: Passing options to the assembler.
141 * Link Options:: Specifying libraries and so on.
142 * Directory Options:: Where to find header files and libraries.
143 Where to find the compiler executable files.
144 * Spec Files:: How to pass switches to sub-processes.
145 * Target Options:: Running a cross-compiler, or an old version of GCC.
146 * Submodel Options:: Specifying minor hardware or convention variations,
147 such as 68010 vs 68020.
148 * Code Gen Options:: Specifying conventions for function calls, data layout
150 * Environment Variables:: Env vars that affect GCC.
151 * Precompiled Headers:: Compiling a header once, and using it many times.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
166 -pipe -pass-exit-codes @gol
167 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
168 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
169 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
171 @item C Language Options
172 @xref{C Dialect Options,,Options Controlling C Dialect}.
173 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
174 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
175 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
176 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
177 -trigraphs -traditional -traditional-cpp @gol
178 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
179 -fsigned-bitfields -fsigned-char @gol
180 -funsigned-bitfields -funsigned-char}
182 @item C++ Language Options
183 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
184 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
185 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
186 -fno-elide-constructors @gol
187 -fno-enforce-eh-specs @gol
188 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
189 -fno-implicit-templates @gol
190 -fno-implicit-inline-templates @gol
191 -fno-implement-inlines -fms-extensions @gol
192 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
193 -fno-optional-diags -fpermissive @gol
194 -fno-pretty-templates @gol
195 -frepo -fno-rtti -fstats -ftemplate-backtrace-limit=@var{n} @gol
196 -ftemplate-depth=@var{n} @gol
197 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
198 -fno-default-inline -fvisibility-inlines-hidden @gol
199 -fvisibility-ms-compat @gol
200 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
201 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wnarrowing @gol
202 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
203 -Weffc++ -Wstrict-null-sentinel @gol
204 -Wno-non-template-friend -Wold-style-cast @gol
205 -Woverloaded-virtual -Wno-pmf-conversions @gol
208 @item Objective-C and Objective-C++ Language Options
209 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
210 Objective-C and Objective-C++ Dialects}.
211 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
212 -fgnu-runtime -fnext-runtime @gol
213 -fno-nil-receivers @gol
214 -fobjc-abi-version=@var{n} @gol
215 -fobjc-call-cxx-cdtors @gol
216 -fobjc-direct-dispatch @gol
217 -fobjc-exceptions @gol
220 -fobjc-std=objc1 @gol
221 -freplace-objc-classes @gol
224 -Wassign-intercept @gol
225 -Wno-protocol -Wselector @gol
226 -Wstrict-selector-match @gol
227 -Wundeclared-selector}
229 @item Language Independent Options
230 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
231 @gccoptlist{-fmessage-length=@var{n} @gol
232 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
233 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
235 @item Warning Options
236 @xref{Warning Options,,Options to Request or Suppress Warnings}.
237 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
238 -pedantic-errors @gol
239 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
240 -Wno-attributes -Wno-builtin-macro-redefined @gol
241 -Wc++-compat -Wc++11-compat -Wcast-align -Wcast-qual @gol
242 -Wchar-subscripts -Wclobbered -Wcomment @gol
243 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
244 -Wno-deprecated-declarations -Wdisabled-optimization @gol
245 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
246 -Wno-endif-labels -Werror -Werror=* @gol
247 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
248 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
249 -Wformat-security -Wformat-y2k @gol
250 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
251 -Wignored-qualifiers @gol
252 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
253 -Winit-self -Winline -Wmaybe-uninitialized @gol
254 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
255 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
256 -Wlogical-op -Wlong-long @gol
257 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
258 -Wmissing-include-dirs @gol
260 -Wno-multichar -Wnonnull -Wno-overflow @gol
261 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
262 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
263 -Wpointer-arith -Wno-pointer-to-int-cast @gol
264 -Wredundant-decls @gol
265 -Wreturn-type -Wsequence-point -Wshadow @gol
266 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
267 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
268 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
269 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
270 -Wmissing-format-attribute @gol
271 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
272 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
273 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
274 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
275 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
276 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
277 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
278 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
279 -Wvla -Wvolatile-register-var -Wwrite-strings -Wzero-as-null-pointer-constant}
281 @item C and Objective-C-only Warning Options
282 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
283 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
284 -Wold-style-declaration -Wold-style-definition @gol
285 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
286 -Wdeclaration-after-statement -Wpointer-sign}
288 @item Debugging Options
289 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
290 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
291 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
292 -fdisable-ipa-@var{pass_name} @gol
293 -fdisable-rtl-@var{pass_name} @gol
294 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
295 -fdisable-tree-@var{pass_name} @gol
296 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
297 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
298 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
299 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
300 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
302 -fdump-statistics @gol
304 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
308 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-nrv -fdump-tree-vect @gol
318 -fdump-tree-sink @gol
319 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
320 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
321 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
322 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
323 -ftree-vectorizer-verbose=@var{n} @gol
324 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
325 -fdump-final-insns=@var{file} @gol
326 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
327 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
328 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
329 -fenable-@var{kind}-@var{pass} @gol
330 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
331 -fdebug-types-section @gol
332 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
333 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
334 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
335 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
336 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
337 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
338 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
339 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
340 -gvms -gxcoff -gxcoff+ @gol
341 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
342 -fdebug-prefix-map=@var{old}=@var{new} @gol
343 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
344 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
345 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
346 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
347 -print-prog-name=@var{program} -print-search-dirs -Q @gol
348 -print-sysroot -print-sysroot-headers-suffix @gol
349 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
351 @item Optimization Options
352 @xref{Optimize Options,,Options that Control Optimization}.
353 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
354 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
355 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
356 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
357 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
358 -fcompare-elim -fcprop-registers -fcrossjumping @gol
359 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
360 -fcx-limited-range @gol
361 -fdata-sections -fdce -fdelayed-branch @gol
362 -fdelete-null-pointer-checks -fdevirtualize -fdse @gol
363 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
364 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
365 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
366 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
367 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
368 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
369 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
370 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
371 -fira-algorithm=@var{algorithm} @gol
372 -fira-region=@var{region} @gol
373 -fira-loop-pressure -fno-ira-share-save-slots @gol
374 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
375 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
376 -floop-block -floop-interchange -floop-strip-mine @gol
377 -floop-parallelize-all -flto -flto-compression-level @gol
378 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
379 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
380 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
381 -fno-default-inline @gol
382 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
383 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
384 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
385 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
386 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
387 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
388 -fprefetch-loop-arrays @gol
389 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
390 -fprofile-generate=@var{path} @gol
391 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
392 -freciprocal-math -free -fregmove -frename-registers -freorder-blocks @gol
393 -freorder-blocks-and-partition -freorder-functions @gol
394 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
395 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
396 -fsched-spec-load -fsched-spec-load-dangerous @gol
397 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
398 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
399 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
400 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
401 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
402 -fselective-scheduling -fselective-scheduling2 @gol
403 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
404 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
405 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
406 -fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol
407 -fthread-jumps -ftracer -ftree-bit-ccp @gol
408 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
409 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
410 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
411 -ftree-loop-if-convert-stores -ftree-loop-im @gol
412 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
413 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
414 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
416 -ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol
417 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
418 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
419 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
420 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
421 -fwhole-program -fwpa -fuse-linker-plugin @gol
422 --param @var{name}=@var{value}
423 -O -O0 -O1 -O2 -O3 -Os -Ofast}
425 @item Preprocessor Options
426 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
427 @gccoptlist{-A@var{question}=@var{answer} @gol
428 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
429 -C -dD -dI -dM -dN @gol
430 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
431 -idirafter @var{dir} @gol
432 -include @var{file} -imacros @var{file} @gol
433 -iprefix @var{file} -iwithprefix @var{dir} @gol
434 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
435 -imultilib @var{dir} -isysroot @var{dir} @gol
436 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
437 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
438 -remap -trigraphs -undef -U@var{macro} @gol
439 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
441 @item Assembler Option
442 @xref{Assembler Options,,Passing Options to the Assembler}.
443 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
446 @xref{Link Options,,Options for Linking}.
447 @gccoptlist{@var{object-file-name} -l@var{library} @gol
448 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
449 -s -static -static-libgcc -static-libstdc++ -shared @gol
450 -shared-libgcc -symbolic @gol
451 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
454 @item Directory Options
455 @xref{Directory Options,,Options for Directory Search}.
456 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
457 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
460 @item Machine Dependent Options
461 @xref{Submodel Options,,Hardware Models and Configurations}.
462 @c This list is ordered alphanumerically by subsection name.
463 @c Try and put the significant identifier (CPU or system) first,
464 @c so users have a clue at guessing where the ones they want will be.
466 @emph{Adapteva Epiphany Options}
467 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
468 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
469 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
470 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
471 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
472 -msplit-vecmove-early -m1reg-@var{reg}}
475 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
476 -mabi=@var{name} @gol
477 -mapcs-stack-check -mno-apcs-stack-check @gol
478 -mapcs-float -mno-apcs-float @gol
479 -mapcs-reentrant -mno-apcs-reentrant @gol
480 -msched-prolog -mno-sched-prolog @gol
481 -mlittle-endian -mbig-endian -mwords-little-endian @gol
482 -mfloat-abi=@var{name} -mfpe @gol
483 -mfp16-format=@var{name}
484 -mthumb-interwork -mno-thumb-interwork @gol
485 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
486 -mstructure-size-boundary=@var{n} @gol
487 -mabort-on-noreturn @gol
488 -mlong-calls -mno-long-calls @gol
489 -msingle-pic-base -mno-single-pic-base @gol
490 -mpic-register=@var{reg} @gol
491 -mnop-fun-dllimport @gol
492 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
493 -mpoke-function-name @gol
495 -mtpcs-frame -mtpcs-leaf-frame @gol
496 -mcaller-super-interworking -mcallee-super-interworking @gol
497 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
498 -mword-relocations @gol
499 -mfix-cortex-m3-ldrd}
502 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
503 -mcall-prologues -mint8 -mno-interrupts -mrelax -mshort-calls @gol
504 -mstrict-X -mtiny-stack}
506 @emph{Blackfin Options}
507 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
508 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
509 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
510 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
511 -mno-id-shared-library -mshared-library-id=@var{n} @gol
512 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
513 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
514 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
518 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
519 -msim -msdata=@var{sdata-type}}
522 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
523 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
524 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
525 -mstack-align -mdata-align -mconst-align @gol
526 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
527 -melf -maout -melinux -mlinux -sim -sim2 @gol
528 -mmul-bug-workaround -mno-mul-bug-workaround}
531 @gccoptlist{-mmac @gol
532 -mcr16cplus -mcr16c @gol
533 -msim -mint32 -mbit-ops
534 -mdata-model=@var{model}}
536 @emph{Darwin Options}
537 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
538 -arch_only -bind_at_load -bundle -bundle_loader @gol
539 -client_name -compatibility_version -current_version @gol
541 -dependency-file -dylib_file -dylinker_install_name @gol
542 -dynamic -dynamiclib -exported_symbols_list @gol
543 -filelist -flat_namespace -force_cpusubtype_ALL @gol
544 -force_flat_namespace -headerpad_max_install_names @gol
546 -image_base -init -install_name -keep_private_externs @gol
547 -multi_module -multiply_defined -multiply_defined_unused @gol
548 -noall_load -no_dead_strip_inits_and_terms @gol
549 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
550 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
551 -private_bundle -read_only_relocs -sectalign @gol
552 -sectobjectsymbols -whyload -seg1addr @gol
553 -sectcreate -sectobjectsymbols -sectorder @gol
554 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
555 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
556 -segprot -segs_read_only_addr -segs_read_write_addr @gol
557 -single_module -static -sub_library -sub_umbrella @gol
558 -twolevel_namespace -umbrella -undefined @gol
559 -unexported_symbols_list -weak_reference_mismatches @gol
560 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
561 -mkernel -mone-byte-bool}
563 @emph{DEC Alpha Options}
564 @gccoptlist{-mno-fp-regs -msoft-float @gol
565 -mieee -mieee-with-inexact -mieee-conformant @gol
566 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
567 -mtrap-precision=@var{mode} -mbuild-constants @gol
568 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
569 -mbwx -mmax -mfix -mcix @gol
570 -mfloat-vax -mfloat-ieee @gol
571 -mexplicit-relocs -msmall-data -mlarge-data @gol
572 -msmall-text -mlarge-text @gol
573 -mmemory-latency=@var{time}}
576 @gccoptlist{-msmall-model -mno-lsim}
579 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
580 -mhard-float -msoft-float @gol
581 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
582 -mdouble -mno-double @gol
583 -mmedia -mno-media -mmuladd -mno-muladd @gol
584 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
585 -mlinked-fp -mlong-calls -malign-labels @gol
586 -mlibrary-pic -macc-4 -macc-8 @gol
587 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
588 -moptimize-membar -mno-optimize-membar @gol
589 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
590 -mvliw-branch -mno-vliw-branch @gol
591 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
592 -mno-nested-cond-exec -mtomcat-stats @gol
596 @emph{GNU/Linux Options}
597 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
598 -tno-android-cc -tno-android-ld}
600 @emph{H8/300 Options}
601 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
604 @gccoptlist{-march=@var{architecture-type} @gol
605 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
606 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
607 -mfixed-range=@var{register-range} @gol
608 -mjump-in-delay -mlinker-opt -mlong-calls @gol
609 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
610 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
611 -mno-jump-in-delay -mno-long-load-store @gol
612 -mno-portable-runtime -mno-soft-float @gol
613 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
614 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
615 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
616 -munix=@var{unix-std} -nolibdld -static -threads}
618 @emph{i386 and x86-64 Options}
619 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
620 -mfpmath=@var{unit} @gol
621 -masm=@var{dialect} -mno-fancy-math-387 @gol
622 -mno-fp-ret-in-387 -msoft-float @gol
623 -mno-wide-multiply -mrtd -malign-double @gol
624 -mpreferred-stack-boundary=@var{num} @gol
625 -mincoming-stack-boundary=@var{num} @gol
626 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
627 -mrecip -mrecip=@var{opt} @gol
629 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
630 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
631 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
632 -mbmi2 -mrtm -mlwp -mthreads @gol
633 -mno-align-stringops -minline-all-stringops @gol
634 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
635 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
636 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
637 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
638 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
639 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
640 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
641 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
642 -msse2avx -mfentry -m8bit-idiv @gol
643 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
645 @emph{i386 and x86-64 Windows Options}
646 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
647 -mnop-fun-dllimport -mthread @gol
648 -municode -mwin32 -mwindows -fno-set-stack-executable}
651 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
652 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
653 -mconstant-gp -mauto-pic -mfused-madd @gol
654 -minline-float-divide-min-latency @gol
655 -minline-float-divide-max-throughput @gol
656 -mno-inline-float-divide @gol
657 -minline-int-divide-min-latency @gol
658 -minline-int-divide-max-throughput @gol
659 -mno-inline-int-divide @gol
660 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
661 -mno-inline-sqrt @gol
662 -mdwarf2-asm -mearly-stop-bits @gol
663 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
664 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
665 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
666 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
667 -msched-spec-ldc -msched-spec-control-ldc @gol
668 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
669 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
670 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
671 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
674 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
675 -msign-extend-enabled -muser-enabled}
677 @emph{M32R/D Options}
678 @gccoptlist{-m32r2 -m32rx -m32r @gol
680 -malign-loops -mno-align-loops @gol
681 -missue-rate=@var{number} @gol
682 -mbranch-cost=@var{number} @gol
683 -mmodel=@var{code-size-model-type} @gol
684 -msdata=@var{sdata-type} @gol
685 -mno-flush-func -mflush-func=@var{name} @gol
686 -mno-flush-trap -mflush-trap=@var{number} @gol
690 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
692 @emph{M680x0 Options}
693 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
694 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
695 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
696 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
697 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
698 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
699 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
700 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
704 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
705 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
706 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
707 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
708 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
711 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
712 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
713 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
714 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
717 @emph{MicroBlaze Options}
718 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
719 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
720 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
721 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
722 -mxl-mode-@var{app-model}}
725 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
726 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
727 -mips64 -mips64r2 @gol
728 -mips16 -mno-mips16 -mflip-mips16 @gol
729 -minterlink-mips16 -mno-interlink-mips16 @gol
730 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
731 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
732 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
733 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
734 -mfpu=@var{fpu-type} @gol
735 -msmartmips -mno-smartmips @gol
736 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
737 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
738 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
739 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
740 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
741 -membedded-data -mno-embedded-data @gol
742 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
743 -mcode-readable=@var{setting} @gol
744 -msplit-addresses -mno-split-addresses @gol
745 -mexplicit-relocs -mno-explicit-relocs @gol
746 -mcheck-zero-division -mno-check-zero-division @gol
747 -mdivide-traps -mdivide-breaks @gol
748 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
749 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
750 -mfix-24k -mno-fix-24k @gol
751 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
752 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
753 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
754 -mflush-func=@var{func} -mno-flush-func @gol
755 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
756 -mfp-exceptions -mno-fp-exceptions @gol
757 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
758 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
761 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
762 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
763 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
764 -mno-base-addresses -msingle-exit -mno-single-exit}
766 @emph{MN10300 Options}
767 @gccoptlist{-mmult-bug -mno-mult-bug @gol
768 -mno-am33 -mam33 -mam33-2 -mam34 @gol
769 -mtune=@var{cpu-type} @gol
770 -mreturn-pointer-on-d0 @gol
771 -mno-crt0 -mrelax -mliw -msetlb}
773 @emph{PDP-11 Options}
774 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
775 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
776 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
777 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
778 -mbranch-expensive -mbranch-cheap @gol
779 -munix-asm -mdec-asm}
781 @emph{picoChip Options}
782 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
783 -msymbol-as-address -mno-inefficient-warnings}
785 @emph{PowerPC Options}
786 See RS/6000 and PowerPC Options.
789 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78}
791 @emph{RS/6000 and PowerPC Options}
792 @gccoptlist{-mcpu=@var{cpu-type} @gol
793 -mtune=@var{cpu-type} @gol
794 -mcmodel=@var{code-model} @gol
795 -mpower -mno-power -mpower2 -mno-power2 @gol
796 -mpowerpc -mpowerpc64 -mno-powerpc @gol
797 -maltivec -mno-altivec @gol
798 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
799 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
800 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
801 -mfprnd -mno-fprnd @gol
802 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
803 -mnew-mnemonics -mold-mnemonics @gol
804 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
805 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
806 -malign-power -malign-natural @gol
807 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
808 -msingle-float -mdouble-float -msimple-fpu @gol
809 -mstring -mno-string -mupdate -mno-update @gol
810 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
811 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
812 -mstrict-align -mno-strict-align -mrelocatable @gol
813 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
814 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
815 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
816 -mprioritize-restricted-insns=@var{priority} @gol
817 -msched-costly-dep=@var{dependence_type} @gol
818 -minsert-sched-nops=@var{scheme} @gol
819 -mcall-sysv -mcall-netbsd @gol
820 -maix-struct-return -msvr4-struct-return @gol
821 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
822 -mblock-move-inline-limit=@var{num} @gol
823 -misel -mno-isel @gol
824 -misel=yes -misel=no @gol
826 -mspe=yes -mspe=no @gol
828 -mgen-cell-microcode -mwarn-cell-microcode @gol
829 -mvrsave -mno-vrsave @gol
830 -mmulhw -mno-mulhw @gol
831 -mdlmzb -mno-dlmzb @gol
832 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
833 -mprototype -mno-prototype @gol
834 -msim -mmvme -mads -myellowknife -memb -msdata @gol
835 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
836 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
837 -mno-recip-precision @gol
838 -mveclibabi=@var{type} -mfriz -mno-friz @gol
839 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
840 -msave-toc-indirect -mno-save-toc-indirect}
843 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
845 -mbig-endian-data -mlittle-endian-data @gol
848 -mas100-syntax -mno-as100-syntax@gol
850 -mmax-constant-size=@gol
853 -msave-acc-in-interrupts}
855 @emph{S/390 and zSeries Options}
856 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
857 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
858 -mlong-double-64 -mlong-double-128 @gol
859 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
860 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
861 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
862 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
863 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
866 @gccoptlist{-meb -mel @gol
870 -mscore5 -mscore5u -mscore7 -mscore7d}
873 @gccoptlist{-m1 -m2 -m2e @gol
874 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
876 -m4-nofpu -m4-single-only -m4-single -m4 @gol
877 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
878 -m5-64media -m5-64media-nofpu @gol
879 -m5-32media -m5-32media-nofpu @gol
880 -m5-compact -m5-compact-nofpu @gol
881 -mb -ml -mdalign -mrelax @gol
882 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
883 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
884 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
885 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
886 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
887 -maccumulate-outgoing-args -minvalid-symbols -msoft-atomic -mhard-atomic @gol
888 -mbranch-cost=@var{num} -mcbranchdi -mcmpeqdi -mfused-madd -mno-fused-madd @gol
889 -mfsca -mno-fsca -mfsrra -mno-fsrra -mpretend-cmove -menable-tas}
891 @emph{Solaris 2 Options}
892 @gccoptlist{-mimpure-text -mno-impure-text @gol
896 @gccoptlist{-mcpu=@var{cpu-type} @gol
897 -mtune=@var{cpu-type} @gol
898 -mcmodel=@var{code-model} @gol
899 -mmemory-model=@var{mem-model} @gol
900 -m32 -m64 -mapp-regs -mno-app-regs @gol
901 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
902 -mfpu -mno-fpu -mhard-float -msoft-float @gol
903 -mhard-quad-float -msoft-quad-float @gol
905 -mstack-bias -mno-stack-bias @gol
906 -munaligned-doubles -mno-unaligned-doubles @gol
907 -mv8plus -mno-v8plus -mvis -mno-vis @gol
908 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
909 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
913 @gccoptlist{-mwarn-reloc -merror-reloc @gol
914 -msafe-dma -munsafe-dma @gol
916 -msmall-mem -mlarge-mem -mstdmain @gol
917 -mfixed-range=@var{register-range} @gol
919 -maddress-space-conversion -mno-address-space-conversion @gol
920 -mcache-size=@var{cache-size} @gol
921 -matomic-updates -mno-atomic-updates}
923 @emph{System V Options}
924 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
926 @emph{TILE-Gx Options}
927 @gccoptlist{-mcpu=CPU -m32 -m64}
929 @emph{TILEPro Options}
930 @gccoptlist{-mcpu=CPU -m32}
933 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
934 -mprolog-function -mno-prolog-function -mspace @gol
935 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
936 -mapp-regs -mno-app-regs @gol
937 -mdisable-callt -mno-disable-callt @gol
940 -mv850e1 -mv850es @gol
945 @gccoptlist{-mg -mgnu -munix}
948 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
949 -mpointer-size=@var{size}}
951 @emph{VxWorks Options}
952 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
953 -Xbind-lazy -Xbind-now}
955 @emph{x86-64 Options}
956 See i386 and x86-64 Options.
958 @emph{Xstormy16 Options}
961 @emph{Xtensa Options}
962 @gccoptlist{-mconst16 -mno-const16 @gol
963 -mfused-madd -mno-fused-madd @gol
965 -mserialize-volatile -mno-serialize-volatile @gol
966 -mtext-section-literals -mno-text-section-literals @gol
967 -mtarget-align -mno-target-align @gol
968 -mlongcalls -mno-longcalls}
970 @emph{zSeries Options}
971 See S/390 and zSeries Options.
973 @item Code Generation Options
974 @xref{Code Gen Options,,Options for Code Generation Conventions}.
975 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
976 -ffixed-@var{reg} -fexceptions @gol
977 -fnon-call-exceptions -funwind-tables @gol
978 -fasynchronous-unwind-tables @gol
979 -finhibit-size-directive -finstrument-functions @gol
980 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
981 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
982 -fno-common -fno-ident @gol
983 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
984 -fno-jump-tables @gol
985 -frecord-gcc-switches @gol
986 -freg-struct-return -fshort-enums @gol
987 -fshort-double -fshort-wchar @gol
988 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
989 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
990 -fno-stack-limit -fsplit-stack @gol
991 -fleading-underscore -ftls-model=@var{model} @gol
992 -ftrapv -fwrapv -fbounds-check @gol
993 -fvisibility -fstrict-volatile-bitfields -fsync-libcalls}
997 * Overall Options:: Controlling the kind of output:
998 an executable, object files, assembler files,
999 or preprocessed source.
1000 * C Dialect Options:: Controlling the variant of C language compiled.
1001 * C++ Dialect Options:: Variations on C++.
1002 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
1004 * Language Independent Options:: Controlling how diagnostics should be
1006 * Warning Options:: How picky should the compiler be?
1007 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
1008 * Optimize Options:: How much optimization?
1009 * Preprocessor Options:: Controlling header files and macro definitions.
1010 Also, getting dependency information for Make.
1011 * Assembler Options:: Passing options to the assembler.
1012 * Link Options:: Specifying libraries and so on.
1013 * Directory Options:: Where to find header files and libraries.
1014 Where to find the compiler executable files.
1015 * Spec Files:: How to pass switches to sub-processes.
1016 * Target Options:: Running a cross-compiler, or an old version of GCC.
1019 @node Overall Options
1020 @section Options Controlling the Kind of Output
1022 Compilation can involve up to four stages: preprocessing, compilation
1023 proper, assembly and linking, always in that order. GCC is capable of
1024 preprocessing and compiling several files either into several
1025 assembler input files, or into one assembler input file; then each
1026 assembler input file produces an object file, and linking combines all
1027 the object files (those newly compiled, and those specified as input)
1028 into an executable file.
1030 @cindex file name suffix
1031 For any given input file, the file name suffix determines what kind of
1032 compilation is done:
1036 C source code that must be preprocessed.
1039 C source code that should not be preprocessed.
1042 C++ source code that should not be preprocessed.
1045 Objective-C source code. Note that you must link with the @file{libobjc}
1046 library to make an Objective-C program work.
1049 Objective-C source code that should not be preprocessed.
1053 Objective-C++ source code. Note that you must link with the @file{libobjc}
1054 library to make an Objective-C++ program work. Note that @samp{.M} refers
1055 to a literal capital M@.
1057 @item @var{file}.mii
1058 Objective-C++ source code that should not be preprocessed.
1061 C, C++, Objective-C or Objective-C++ header file to be turned into a
1062 precompiled header (default), or C, C++ header file to be turned into an
1063 Ada spec (via the @option{-fdump-ada-spec} switch).
1066 @itemx @var{file}.cp
1067 @itemx @var{file}.cxx
1068 @itemx @var{file}.cpp
1069 @itemx @var{file}.CPP
1070 @itemx @var{file}.c++
1072 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1073 the last two letters must both be literally @samp{x}. Likewise,
1074 @samp{.C} refers to a literal capital C@.
1078 Objective-C++ source code that must be preprocessed.
1080 @item @var{file}.mii
1081 Objective-C++ source code that should not be preprocessed.
1085 @itemx @var{file}.hp
1086 @itemx @var{file}.hxx
1087 @itemx @var{file}.hpp
1088 @itemx @var{file}.HPP
1089 @itemx @var{file}.h++
1090 @itemx @var{file}.tcc
1091 C++ header file to be turned into a precompiled header or Ada spec.
1094 @itemx @var{file}.for
1095 @itemx @var{file}.ftn
1096 Fixed form Fortran source code that should not be preprocessed.
1099 @itemx @var{file}.FOR
1100 @itemx @var{file}.fpp
1101 @itemx @var{file}.FPP
1102 @itemx @var{file}.FTN
1103 Fixed form Fortran source code that must be preprocessed (with the traditional
1106 @item @var{file}.f90
1107 @itemx @var{file}.f95
1108 @itemx @var{file}.f03
1109 @itemx @var{file}.f08
1110 Free form Fortran source code that should not be preprocessed.
1112 @item @var{file}.F90
1113 @itemx @var{file}.F95
1114 @itemx @var{file}.F03
1115 @itemx @var{file}.F08
1116 Free form Fortran source code that must be preprocessed (with the
1117 traditional preprocessor).
1122 @c FIXME: Descriptions of Java file types.
1128 @item @var{file}.ads
1129 Ada source code file that contains a library unit declaration (a
1130 declaration of a package, subprogram, or generic, or a generic
1131 instantiation), or a library unit renaming declaration (a package,
1132 generic, or subprogram renaming declaration). Such files are also
1135 @item @var{file}.adb
1136 Ada source code file containing a library unit body (a subprogram or
1137 package body). Such files are also called @dfn{bodies}.
1139 @c GCC also knows about some suffixes for languages not yet included:
1150 @itemx @var{file}.sx
1151 Assembler code that must be preprocessed.
1154 An object file to be fed straight into linking.
1155 Any file name with no recognized suffix is treated this way.
1159 You can specify the input language explicitly with the @option{-x} option:
1162 @item -x @var{language}
1163 Specify explicitly the @var{language} for the following input files
1164 (rather than letting the compiler choose a default based on the file
1165 name suffix). This option applies to all following input files until
1166 the next @option{-x} option. Possible values for @var{language} are:
1168 c c-header cpp-output
1169 c++ c++-header c++-cpp-output
1170 objective-c objective-c-header objective-c-cpp-output
1171 objective-c++ objective-c++-header objective-c++-cpp-output
1172 assembler assembler-with-cpp
1174 f77 f77-cpp-input f95 f95-cpp-input
1180 Turn off any specification of a language, so that subsequent files are
1181 handled according to their file name suffixes (as they are if @option{-x}
1182 has not been used at all).
1184 @item -pass-exit-codes
1185 @opindex pass-exit-codes
1186 Normally the @command{gcc} program exits with the code of 1 if any
1187 phase of the compiler returns a non-success return code. If you specify
1188 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1189 the numerically highest error produced by any phase returning an error
1190 indication. The C, C++, and Fortran front ends return 4 if an internal
1191 compiler error is encountered.
1194 If you only want some of the stages of compilation, you can use
1195 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1196 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1197 @command{gcc} is to stop. Note that some combinations (for example,
1198 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1203 Compile or assemble the source files, but do not link. The linking
1204 stage simply is not done. The ultimate output is in the form of an
1205 object file for each source file.
1207 By default, the object file name for a source file is made by replacing
1208 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1210 Unrecognized input files, not requiring compilation or assembly, are
1215 Stop after the stage of compilation proper; do not assemble. The output
1216 is in the form of an assembler code file for each non-assembler input
1219 By default, the assembler file name for a source file is made by
1220 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1222 Input files that don't require compilation are ignored.
1226 Stop after the preprocessing stage; do not run the compiler proper. The
1227 output is in the form of preprocessed source code, which is sent to the
1230 Input files that don't require preprocessing are ignored.
1232 @cindex output file option
1235 Place output in file @var{file}. This applies regardless to whatever
1236 sort of output is being produced, whether it be an executable file,
1237 an object file, an assembler file or preprocessed C code.
1239 If @option{-o} is not specified, the default is to put an executable
1240 file in @file{a.out}, the object file for
1241 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1242 assembler file in @file{@var{source}.s}, a precompiled header file in
1243 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1248 Print (on standard error output) the commands executed to run the stages
1249 of compilation. Also print the version number of the compiler driver
1250 program and of the preprocessor and the compiler proper.
1254 Like @option{-v} except the commands are not executed and arguments
1255 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1256 This is useful for shell scripts to capture the driver-generated command lines.
1260 Use pipes rather than temporary files for communication between the
1261 various stages of compilation. This fails to work on some systems where
1262 the assembler is unable to read from a pipe; but the GNU assembler has
1267 Print (on the standard output) a description of the command-line options
1268 understood by @command{gcc}. If the @option{-v} option is also specified
1269 then @option{--help} is also passed on to the various processes
1270 invoked by @command{gcc}, so that they can display the command-line options
1271 they accept. If the @option{-Wextra} option has also been specified
1272 (prior to the @option{--help} option), then command-line options that
1273 have no documentation associated with them are also displayed.
1276 @opindex target-help
1277 Print (on the standard output) a description of target-specific command-line
1278 options for each tool. For some targets extra target-specific
1279 information may also be printed.
1281 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1282 Print (on the standard output) a description of the command-line
1283 options understood by the compiler that fit into all specified classes
1284 and qualifiers. These are the supported classes:
1287 @item @samp{optimizers}
1288 Display all of the optimization options supported by the
1291 @item @samp{warnings}
1292 Display all of the options controlling warning messages
1293 produced by the compiler.
1296 Display target-specific options. Unlike the
1297 @option{--target-help} option however, target-specific options of the
1298 linker and assembler are not displayed. This is because those
1299 tools do not currently support the extended @option{--help=} syntax.
1302 Display the values recognized by the @option{--param}
1305 @item @var{language}
1306 Display the options supported for @var{language}, where
1307 @var{language} is the name of one of the languages supported in this
1311 Display the options that are common to all languages.
1314 These are the supported qualifiers:
1317 @item @samp{undocumented}
1318 Display only those options that are undocumented.
1321 Display options taking an argument that appears after an equal
1322 sign in the same continuous piece of text, such as:
1323 @samp{--help=target}.
1325 @item @samp{separate}
1326 Display options taking an argument that appears as a separate word
1327 following the original option, such as: @samp{-o output-file}.
1330 Thus for example to display all the undocumented target-specific
1331 switches supported by the compiler the following can be used:
1334 --help=target,undocumented
1337 The sense of a qualifier can be inverted by prefixing it with the
1338 @samp{^} character, so for example to display all binary warning
1339 options (i.e., ones that are either on or off and that do not take an
1340 argument) that have a description, use:
1343 --help=warnings,^joined,^undocumented
1346 The argument to @option{--help=} should not consist solely of inverted
1349 Combining several classes is possible, although this usually
1350 restricts the output by so much that there is nothing to display. One
1351 case where it does work however is when one of the classes is
1352 @var{target}. So for example to display all the target-specific
1353 optimization options the following can be used:
1356 --help=target,optimizers
1359 The @option{--help=} option can be repeated on the command line. Each
1360 successive use displays its requested class of options, skipping
1361 those that have already been displayed.
1363 If the @option{-Q} option appears on the command line before the
1364 @option{--help=} option, then the descriptive text displayed by
1365 @option{--help=} is changed. Instead of describing the displayed
1366 options, an indication is given as to whether the option is enabled,
1367 disabled or set to a specific value (assuming that the compiler
1368 knows this at the point where the @option{--help=} option is used).
1370 Here is a truncated example from the ARM port of @command{gcc}:
1373 % gcc -Q -mabi=2 --help=target -c
1374 The following options are target specific:
1376 -mabort-on-noreturn [disabled]
1380 The output is sensitive to the effects of previous command-line
1381 options, so for example it is possible to find out which optimizations
1382 are enabled at @option{-O2} by using:
1385 -Q -O2 --help=optimizers
1388 Alternatively you can discover which binary optimizations are enabled
1389 by @option{-O3} by using:
1392 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1393 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1394 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1397 @item -no-canonical-prefixes
1398 @opindex no-canonical-prefixes
1399 Do not expand any symbolic links, resolve references to @samp{/../}
1400 or @samp{/./}, or make the path absolute when generating a relative
1405 Display the version number and copyrights of the invoked GCC@.
1409 Invoke all subcommands under a wrapper program. The name of the
1410 wrapper program and its parameters are passed as a comma separated
1414 gcc -c t.c -wrapper gdb,--args
1418 This invokes all subprograms of @command{gcc} under
1419 @samp{gdb --args}, thus the invocation of @command{cc1} is
1420 @samp{gdb --args cc1 @dots{}}.
1422 @item -fplugin=@var{name}.so
1423 Load the plugin code in file @var{name}.so, assumed to be a
1424 shared object to be dlopen'd by the compiler. The base name of
1425 the shared object file is used to identify the plugin for the
1426 purposes of argument parsing (See
1427 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1428 Each plugin should define the callback functions specified in the
1431 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1432 Define an argument called @var{key} with a value of @var{value}
1433 for the plugin called @var{name}.
1435 @item -fdump-ada-spec@r{[}-slim@r{]}
1436 For C and C++ source and include files, generate corresponding Ada
1437 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1438 GNAT User's Guide}, which provides detailed documentation on this feature.
1440 @item -fdump-go-spec=@var{file}
1441 For input files in any language, generate corresponding Go
1442 declarations in @var{file}. This generates Go @code{const},
1443 @code{type}, @code{var}, and @code{func} declarations which may be a
1444 useful way to start writing a Go interface to code written in some
1447 @include @value{srcdir}/../libiberty/at-file.texi
1451 @section Compiling C++ Programs
1453 @cindex suffixes for C++ source
1454 @cindex C++ source file suffixes
1455 C++ source files conventionally use one of the suffixes @samp{.C},
1456 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1457 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1458 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1459 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1460 files with these names and compiles them as C++ programs even if you
1461 call the compiler the same way as for compiling C programs (usually
1462 with the name @command{gcc}).
1466 However, the use of @command{gcc} does not add the C++ library.
1467 @command{g++} is a program that calls GCC and treats @samp{.c},
1468 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1469 files unless @option{-x} is used, and automatically specifies linking
1470 against the C++ library. This program is also useful when
1471 precompiling a C header file with a @samp{.h} extension for use in C++
1472 compilations. On many systems, @command{g++} is also installed with
1473 the name @command{c++}.
1475 @cindex invoking @command{g++}
1476 When you compile C++ programs, you may specify many of the same
1477 command-line options that you use for compiling programs in any
1478 language; or command-line options meaningful for C and related
1479 languages; or options that are meaningful only for C++ programs.
1480 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1481 explanations of options for languages related to C@.
1482 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1483 explanations of options that are meaningful only for C++ programs.
1485 @node C Dialect Options
1486 @section Options Controlling C Dialect
1487 @cindex dialect options
1488 @cindex language dialect options
1489 @cindex options, dialect
1491 The following options control the dialect of C (or languages derived
1492 from C, such as C++, Objective-C and Objective-C++) that the compiler
1496 @cindex ANSI support
1500 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1501 equivalent to @option{-std=c++98}.
1503 This turns off certain features of GCC that are incompatible with ISO
1504 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1505 such as the @code{asm} and @code{typeof} keywords, and
1506 predefined macros such as @code{unix} and @code{vax} that identify the
1507 type of system you are using. It also enables the undesirable and
1508 rarely used ISO trigraph feature. For the C compiler,
1509 it disables recognition of C++ style @samp{//} comments as well as
1510 the @code{inline} keyword.
1512 The alternate keywords @code{__asm__}, @code{__extension__},
1513 @code{__inline__} and @code{__typeof__} continue to work despite
1514 @option{-ansi}. You would not want to use them in an ISO C program, of
1515 course, but it is useful to put them in header files that might be included
1516 in compilations done with @option{-ansi}. Alternate predefined macros
1517 such as @code{__unix__} and @code{__vax__} are also available, with or
1518 without @option{-ansi}.
1520 The @option{-ansi} option does not cause non-ISO programs to be
1521 rejected gratuitously. For that, @option{-Wpedantic} is required in
1522 addition to @option{-ansi}. @xref{Warning Options}.
1524 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1525 option is used. Some header files may notice this macro and refrain
1526 from declaring certain functions or defining certain macros that the
1527 ISO standard doesn't call for; this is to avoid interfering with any
1528 programs that might use these names for other things.
1530 Functions that are normally built in but do not have semantics
1531 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1532 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1533 built-in functions provided by GCC}, for details of the functions
1538 Determine the language standard. @xref{Standards,,Language Standards
1539 Supported by GCC}, for details of these standard versions. This option
1540 is currently only supported when compiling C or C++.
1542 The compiler can accept several base standards, such as @samp{c90} or
1543 @samp{c++98}, and GNU dialects of those standards, such as
1544 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1545 compiler accepts all programs following that standard plus those
1546 using GNU extensions that do not contradict it. For example,
1547 @option{-std=c90} turns off certain features of GCC that are
1548 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1549 keywords, but not other GNU extensions that do not have a meaning in
1550 ISO C90, such as omitting the middle term of a @code{?:}
1551 expression. On the other hand, by specifying a GNU dialect of a
1552 standard, all features the compiler support are enabled, even when
1553 those features change the meaning of the base standard and some
1554 strict-conforming programs may be rejected. The particular standard
1555 is used by @option{-Wpedantic} to identify which features are GNU
1556 extensions given that version of the standard. For example
1557 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1558 comments, while @option{-std=gnu99 -Wpedantic} does not.
1560 A value for this option must be provided; possible values are
1566 Support all ISO C90 programs (certain GNU extensions that conflict
1567 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1569 @item iso9899:199409
1570 ISO C90 as modified in amendment 1.
1576 ISO C99. Note that this standard is not yet fully supported; see
1577 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1578 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1583 ISO C11, the 2011 revision of the ISO C standard.
1584 Support is incomplete and experimental. The name @samp{c1x} is
1589 GNU dialect of ISO C90 (including some C99 features). This
1590 is the default for C code.
1594 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1595 this will become the default. The name @samp{gnu9x} is deprecated.
1599 GNU dialect of ISO C11. Support is incomplete and experimental. The
1600 name @samp{gnu1x} is deprecated.
1604 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1605 additional defect reports. Same as @option{-ansi} for C++ code.
1609 GNU dialect of @option{-std=c++98}. This is the default for
1614 The 2011 ISO C++ standard plus amendments. Support for C++11 is still
1615 experimental, and may change in incompatible ways in future releases.
1616 The name @samp{c++0x} is deprecated.
1620 GNU dialect of @option{-std=c++11}. Support for C++11 is still
1621 experimental, and may change in incompatible ways in future releases.
1622 The name @samp{gnu++0x} is deprecated.
1625 The next revision of the ISO C++ standard, tentatively planned for
1626 2017. Support is highly experimental, and will almost certainly
1627 change in incompatible ways in future releases.
1630 GNU dialect of @option{-std=c++1y}. Support is highly experimental,
1631 and will almost certainly change in incompatible ways in future
1635 @item -fgnu89-inline
1636 @opindex fgnu89-inline
1637 The option @option{-fgnu89-inline} tells GCC to use the traditional
1638 GNU semantics for @code{inline} functions when in C99 mode.
1639 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1640 is accepted and ignored by GCC versions 4.1.3 up to but not including
1641 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1642 C99 mode. Using this option is roughly equivalent to adding the
1643 @code{gnu_inline} function attribute to all inline functions
1644 (@pxref{Function Attributes}).
1646 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1647 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1648 specifies the default behavior). This option was first supported in
1649 GCC 4.3. This option is not supported in @option{-std=c90} or
1650 @option{-std=gnu90} mode.
1652 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1653 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1654 in effect for @code{inline} functions. @xref{Common Predefined
1655 Macros,,,cpp,The C Preprocessor}.
1657 @item -aux-info @var{filename}
1659 Output to the given filename prototyped declarations for all functions
1660 declared and/or defined in a translation unit, including those in header
1661 files. This option is silently ignored in any language other than C@.
1663 Besides declarations, the file indicates, in comments, the origin of
1664 each declaration (source file and line), whether the declaration was
1665 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1666 @samp{O} for old, respectively, in the first character after the line
1667 number and the colon), and whether it came from a declaration or a
1668 definition (@samp{C} or @samp{F}, respectively, in the following
1669 character). In the case of function definitions, a K&R-style list of
1670 arguments followed by their declarations is also provided, inside
1671 comments, after the declaration.
1673 @item -fallow-parameterless-variadic-functions
1674 Accept variadic functions without named parameters.
1676 Although it is possible to define such a function, this is not very
1677 useful as it is not possible to read the arguments. This is only
1678 supported for C as this construct is allowed by C++.
1682 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1683 keyword, so that code can use these words as identifiers. You can use
1684 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1685 instead. @option{-ansi} implies @option{-fno-asm}.
1687 In C++, this switch only affects the @code{typeof} keyword, since
1688 @code{asm} and @code{inline} are standard keywords. You may want to
1689 use the @option{-fno-gnu-keywords} flag instead, which has the same
1690 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1691 switch only affects the @code{asm} and @code{typeof} keywords, since
1692 @code{inline} is a standard keyword in ISO C99.
1695 @itemx -fno-builtin-@var{function}
1696 @opindex fno-builtin
1697 @cindex built-in functions
1698 Don't recognize built-in functions that do not begin with
1699 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1700 functions provided by GCC}, for details of the functions affected,
1701 including those which are not built-in functions when @option{-ansi} or
1702 @option{-std} options for strict ISO C conformance are used because they
1703 do not have an ISO standard meaning.
1705 GCC normally generates special code to handle certain built-in functions
1706 more efficiently; for instance, calls to @code{alloca} may become single
1707 instructions which adjust the stack directly, and calls to @code{memcpy}
1708 may become inline copy loops. The resulting code is often both smaller
1709 and faster, but since the function calls no longer appear as such, you
1710 cannot set a breakpoint on those calls, nor can you change the behavior
1711 of the functions by linking with a different library. In addition,
1712 when a function is recognized as a built-in function, GCC may use
1713 information about that function to warn about problems with calls to
1714 that function, or to generate more efficient code, even if the
1715 resulting code still contains calls to that function. For example,
1716 warnings are given with @option{-Wformat} for bad calls to
1717 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1718 known not to modify global memory.
1720 With the @option{-fno-builtin-@var{function}} option
1721 only the built-in function @var{function} is
1722 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1723 function is named that is not built-in in this version of GCC, this
1724 option is ignored. There is no corresponding
1725 @option{-fbuiltin-@var{function}} option; if you wish to enable
1726 built-in functions selectively when using @option{-fno-builtin} or
1727 @option{-ffreestanding}, you may define macros such as:
1730 #define abs(n) __builtin_abs ((n))
1731 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1736 @cindex hosted environment
1738 Assert that compilation takes place in a hosted environment. This implies
1739 @option{-fbuiltin}. A hosted environment is one in which the
1740 entire standard library is available, and in which @code{main} has a return
1741 type of @code{int}. Examples are nearly everything except a kernel.
1742 This is equivalent to @option{-fno-freestanding}.
1744 @item -ffreestanding
1745 @opindex ffreestanding
1746 @cindex hosted environment
1748 Assert that compilation takes place in a freestanding environment. This
1749 implies @option{-fno-builtin}. A freestanding environment
1750 is one in which the standard library may not exist, and program startup may
1751 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1752 This is equivalent to @option{-fno-hosted}.
1754 @xref{Standards,,Language Standards Supported by GCC}, for details of
1755 freestanding and hosted environments.
1759 @cindex OpenMP parallel
1760 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1761 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1762 compiler generates parallel code according to the OpenMP Application
1763 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1764 implies @option{-pthread}, and thus is only supported on targets that
1765 have support for @option{-pthread}.
1769 When the option @option{-fgnu-tm} is specified, the compiler
1770 generates code for the Linux variant of Intel's current Transactional
1771 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1772 an experimental feature whose interface may change in future versions
1773 of GCC, as the official specification changes. Please note that not
1774 all architectures are supported for this feature.
1776 For more information on GCC's support for transactional memory,
1777 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1778 Transactional Memory Library}.
1780 Note that the transactional memory feature is not supported with
1781 non-call exceptions (@option{-fnon-call-exceptions}).
1783 @item -fms-extensions
1784 @opindex fms-extensions
1785 Accept some non-standard constructs used in Microsoft header files.
1787 In C++ code, this allows member names in structures to be similar
1788 to previous types declarations.
1797 Some cases of unnamed fields in structures and unions are only
1798 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1799 fields within structs/unions}, for details.
1801 @item -fplan9-extensions
1802 Accept some non-standard constructs used in Plan 9 code.
1804 This enables @option{-fms-extensions}, permits passing pointers to
1805 structures with anonymous fields to functions that expect pointers to
1806 elements of the type of the field, and permits referring to anonymous
1807 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1808 struct/union fields within structs/unions}, for details. This is only
1809 supported for C, not C++.
1813 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1814 options for strict ISO C conformance) implies @option{-trigraphs}.
1816 @cindex traditional C language
1817 @cindex C language, traditional
1819 @itemx -traditional-cpp
1820 @opindex traditional-cpp
1821 @opindex traditional
1822 Formerly, these options caused GCC to attempt to emulate a pre-standard
1823 C compiler. They are now only supported with the @option{-E} switch.
1824 The preprocessor continues to support a pre-standard mode. See the GNU
1825 CPP manual for details.
1827 @item -fcond-mismatch
1828 @opindex fcond-mismatch
1829 Allow conditional expressions with mismatched types in the second and
1830 third arguments. The value of such an expression is void. This option
1831 is not supported for C++.
1833 @item -flax-vector-conversions
1834 @opindex flax-vector-conversions
1835 Allow implicit conversions between vectors with differing numbers of
1836 elements and/or incompatible element types. This option should not be
1839 @item -funsigned-char
1840 @opindex funsigned-char
1841 Let the type @code{char} be unsigned, like @code{unsigned char}.
1843 Each kind of machine has a default for what @code{char} should
1844 be. It is either like @code{unsigned char} by default or like
1845 @code{signed char} by default.
1847 Ideally, a portable program should always use @code{signed char} or
1848 @code{unsigned char} when it depends on the signedness of an object.
1849 But many programs have been written to use plain @code{char} and
1850 expect it to be signed, or expect it to be unsigned, depending on the
1851 machines they were written for. This option, and its inverse, let you
1852 make such a program work with the opposite default.
1854 The type @code{char} is always a distinct type from each of
1855 @code{signed char} or @code{unsigned char}, even though its behavior
1856 is always just like one of those two.
1859 @opindex fsigned-char
1860 Let the type @code{char} be signed, like @code{signed char}.
1862 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1863 the negative form of @option{-funsigned-char}. Likewise, the option
1864 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1866 @item -fsigned-bitfields
1867 @itemx -funsigned-bitfields
1868 @itemx -fno-signed-bitfields
1869 @itemx -fno-unsigned-bitfields
1870 @opindex fsigned-bitfields
1871 @opindex funsigned-bitfields
1872 @opindex fno-signed-bitfields
1873 @opindex fno-unsigned-bitfields
1874 These options control whether a bit-field is signed or unsigned, when the
1875 declaration does not use either @code{signed} or @code{unsigned}. By
1876 default, such a bit-field is signed, because this is consistent: the
1877 basic integer types such as @code{int} are signed types.
1880 @node C++ Dialect Options
1881 @section Options Controlling C++ Dialect
1883 @cindex compiler options, C++
1884 @cindex C++ options, command-line
1885 @cindex options, C++
1886 This section describes the command-line options that are only meaningful
1887 for C++ programs; but you can also use most of the GNU compiler options
1888 regardless of what language your program is in. For example, you
1889 might compile a file @code{firstClass.C} like this:
1892 g++ -g -frepo -O -c firstClass.C
1896 In this example, only @option{-frepo} is an option meant
1897 only for C++ programs; you can use the other options with any
1898 language supported by GCC@.
1900 Here is a list of options that are @emph{only} for compiling C++ programs:
1904 @item -fabi-version=@var{n}
1905 @opindex fabi-version
1906 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1907 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1908 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1909 the version that conforms most closely to the C++ ABI specification.
1910 Therefore, the ABI obtained using version 0 will change as ABI bugs
1913 The default is version 2.
1915 Version 3 corrects an error in mangling a constant address as a
1918 Version 4, which first appeared in G++ 4.5, implements a standard
1919 mangling for vector types.
1921 Version 5, which first appeared in G++ 4.6, corrects the mangling of
1922 attribute const/volatile on function pointer types, decltype of a
1923 plain decl, and use of a function parameter in the declaration of
1926 Version 6, which first appeared in G++ 4.7, corrects the promotion
1927 behavior of C++11 scoped enums and the mangling of template argument
1928 packs, const/static_cast, prefix ++ and --, and a class scope function
1929 used as a template argument.
1931 See also @option{-Wabi}.
1933 @item -fno-access-control
1934 @opindex fno-access-control
1935 Turn off all access checking. This switch is mainly useful for working
1936 around bugs in the access control code.
1940 Check that the pointer returned by @code{operator new} is non-null
1941 before attempting to modify the storage allocated. This check is
1942 normally unnecessary because the C++ standard specifies that
1943 @code{operator new} only returns @code{0} if it is declared
1944 @samp{throw()}, in which case the compiler always checks the
1945 return value even without this option. In all other cases, when
1946 @code{operator new} has a non-empty exception specification, memory
1947 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1948 @samp{new (nothrow)}.
1950 @item -fconserve-space
1951 @opindex fconserve-space
1952 Put uninitialized or run-time-initialized global variables into the
1953 common segment, as C does. This saves space in the executable at the
1954 cost of not diagnosing duplicate definitions. If you compile with this
1955 flag and your program mysteriously crashes after @code{main()} has
1956 completed, you may have an object that is being destroyed twice because
1957 two definitions are merged.
1959 This option is no longer useful on most targets, now that support has
1960 been added for putting variables into BSS without making them common.
1962 @item -fconstexpr-depth=@var{n}
1963 @opindex fconstexpr-depth
1964 Set the maximum nested evaluation depth for C++11 constexpr functions
1965 to @var{n}. A limit is needed to detect endless recursion during
1966 constant expression evaluation. The minimum specified by the standard
1969 @item -fdeduce-init-list
1970 @opindex fdeduce-init-list
1971 Enable deduction of a template type parameter as
1972 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
1975 template <class T> auto forward(T t) -> decltype (realfn (t))
1982 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1986 This deduction was implemented as a possible extension to the
1987 originally proposed semantics for the C++11 standard, but was not part
1988 of the final standard, so it is disabled by default. This option is
1989 deprecated, and may be removed in a future version of G++.
1991 @item -ffriend-injection
1992 @opindex ffriend-injection
1993 Inject friend functions into the enclosing namespace, so that they are
1994 visible outside the scope of the class in which they are declared.
1995 Friend functions were documented to work this way in the old Annotated
1996 C++ Reference Manual, and versions of G++ before 4.1 always worked
1997 that way. However, in ISO C++ a friend function that is not declared
1998 in an enclosing scope can only be found using argument dependent
1999 lookup. This option causes friends to be injected as they were in
2002 This option is for compatibility, and may be removed in a future
2005 @item -fno-elide-constructors
2006 @opindex fno-elide-constructors
2007 The C++ standard allows an implementation to omit creating a temporary
2008 that is only used to initialize another object of the same type.
2009 Specifying this option disables that optimization, and forces G++ to
2010 call the copy constructor in all cases.
2012 @item -fno-enforce-eh-specs
2013 @opindex fno-enforce-eh-specs
2014 Don't generate code to check for violation of exception specifications
2015 at run time. This option violates the C++ standard, but may be useful
2016 for reducing code size in production builds, much like defining
2017 @samp{NDEBUG}. This does not give user code permission to throw
2018 exceptions in violation of the exception specifications; the compiler
2019 still optimizes based on the specifications, so throwing an
2020 unexpected exception results in undefined behavior at run time.
2023 @itemx -fno-for-scope
2025 @opindex fno-for-scope
2026 If @option{-ffor-scope} is specified, the scope of variables declared in
2027 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2028 as specified by the C++ standard.
2029 If @option{-fno-for-scope} is specified, the scope of variables declared in
2030 a @i{for-init-statement} extends to the end of the enclosing scope,
2031 as was the case in old versions of G++, and other (traditional)
2032 implementations of C++.
2034 The default if neither flag is given to follow the standard,
2035 but to allow and give a warning for old-style code that would
2036 otherwise be invalid, or have different behavior.
2038 @item -fno-gnu-keywords
2039 @opindex fno-gnu-keywords
2040 Do not recognize @code{typeof} as a keyword, so that code can use this
2041 word as an identifier. You can use the keyword @code{__typeof__} instead.
2042 @option{-ansi} implies @option{-fno-gnu-keywords}.
2044 @item -fno-implicit-templates
2045 @opindex fno-implicit-templates
2046 Never emit code for non-inline templates that are instantiated
2047 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2048 @xref{Template Instantiation}, for more information.
2050 @item -fno-implicit-inline-templates
2051 @opindex fno-implicit-inline-templates
2052 Don't emit code for implicit instantiations of inline templates, either.
2053 The default is to handle inlines differently so that compiles with and
2054 without optimization need the same set of explicit instantiations.
2056 @item -fno-implement-inlines
2057 @opindex fno-implement-inlines
2058 To save space, do not emit out-of-line copies of inline functions
2059 controlled by @samp{#pragma implementation}. This causes linker
2060 errors if these functions are not inlined everywhere they are called.
2062 @item -fms-extensions
2063 @opindex fms-extensions
2064 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2065 int and getting a pointer to member function via non-standard syntax.
2067 @item -fno-nonansi-builtins
2068 @opindex fno-nonansi-builtins
2069 Disable built-in declarations of functions that are not mandated by
2070 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2071 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2074 @opindex fnothrow-opt
2075 Treat a @code{throw()} exception specification as if it were a
2076 @code{noexcept} specification to reduce or eliminate the text size
2077 overhead relative to a function with no exception specification. If
2078 the function has local variables of types with non-trivial
2079 destructors, the exception specification actually makes the
2080 function smaller because the EH cleanups for those variables can be
2081 optimized away. The semantic effect is that an exception thrown out of
2082 a function with such an exception specification results in a call
2083 to @code{terminate} rather than @code{unexpected}.
2085 @item -fno-operator-names
2086 @opindex fno-operator-names
2087 Do not treat the operator name keywords @code{and}, @code{bitand},
2088 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2089 synonyms as keywords.
2091 @item -fno-optional-diags
2092 @opindex fno-optional-diags
2093 Disable diagnostics that the standard says a compiler does not need to
2094 issue. Currently, the only such diagnostic issued by G++ is the one for
2095 a name having multiple meanings within a class.
2098 @opindex fpermissive
2099 Downgrade some diagnostics about nonconformant code from errors to
2100 warnings. Thus, using @option{-fpermissive} allows some
2101 nonconforming code to compile.
2103 @item -fno-pretty-templates
2104 @opindex fno-pretty-templates
2105 When an error message refers to a specialization of a function
2106 template, the compiler normally prints the signature of the
2107 template followed by the template arguments and any typedefs or
2108 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2109 rather than @code{void f(int)}) so that it's clear which template is
2110 involved. When an error message refers to a specialization of a class
2111 template, the compiler omits any template arguments that match
2112 the default template arguments for that template. If either of these
2113 behaviors make it harder to understand the error message rather than
2114 easier, you can use @option{-fno-pretty-templates} to disable them.
2118 Enable automatic template instantiation at link time. This option also
2119 implies @option{-fno-implicit-templates}. @xref{Template
2120 Instantiation}, for more information.
2124 Disable generation of information about every class with virtual
2125 functions for use by the C++ run-time type identification features
2126 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2127 of the language, you can save some space by using this flag. Note that
2128 exception handling uses the same information, but G++ generates it as
2129 needed. The @samp{dynamic_cast} operator can still be used for casts that
2130 do not require run-time type information, i.e.@: casts to @code{void *} or to
2131 unambiguous base classes.
2135 Emit statistics about front-end processing at the end of the compilation.
2136 This information is generally only useful to the G++ development team.
2138 @item -fstrict-enums
2139 @opindex fstrict-enums
2140 Allow the compiler to optimize using the assumption that a value of
2141 enumerated type can only be one of the values of the enumeration (as
2142 defined in the C++ standard; basically, a value that can be
2143 represented in the minimum number of bits needed to represent all the
2144 enumerators). This assumption may not be valid if the program uses a
2145 cast to convert an arbitrary integer value to the enumerated type.
2147 @item -ftemplate-backtrace-limit=@var{n}
2148 @opindex ftemplate-backtrace-limit
2149 Set the maximum number of template instantiation notes for a single
2150 warning or error to @var{n}. The default value is 10.
2152 @item -ftemplate-depth=@var{n}
2153 @opindex ftemplate-depth
2154 Set the maximum instantiation depth for template classes to @var{n}.
2155 A limit on the template instantiation depth is needed to detect
2156 endless recursions during template class instantiation. ANSI/ISO C++
2157 conforming programs must not rely on a maximum depth greater than 17
2158 (changed to 1024 in C++11). The default value is 900, as the compiler
2159 can run out of stack space before hitting 1024 in some situations.
2161 @item -fno-threadsafe-statics
2162 @opindex fno-threadsafe-statics
2163 Do not emit the extra code to use the routines specified in the C++
2164 ABI for thread-safe initialization of local statics. You can use this
2165 option to reduce code size slightly in code that doesn't need to be
2168 @item -fuse-cxa-atexit
2169 @opindex fuse-cxa-atexit
2170 Register destructors for objects with static storage duration with the
2171 @code{__cxa_atexit} function rather than the @code{atexit} function.
2172 This option is required for fully standards-compliant handling of static
2173 destructors, but only works if your C library supports
2174 @code{__cxa_atexit}.
2176 @item -fno-use-cxa-get-exception-ptr
2177 @opindex fno-use-cxa-get-exception-ptr
2178 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2179 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2180 if the runtime routine is not available.
2182 @item -fvisibility-inlines-hidden
2183 @opindex fvisibility-inlines-hidden
2184 This switch declares that the user does not attempt to compare
2185 pointers to inline functions or methods where the addresses of the two functions
2186 are taken in different shared objects.
2188 The effect of this is that GCC may, effectively, mark inline methods with
2189 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2190 appear in the export table of a DSO and do not require a PLT indirection
2191 when used within the DSO@. Enabling this option can have a dramatic effect
2192 on load and link times of a DSO as it massively reduces the size of the
2193 dynamic export table when the library makes heavy use of templates.
2195 The behavior of this switch is not quite the same as marking the
2196 methods as hidden directly, because it does not affect static variables
2197 local to the function or cause the compiler to deduce that
2198 the function is defined in only one shared object.
2200 You may mark a method as having a visibility explicitly to negate the
2201 effect of the switch for that method. For example, if you do want to
2202 compare pointers to a particular inline method, you might mark it as
2203 having default visibility. Marking the enclosing class with explicit
2204 visibility has no effect.
2206 Explicitly instantiated inline methods are unaffected by this option
2207 as their linkage might otherwise cross a shared library boundary.
2208 @xref{Template Instantiation}.
2210 @item -fvisibility-ms-compat
2211 @opindex fvisibility-ms-compat
2212 This flag attempts to use visibility settings to make GCC's C++
2213 linkage model compatible with that of Microsoft Visual Studio.
2215 The flag makes these changes to GCC's linkage model:
2219 It sets the default visibility to @code{hidden}, like
2220 @option{-fvisibility=hidden}.
2223 Types, but not their members, are not hidden by default.
2226 The One Definition Rule is relaxed for types without explicit
2227 visibility specifications that are defined in more than one
2228 shared object: those declarations are permitted if they are
2229 permitted when this option is not used.
2232 In new code it is better to use @option{-fvisibility=hidden} and
2233 export those classes that are intended to be externally visible.
2234 Unfortunately it is possible for code to rely, perhaps accidentally,
2235 on the Visual Studio behavior.
2237 Among the consequences of these changes are that static data members
2238 of the same type with the same name but defined in different shared
2239 objects are different, so changing one does not change the other;
2240 and that pointers to function members defined in different shared
2241 objects may not compare equal. When this flag is given, it is a
2242 violation of the ODR to define types with the same name differently.
2246 Do not use weak symbol support, even if it is provided by the linker.
2247 By default, G++ uses weak symbols if they are available. This
2248 option exists only for testing, and should not be used by end-users;
2249 it results in inferior code and has no benefits. This option may
2250 be removed in a future release of G++.
2254 Do not search for header files in the standard directories specific to
2255 C++, but do still search the other standard directories. (This option
2256 is used when building the C++ library.)
2259 In addition, these optimization, warning, and code generation options
2260 have meanings only for C++ programs:
2263 @item -fno-default-inline
2264 @opindex fno-default-inline
2265 Do not assume @samp{inline} for functions defined inside a class scope.
2266 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2267 functions have linkage like inline functions; they just aren't
2270 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2273 Warn when G++ generates code that is probably not compatible with the
2274 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2275 all such cases, there are probably some cases that are not warned about,
2276 even though G++ is generating incompatible code. There may also be
2277 cases where warnings are emitted even though the code that is generated
2280 You should rewrite your code to avoid these warnings if you are
2281 concerned about the fact that code generated by G++ may not be binary
2282 compatible with code generated by other compilers.
2284 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2289 A template with a non-type template parameter of reference type is
2290 mangled incorrectly:
2293 template <int &> struct S @{@};
2297 This is fixed in @option{-fabi-version=3}.
2300 SIMD vector types declared using @code{__attribute ((vector_size))} are
2301 mangled in a non-standard way that does not allow for overloading of
2302 functions taking vectors of different sizes.
2304 The mangling is changed in @option{-fabi-version=4}.
2307 The known incompatibilities in @option{-fabi-version=1} include:
2312 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2313 pack data into the same byte as a base class. For example:
2316 struct A @{ virtual void f(); int f1 : 1; @};
2317 struct B : public A @{ int f2 : 1; @};
2321 In this case, G++ places @code{B::f2} into the same byte
2322 as@code{A::f1}; other compilers do not. You can avoid this problem
2323 by explicitly padding @code{A} so that its size is a multiple of the
2324 byte size on your platform; that causes G++ and other compilers to
2325 lay out @code{B} identically.
2328 Incorrect handling of tail-padding for virtual bases. G++ does not use
2329 tail padding when laying out virtual bases. For example:
2332 struct A @{ virtual void f(); char c1; @};
2333 struct B @{ B(); char c2; @};
2334 struct C : public A, public virtual B @{@};
2338 In this case, G++ does not place @code{B} into the tail-padding for
2339 @code{A}; other compilers do. You can avoid this problem by
2340 explicitly padding @code{A} so that its size is a multiple of its
2341 alignment (ignoring virtual base classes); that causes G++ and other
2342 compilers to lay out @code{C} identically.
2345 Incorrect handling of bit-fields with declared widths greater than that
2346 of their underlying types, when the bit-fields appear in a union. For
2350 union U @{ int i : 4096; @};
2354 Assuming that an @code{int} does not have 4096 bits, G++ makes the
2355 union too small by the number of bits in an @code{int}.
2358 Empty classes can be placed at incorrect offsets. For example:
2368 struct C : public B, public A @{@};
2372 G++ places the @code{A} base class of @code{C} at a nonzero offset;
2373 it should be placed at offset zero. G++ mistakenly believes that the
2374 @code{A} data member of @code{B} is already at offset zero.
2377 Names of template functions whose types involve @code{typename} or
2378 template template parameters can be mangled incorrectly.
2381 template <typename Q>
2382 void f(typename Q::X) @{@}
2384 template <template <typename> class Q>
2385 void f(typename Q<int>::X) @{@}
2389 Instantiations of these templates may be mangled incorrectly.
2393 It also warns about psABI-related changes. The known psABI changes at this
2399 For SysV/x86-64, unions with @code{long double} members are
2400 passed in memory as specified in psABI. For example:
2410 @code{union U} is always passed in memory.
2414 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2415 @opindex Wctor-dtor-privacy
2416 @opindex Wno-ctor-dtor-privacy
2417 Warn when a class seems unusable because all the constructors or
2418 destructors in that class are private, and it has neither friends nor
2419 public static member functions.
2421 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2422 @opindex Wdelete-non-virtual-dtor
2423 @opindex Wno-delete-non-virtual-dtor
2424 Warn when @samp{delete} is used to destroy an instance of a class that
2425 has virtual functions and non-virtual destructor. It is unsafe to delete
2426 an instance of a derived class through a pointer to a base class if the
2427 base class does not have a virtual destructor. This warning is enabled
2430 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2431 @opindex Wliteral-suffix
2432 @opindex Wno-literal-suffix
2433 Warn when a string or character literal is followed by a ud-suffix which does
2434 not begin with an underscore. As a conforming extension, GCC treats such
2435 suffixes as separate preprocessing tokens in order to maintain backwards
2436 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2440 #define __STDC_FORMAT_MACROS
2441 #include <inttypes.h>
2446 printf("My int64: %"PRId64"\n", i64);
2450 In this case, @code{PRId64} is treated as a separate preprocessing token.
2452 This warning is enabled by default.
2454 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2456 @opindex Wno-narrowing
2457 Warn when a narrowing conversion prohibited by C++11 occurs within
2461 int i = @{ 2.2 @}; // error: narrowing from double to int
2464 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2466 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2467 required by the standard. Note that this does not affect the meaning
2468 of well-formed code; narrowing conversions are still considered
2469 ill-formed in SFINAE context.
2471 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2473 @opindex Wno-noexcept
2474 Warn when a noexcept-expression evaluates to false because of a call
2475 to a function that does not have a non-throwing exception
2476 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2477 the compiler to never throw an exception.
2479 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2480 @opindex Wnon-virtual-dtor
2481 @opindex Wno-non-virtual-dtor
2482 Warn when a class has virtual functions and an accessible non-virtual
2483 destructor, in which case it is possible but unsafe to delete
2484 an instance of a derived class through a pointer to the base class.
2485 This warning is also enabled if @option{-Weffc++} is specified.
2487 @item -Wreorder @r{(C++ and Objective-C++ only)}
2489 @opindex Wno-reorder
2490 @cindex reordering, warning
2491 @cindex warning for reordering of member initializers
2492 Warn when the order of member initializers given in the code does not
2493 match the order in which they must be executed. For instance:
2499 A(): j (0), i (1) @{ @}
2504 The compiler rearranges the member initializers for @samp{i}
2505 and @samp{j} to match the declaration order of the members, emitting
2506 a warning to that effect. This warning is enabled by @option{-Wall}.
2509 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2512 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2515 Warn about violations of the following style guidelines from Scott Meyers'
2516 @cite{Effective C++, Second Edition} book:
2520 Item 11: Define a copy constructor and an assignment operator for classes
2521 with dynamically allocated memory.
2524 Item 12: Prefer initialization to assignment in constructors.
2527 Item 14: Make destructors virtual in base classes.
2530 Item 15: Have @code{operator=} return a reference to @code{*this}.
2533 Item 23: Don't try to return a reference when you must return an object.
2537 Also warn about violations of the following style guidelines from
2538 Scott Meyers' @cite{More Effective C++} book:
2542 Item 6: Distinguish between prefix and postfix forms of increment and
2543 decrement operators.
2546 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2550 When selecting this option, be aware that the standard library
2551 headers do not obey all of these guidelines; use @samp{grep -v}
2552 to filter out those warnings.
2554 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2555 @opindex Wstrict-null-sentinel
2556 @opindex Wno-strict-null-sentinel
2557 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2558 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2559 to @code{__null}. Although it is a null pointer constant not a null pointer,
2560 it is guaranteed to be of the same size as a pointer. But this use is
2561 not portable across different compilers.
2563 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2564 @opindex Wno-non-template-friend
2565 @opindex Wnon-template-friend
2566 Disable warnings when non-templatized friend functions are declared
2567 within a template. Since the advent of explicit template specification
2568 support in G++, if the name of the friend is an unqualified-id (i.e.,
2569 @samp{friend foo(int)}), the C++ language specification demands that the
2570 friend declare or define an ordinary, nontemplate function. (Section
2571 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2572 could be interpreted as a particular specialization of a templatized
2573 function. Because this non-conforming behavior is no longer the default
2574 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2575 check existing code for potential trouble spots and is on by default.
2576 This new compiler behavior can be turned off with
2577 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2578 but disables the helpful warning.
2580 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2581 @opindex Wold-style-cast
2582 @opindex Wno-old-style-cast
2583 Warn if an old-style (C-style) cast to a non-void type is used within
2584 a C++ program. The new-style casts (@samp{dynamic_cast},
2585 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2586 less vulnerable to unintended effects and much easier to search for.
2588 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2589 @opindex Woverloaded-virtual
2590 @opindex Wno-overloaded-virtual
2591 @cindex overloaded virtual function, warning
2592 @cindex warning for overloaded virtual function
2593 Warn when a function declaration hides virtual functions from a
2594 base class. For example, in:
2601 struct B: public A @{
2606 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2617 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2618 @opindex Wno-pmf-conversions
2619 @opindex Wpmf-conversions
2620 Disable the diagnostic for converting a bound pointer to member function
2623 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2624 @opindex Wsign-promo
2625 @opindex Wno-sign-promo
2626 Warn when overload resolution chooses a promotion from unsigned or
2627 enumerated type to a signed type, over a conversion to an unsigned type of
2628 the same size. Previous versions of G++ tried to preserve
2629 unsignedness, but the standard mandates the current behavior.
2634 A& operator = (int);
2645 In this example, G++ synthesizes a default @samp{A& operator =
2646 (const A&);}, while cfront uses the user-defined @samp{operator =}.
2649 @node Objective-C and Objective-C++ Dialect Options
2650 @section Options Controlling Objective-C and Objective-C++ Dialects
2652 @cindex compiler options, Objective-C and Objective-C++
2653 @cindex Objective-C and Objective-C++ options, command-line
2654 @cindex options, Objective-C and Objective-C++
2655 (NOTE: This manual does not describe the Objective-C and Objective-C++
2656 languages themselves. @xref{Standards,,Language Standards
2657 Supported by GCC}, for references.)
2659 This section describes the command-line options that are only meaningful
2660 for Objective-C and Objective-C++ programs, but you can also use most of
2661 the language-independent GNU compiler options.
2662 For example, you might compile a file @code{some_class.m} like this:
2665 gcc -g -fgnu-runtime -O -c some_class.m
2669 In this example, @option{-fgnu-runtime} is an option meant only for
2670 Objective-C and Objective-C++ programs; you can use the other options with
2671 any language supported by GCC@.
2673 Note that since Objective-C is an extension of the C language, Objective-C
2674 compilations may also use options specific to the C front-end (e.g.,
2675 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2676 C++-specific options (e.g., @option{-Wabi}).
2678 Here is a list of options that are @emph{only} for compiling Objective-C
2679 and Objective-C++ programs:
2682 @item -fconstant-string-class=@var{class-name}
2683 @opindex fconstant-string-class
2684 Use @var{class-name} as the name of the class to instantiate for each
2685 literal string specified with the syntax @code{@@"@dots{}"}. The default
2686 class name is @code{NXConstantString} if the GNU runtime is being used, and
2687 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2688 @option{-fconstant-cfstrings} option, if also present, overrides the
2689 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2690 to be laid out as constant CoreFoundation strings.
2693 @opindex fgnu-runtime
2694 Generate object code compatible with the standard GNU Objective-C
2695 runtime. This is the default for most types of systems.
2697 @item -fnext-runtime
2698 @opindex fnext-runtime
2699 Generate output compatible with the NeXT runtime. This is the default
2700 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2701 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2704 @item -fno-nil-receivers
2705 @opindex fno-nil-receivers
2706 Assume that all Objective-C message dispatches (@code{[receiver
2707 message:arg]}) in this translation unit ensure that the receiver is
2708 not @code{nil}. This allows for more efficient entry points in the
2709 runtime to be used. This option is only available in conjunction with
2710 the NeXT runtime and ABI version 0 or 1.
2712 @item -fobjc-abi-version=@var{n}
2713 @opindex fobjc-abi-version
2714 Use version @var{n} of the Objective-C ABI for the selected runtime.
2715 This option is currently supported only for the NeXT runtime. In that
2716 case, Version 0 is the traditional (32-bit) ABI without support for
2717 properties and other Objective-C 2.0 additions. Version 1 is the
2718 traditional (32-bit) ABI with support for properties and other
2719 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2720 nothing is specified, the default is Version 0 on 32-bit target
2721 machines, and Version 2 on 64-bit target machines.
2723 @item -fobjc-call-cxx-cdtors
2724 @opindex fobjc-call-cxx-cdtors
2725 For each Objective-C class, check if any of its instance variables is a
2726 C++ object with a non-trivial default constructor. If so, synthesize a
2727 special @code{- (id) .cxx_construct} instance method which runs
2728 non-trivial default constructors on any such instance variables, in order,
2729 and then return @code{self}. Similarly, check if any instance variable
2730 is a C++ object with a non-trivial destructor, and if so, synthesize a
2731 special @code{- (void) .cxx_destruct} method which runs
2732 all such default destructors, in reverse order.
2734 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2735 methods thusly generated only operate on instance variables
2736 declared in the current Objective-C class, and not those inherited
2737 from superclasses. It is the responsibility of the Objective-C
2738 runtime to invoke all such methods in an object's inheritance
2739 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2740 by the runtime immediately after a new object instance is allocated;
2741 the @code{- (void) .cxx_destruct} methods are invoked immediately
2742 before the runtime deallocates an object instance.
2744 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2745 support for invoking the @code{- (id) .cxx_construct} and
2746 @code{- (void) .cxx_destruct} methods.
2748 @item -fobjc-direct-dispatch
2749 @opindex fobjc-direct-dispatch
2750 Allow fast jumps to the message dispatcher. On Darwin this is
2751 accomplished via the comm page.
2753 @item -fobjc-exceptions
2754 @opindex fobjc-exceptions
2755 Enable syntactic support for structured exception handling in
2756 Objective-C, similar to what is offered by C++ and Java. This option
2757 is required to use the Objective-C keywords @code{@@try},
2758 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2759 @code{@@synchronized}. This option is available with both the GNU
2760 runtime and the NeXT runtime (but not available in conjunction with
2761 the NeXT runtime on Mac OS X 10.2 and earlier).
2765 Enable garbage collection (GC) in Objective-C and Objective-C++
2766 programs. This option is only available with the NeXT runtime; the
2767 GNU runtime has a different garbage collection implementation that
2768 does not require special compiler flags.
2770 @item -fobjc-nilcheck
2771 @opindex fobjc-nilcheck
2772 For the NeXT runtime with version 2 of the ABI, check for a nil
2773 receiver in method invocations before doing the actual method call.
2774 This is the default and can be disabled using
2775 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2776 checked for nil in this way no matter what this flag is set to.
2777 Currently this flag does nothing when the GNU runtime, or an older
2778 version of the NeXT runtime ABI, is used.
2780 @item -fobjc-std=objc1
2782 Conform to the language syntax of Objective-C 1.0, the language
2783 recognized by GCC 4.0. This only affects the Objective-C additions to
2784 the C/C++ language; it does not affect conformance to C/C++ standards,
2785 which is controlled by the separate C/C++ dialect option flags. When
2786 this option is used with the Objective-C or Objective-C++ compiler,
2787 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2788 This is useful if you need to make sure that your Objective-C code can
2789 be compiled with older versions of GCC@.
2791 @item -freplace-objc-classes
2792 @opindex freplace-objc-classes
2793 Emit a special marker instructing @command{ld(1)} not to statically link in
2794 the resulting object file, and allow @command{dyld(1)} to load it in at
2795 run time instead. This is used in conjunction with the Fix-and-Continue
2796 debugging mode, where the object file in question may be recompiled and
2797 dynamically reloaded in the course of program execution, without the need
2798 to restart the program itself. Currently, Fix-and-Continue functionality
2799 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2804 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2805 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2806 compile time) with static class references that get initialized at load time,
2807 which improves run-time performance. Specifying the @option{-fzero-link} flag
2808 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2809 to be retained. This is useful in Zero-Link debugging mode, since it allows
2810 for individual class implementations to be modified during program execution.
2811 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2812 regardless of command-line options.
2816 Dump interface declarations for all classes seen in the source file to a
2817 file named @file{@var{sourcename}.decl}.
2819 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2820 @opindex Wassign-intercept
2821 @opindex Wno-assign-intercept
2822 Warn whenever an Objective-C assignment is being intercepted by the
2825 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2826 @opindex Wno-protocol
2828 If a class is declared to implement a protocol, a warning is issued for
2829 every method in the protocol that is not implemented by the class. The
2830 default behavior is to issue a warning for every method not explicitly
2831 implemented in the class, even if a method implementation is inherited
2832 from the superclass. If you use the @option{-Wno-protocol} option, then
2833 methods inherited from the superclass are considered to be implemented,
2834 and no warning is issued for them.
2836 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2838 @opindex Wno-selector
2839 Warn if multiple methods of different types for the same selector are
2840 found during compilation. The check is performed on the list of methods
2841 in the final stage of compilation. Additionally, a check is performed
2842 for each selector appearing in a @code{@@selector(@dots{})}
2843 expression, and a corresponding method for that selector has been found
2844 during compilation. Because these checks scan the method table only at
2845 the end of compilation, these warnings are not produced if the final
2846 stage of compilation is not reached, for example because an error is
2847 found during compilation, or because the @option{-fsyntax-only} option is
2850 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2851 @opindex Wstrict-selector-match
2852 @opindex Wno-strict-selector-match
2853 Warn if multiple methods with differing argument and/or return types are
2854 found for a given selector when attempting to send a message using this
2855 selector to a receiver of type @code{id} or @code{Class}. When this flag
2856 is off (which is the default behavior), the compiler omits such warnings
2857 if any differences found are confined to types that share the same size
2860 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2861 @opindex Wundeclared-selector
2862 @opindex Wno-undeclared-selector
2863 Warn if a @code{@@selector(@dots{})} expression referring to an
2864 undeclared selector is found. A selector is considered undeclared if no
2865 method with that name has been declared before the
2866 @code{@@selector(@dots{})} expression, either explicitly in an
2867 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2868 an @code{@@implementation} section. This option always performs its
2869 checks as soon as a @code{@@selector(@dots{})} expression is found,
2870 while @option{-Wselector} only performs its checks in the final stage of
2871 compilation. This also enforces the coding style convention
2872 that methods and selectors must be declared before being used.
2874 @item -print-objc-runtime-info
2875 @opindex print-objc-runtime-info
2876 Generate C header describing the largest structure that is passed by
2881 @node Language Independent Options
2882 @section Options to Control Diagnostic Messages Formatting
2883 @cindex options to control diagnostics formatting
2884 @cindex diagnostic messages
2885 @cindex message formatting
2887 Traditionally, diagnostic messages have been formatted irrespective of
2888 the output device's aspect (e.g.@: its width, @dots{}). You can use the
2889 options described below
2890 to control the formatting algorithm for diagnostic messages,
2891 e.g.@: how many characters per line, how often source location
2892 information should be reported. Note that some language front ends may not
2893 honor these options.
2896 @item -fmessage-length=@var{n}
2897 @opindex fmessage-length
2898 Try to format error messages so that they fit on lines of about @var{n}
2899 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2900 the front ends supported by GCC@. If @var{n} is zero, then no
2901 line-wrapping is done; each error message appears on a single
2904 @item -fdiagnostics-show-location=once
2905 @opindex fdiagnostics-show-location
2906 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2907 reporter to emit @emph{once} source location information; that is, in
2908 case the message is too long to fit on a single physical line and has to
2909 be wrapped, the source location won't be emitted (as prefix) again,
2910 over and over, in subsequent continuation lines. This is the default
2913 @item -fdiagnostics-show-location=every-line
2914 Only meaningful in line-wrapping mode. Instructs the diagnostic
2915 messages reporter to emit the same source location information (as
2916 prefix) for physical lines that result from the process of breaking
2917 a message which is too long to fit on a single line.
2919 @item -fno-diagnostics-show-option
2920 @opindex fno-diagnostics-show-option
2921 @opindex fdiagnostics-show-option
2922 By default, each diagnostic emitted includes text indicating the
2923 command-line option that directly controls the diagnostic (if such an
2924 option is known to the diagnostic machinery). Specifying the
2925 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2927 @item -fno-diagnostics-show-caret
2928 @opindex fno-diagnostics-show-caret
2929 @opindex fdiagnostics-show-caret
2930 By default, each diagnostic emitted includes the original source line
2931 and a caret '^' indicating the column. This option suppresses this
2936 @node Warning Options
2937 @section Options to Request or Suppress Warnings
2938 @cindex options to control warnings
2939 @cindex warning messages
2940 @cindex messages, warning
2941 @cindex suppressing warnings
2943 Warnings are diagnostic messages that report constructions that
2944 are not inherently erroneous but that are risky or suggest there
2945 may have been an error.
2947 The following language-independent options do not enable specific
2948 warnings but control the kinds of diagnostics produced by GCC@.
2951 @cindex syntax checking
2953 @opindex fsyntax-only
2954 Check the code for syntax errors, but don't do anything beyond that.
2956 @item -fmax-errors=@var{n}
2957 @opindex fmax-errors
2958 Limits the maximum number of error messages to @var{n}, at which point
2959 GCC bails out rather than attempting to continue processing the source
2960 code. If @var{n} is 0 (the default), there is no limit on the number
2961 of error messages produced. If @option{-Wfatal-errors} is also
2962 specified, then @option{-Wfatal-errors} takes precedence over this
2967 Inhibit all warning messages.
2972 Make all warnings into errors.
2977 Make the specified warning into an error. The specifier for a warning
2978 is appended, for example @option{-Werror=switch} turns the warnings
2979 controlled by @option{-Wswitch} into errors. This switch takes a
2980 negative form, to be used to negate @option{-Werror} for specific
2981 warnings, for example @option{-Wno-error=switch} makes
2982 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2985 The warning message for each controllable warning includes the
2986 option that controls the warning. That option can then be used with
2987 @option{-Werror=} and @option{-Wno-error=} as described above.
2988 (Printing of the option in the warning message can be disabled using the
2989 @option{-fno-diagnostics-show-option} flag.)
2991 Note that specifying @option{-Werror=}@var{foo} automatically implies
2992 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2995 @item -Wfatal-errors
2996 @opindex Wfatal-errors
2997 @opindex Wno-fatal-errors
2998 This option causes the compiler to abort compilation on the first error
2999 occurred rather than trying to keep going and printing further error
3004 You can request many specific warnings with options beginning
3005 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3006 implicit declarations. Each of these specific warning options also
3007 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3008 example, @option{-Wno-implicit}. This manual lists only one of the
3009 two forms, whichever is not the default. For further,
3010 language-specific options also refer to @ref{C++ Dialect Options} and
3011 @ref{Objective-C and Objective-C++ Dialect Options}.
3013 When an unrecognized warning option is requested (e.g.,
3014 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3015 that the option is not recognized. However, if the @option{-Wno-} form
3016 is used, the behavior is slightly different: no diagnostic is
3017 produced for @option{-Wno-unknown-warning} unless other diagnostics
3018 are being produced. This allows the use of new @option{-Wno-} options
3019 with old compilers, but if something goes wrong, the compiler
3020 warns that an unrecognized option is present.
3027 Issue all the warnings demanded by strict ISO C and ISO C++;
3028 reject all programs that use forbidden extensions, and some other
3029 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3030 version of the ISO C standard specified by any @option{-std} option used.
3032 Valid ISO C and ISO C++ programs should compile properly with or without
3033 this option (though a rare few require @option{-ansi} or a
3034 @option{-std} option specifying the required version of ISO C)@. However,
3035 without this option, certain GNU extensions and traditional C and C++
3036 features are supported as well. With this option, they are rejected.
3038 @option{-Wpedantic} does not cause warning messages for use of the
3039 alternate keywords whose names begin and end with @samp{__}. Pedantic
3040 warnings are also disabled in the expression that follows
3041 @code{__extension__}. However, only system header files should use
3042 these escape routes; application programs should avoid them.
3043 @xref{Alternate Keywords}.
3045 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3046 C conformance. They soon find that it does not do quite what they want:
3047 it finds some non-ISO practices, but not all---only those for which
3048 ISO C @emph{requires} a diagnostic, and some others for which
3049 diagnostics have been added.
3051 A feature to report any failure to conform to ISO C might be useful in
3052 some instances, but would require considerable additional work and would
3053 be quite different from @option{-Wpedantic}. We don't have plans to
3054 support such a feature in the near future.
3056 Where the standard specified with @option{-std} represents a GNU
3057 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3058 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3059 extended dialect is based. Warnings from @option{-Wpedantic} are given
3060 where they are required by the base standard. (It does not make sense
3061 for such warnings to be given only for features not in the specified GNU
3062 C dialect, since by definition the GNU dialects of C include all
3063 features the compiler supports with the given option, and there would be
3064 nothing to warn about.)
3066 @item -pedantic-errors
3067 @opindex pedantic-errors
3068 Like @option{-Wpedantic}, except that errors are produced rather than
3074 This enables all the warnings about constructions that some users
3075 consider questionable, and that are easy to avoid (or modify to
3076 prevent the warning), even in conjunction with macros. This also
3077 enables some language-specific warnings described in @ref{C++ Dialect
3078 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3080 @option{-Wall} turns on the following warning flags:
3082 @gccoptlist{-Waddress @gol
3083 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3085 -Wchar-subscripts @gol
3086 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
3087 -Wimplicit-int @r{(C and Objective-C only)} @gol
3088 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3091 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3092 -Wmaybe-uninitialized @gol
3093 -Wmissing-braces @r{(only for C/ObjC)} @gol
3099 -Wsequence-point @gol
3100 -Wsign-compare @r{(only in C++)} @gol
3101 -Wstrict-aliasing @gol
3102 -Wstrict-overflow=1 @gol
3105 -Wuninitialized @gol
3106 -Wunknown-pragmas @gol
3107 -Wunused-function @gol
3110 -Wunused-variable @gol
3111 -Wvolatile-register-var @gol
3114 Note that some warning flags are not implied by @option{-Wall}. Some of
3115 them warn about constructions that users generally do not consider
3116 questionable, but which occasionally you might wish to check for;
3117 others warn about constructions that are necessary or hard to avoid in
3118 some cases, and there is no simple way to modify the code to suppress
3119 the warning. Some of them are enabled by @option{-Wextra} but many of
3120 them must be enabled individually.
3126 This enables some extra warning flags that are not enabled by
3127 @option{-Wall}. (This option used to be called @option{-W}. The older
3128 name is still supported, but the newer name is more descriptive.)
3130 @gccoptlist{-Wclobbered @gol
3132 -Wignored-qualifiers @gol
3133 -Wmissing-field-initializers @gol
3134 -Wmissing-parameter-type @r{(C only)} @gol
3135 -Wold-style-declaration @r{(C only)} @gol
3136 -Woverride-init @gol
3139 -Wuninitialized @gol
3140 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3141 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3144 The option @option{-Wextra} also prints warning messages for the
3150 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3151 @samp{>}, or @samp{>=}.
3154 (C++ only) An enumerator and a non-enumerator both appear in a
3155 conditional expression.
3158 (C++ only) Ambiguous virtual bases.
3161 (C++ only) Subscripting an array that has been declared @samp{register}.
3164 (C++ only) Taking the address of a variable that has been declared
3168 (C++ only) A base class is not initialized in a derived class' copy
3173 @item -Wchar-subscripts
3174 @opindex Wchar-subscripts
3175 @opindex Wno-char-subscripts
3176 Warn if an array subscript has type @code{char}. This is a common cause
3177 of error, as programmers often forget that this type is signed on some
3179 This warning is enabled by @option{-Wall}.
3183 @opindex Wno-comment
3184 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3185 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3186 This warning is enabled by @option{-Wall}.
3188 @item -Wno-coverage-mismatch
3189 @opindex Wno-coverage-mismatch
3190 Warn if feedback profiles do not match when using the
3191 @option{-fprofile-use} option.
3192 If a source file is changed between compiling with @option{-fprofile-gen} and
3193 with @option{-fprofile-use}, the files with the profile feedback can fail
3194 to match the source file and GCC cannot use the profile feedback
3195 information. By default, this warning is enabled and is treated as an
3196 error. @option{-Wno-coverage-mismatch} can be used to disable the
3197 warning or @option{-Wno-error=coverage-mismatch} can be used to
3198 disable the error. Disabling the error for this warning can result in
3199 poorly optimized code and is useful only in the
3200 case of very minor changes such as bug fixes to an existing code-base.
3201 Completely disabling the warning is not recommended.
3204 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3206 Suppress warning messages emitted by @code{#warning} directives.
3208 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3209 @opindex Wdouble-promotion
3210 @opindex Wno-double-promotion
3211 Give a warning when a value of type @code{float} is implicitly
3212 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3213 floating-point unit implement @code{float} in hardware, but emulate
3214 @code{double} in software. On such a machine, doing computations
3215 using @code{double} values is much more expensive because of the
3216 overhead required for software emulation.
3218 It is easy to accidentally do computations with @code{double} because
3219 floating-point literals are implicitly of type @code{double}. For
3223 float area(float radius)
3225 return 3.14159 * radius * radius;
3229 the compiler performs the entire computation with @code{double}
3230 because the floating-point literal is a @code{double}.
3235 @opindex ffreestanding
3236 @opindex fno-builtin
3237 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3238 the arguments supplied have types appropriate to the format string
3239 specified, and that the conversions specified in the format string make
3240 sense. This includes standard functions, and others specified by format
3241 attributes (@pxref{Function Attributes}), in the @code{printf},
3242 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3243 not in the C standard) families (or other target-specific families).
3244 Which functions are checked without format attributes having been
3245 specified depends on the standard version selected, and such checks of
3246 functions without the attribute specified are disabled by
3247 @option{-ffreestanding} or @option{-fno-builtin}.
3249 The formats are checked against the format features supported by GNU
3250 libc version 2.2. These include all ISO C90 and C99 features, as well
3251 as features from the Single Unix Specification and some BSD and GNU
3252 extensions. Other library implementations may not support all these
3253 features; GCC does not support warning about features that go beyond a
3254 particular library's limitations. However, if @option{-Wpedantic} is used
3255 with @option{-Wformat}, warnings are given about format features not
3256 in the selected standard version (but not for @code{strfmon} formats,
3257 since those are not in any version of the C standard). @xref{C Dialect
3258 Options,,Options Controlling C Dialect}.
3260 Since @option{-Wformat} also checks for null format arguments for
3261 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3263 @option{-Wformat} is included in @option{-Wall}. For more control over some
3264 aspects of format checking, the options @option{-Wformat-y2k},
3265 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3266 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3267 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3270 @opindex Wformat-y2k
3271 @opindex Wno-format-y2k
3272 If @option{-Wformat} is specified, also warn about @code{strftime}
3273 formats that may yield only a two-digit year.
3275 @item -Wno-format-contains-nul
3276 @opindex Wno-format-contains-nul
3277 @opindex Wformat-contains-nul
3278 If @option{-Wformat} is specified, do not warn about format strings that
3281 @item -Wno-format-extra-args
3282 @opindex Wno-format-extra-args
3283 @opindex Wformat-extra-args
3284 If @option{-Wformat} is specified, do not warn about excess arguments to a
3285 @code{printf} or @code{scanf} format function. The C standard specifies
3286 that such arguments are ignored.
3288 Where the unused arguments lie between used arguments that are
3289 specified with @samp{$} operand number specifications, normally
3290 warnings are still given, since the implementation could not know what
3291 type to pass to @code{va_arg} to skip the unused arguments. However,
3292 in the case of @code{scanf} formats, this option suppresses the
3293 warning if the unused arguments are all pointers, since the Single
3294 Unix Specification says that such unused arguments are allowed.
3296 @item -Wno-format-zero-length
3297 @opindex Wno-format-zero-length
3298 @opindex Wformat-zero-length
3299 If @option{-Wformat} is specified, do not warn about zero-length formats.
3300 The C standard specifies that zero-length formats are allowed.
3302 @item -Wformat-nonliteral
3303 @opindex Wformat-nonliteral
3304 @opindex Wno-format-nonliteral
3305 If @option{-Wformat} is specified, also warn if the format string is not a
3306 string literal and so cannot be checked, unless the format function
3307 takes its format arguments as a @code{va_list}.
3309 @item -Wformat-security
3310 @opindex Wformat-security
3311 @opindex Wno-format-security
3312 If @option{-Wformat} is specified, also warn about uses of format
3313 functions that represent possible security problems. At present, this
3314 warns about calls to @code{printf} and @code{scanf} functions where the
3315 format string is not a string literal and there are no format arguments,
3316 as in @code{printf (foo);}. This may be a security hole if the format
3317 string came from untrusted input and contains @samp{%n}. (This is
3318 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3319 in future warnings may be added to @option{-Wformat-security} that are not
3320 included in @option{-Wformat-nonliteral}.)
3324 @opindex Wno-format=2
3325 Enable @option{-Wformat} plus format checks not included in
3326 @option{-Wformat}. Currently equivalent to @option{-Wformat
3327 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3331 @opindex Wno-nonnull
3332 Warn about passing a null pointer for arguments marked as
3333 requiring a non-null value by the @code{nonnull} function attribute.
3335 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3336 can be disabled with the @option{-Wno-nonnull} option.
3338 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3340 @opindex Wno-init-self
3341 Warn about uninitialized variables that are initialized with themselves.
3342 Note this option can only be used with the @option{-Wuninitialized} option.
3344 For example, GCC warns about @code{i} being uninitialized in the
3345 following snippet only when @option{-Winit-self} has been specified:
3356 @item -Wimplicit-int @r{(C and Objective-C only)}
3357 @opindex Wimplicit-int
3358 @opindex Wno-implicit-int
3359 Warn when a declaration does not specify a type.
3360 This warning is enabled by @option{-Wall}.
3362 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3363 @opindex Wimplicit-function-declaration
3364 @opindex Wno-implicit-function-declaration
3365 Give a warning whenever a function is used before being declared. In
3366 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3367 enabled by default and it is made into an error by
3368 @option{-pedantic-errors}. This warning is also enabled by
3371 @item -Wimplicit @r{(C and Objective-C only)}
3373 @opindex Wno-implicit
3374 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3375 This warning is enabled by @option{-Wall}.
3377 @item -Wignored-qualifiers @r{(C and C++ only)}
3378 @opindex Wignored-qualifiers
3379 @opindex Wno-ignored-qualifiers
3380 Warn if the return type of a function has a type qualifier
3381 such as @code{const}. For ISO C such a type qualifier has no effect,
3382 since the value returned by a function is not an lvalue.
3383 For C++, the warning is only emitted for scalar types or @code{void}.
3384 ISO C prohibits qualified @code{void} return types on function
3385 definitions, so such return types always receive a warning
3386 even without this option.
3388 This warning is also enabled by @option{-Wextra}.
3393 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3394 a function with external linkage, returning int, taking either zero
3395 arguments, two, or three arguments of appropriate types. This warning
3396 is enabled by default in C++ and is enabled by either @option{-Wall}
3397 or @option{-Wpedantic}.
3399 @item -Wmissing-braces
3400 @opindex Wmissing-braces
3401 @opindex Wno-missing-braces
3402 Warn if an aggregate or union initializer is not fully bracketed. In
3403 the following example, the initializer for @samp{a} is not fully
3404 bracketed, but that for @samp{b} is fully bracketed. This warning is
3405 enabled by @option{-Wall} in C.
3408 int a[2][2] = @{ 0, 1, 2, 3 @};
3409 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3412 This warning is enabled by @option{-Wall}.
3414 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3415 @opindex Wmissing-include-dirs
3416 @opindex Wno-missing-include-dirs
3417 Warn if a user-supplied include directory does not exist.
3420 @opindex Wparentheses
3421 @opindex Wno-parentheses
3422 Warn if parentheses are omitted in certain contexts, such
3423 as when there is an assignment in a context where a truth value
3424 is expected, or when operators are nested whose precedence people
3425 often get confused about.
3427 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3428 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3429 interpretation from that of ordinary mathematical notation.
3431 Also warn about constructions where there may be confusion to which
3432 @code{if} statement an @code{else} branch belongs. Here is an example of
3447 In C/C++, every @code{else} branch belongs to the innermost possible
3448 @code{if} statement, which in this example is @code{if (b)}. This is
3449 often not what the programmer expected, as illustrated in the above
3450 example by indentation the programmer chose. When there is the
3451 potential for this confusion, GCC issues a warning when this flag
3452 is specified. To eliminate the warning, add explicit braces around
3453 the innermost @code{if} statement so there is no way the @code{else}
3454 can belong to the enclosing @code{if}. The resulting code
3471 Also warn for dangerous uses of the GNU extension to
3472 @code{?:} with omitted middle operand. When the condition
3473 in the @code{?}: operator is a boolean expression, the omitted value is
3474 always 1. Often programmers expect it to be a value computed
3475 inside the conditional expression instead.
3477 This warning is enabled by @option{-Wall}.
3479 @item -Wsequence-point
3480 @opindex Wsequence-point
3481 @opindex Wno-sequence-point
3482 Warn about code that may have undefined semantics because of violations
3483 of sequence point rules in the C and C++ standards.
3485 The C and C++ standards defines the order in which expressions in a C/C++
3486 program are evaluated in terms of @dfn{sequence points}, which represent
3487 a partial ordering between the execution of parts of the program: those
3488 executed before the sequence point, and those executed after it. These
3489 occur after the evaluation of a full expression (one which is not part
3490 of a larger expression), after the evaluation of the first operand of a
3491 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3492 function is called (but after the evaluation of its arguments and the
3493 expression denoting the called function), and in certain other places.
3494 Other than as expressed by the sequence point rules, the order of
3495 evaluation of subexpressions of an expression is not specified. All
3496 these rules describe only a partial order rather than a total order,
3497 since, for example, if two functions are called within one expression
3498 with no sequence point between them, the order in which the functions
3499 are called is not specified. However, the standards committee have
3500 ruled that function calls do not overlap.
3502 It is not specified when between sequence points modifications to the
3503 values of objects take effect. Programs whose behavior depends on this
3504 have undefined behavior; the C and C++ standards specify that ``Between
3505 the previous and next sequence point an object shall have its stored
3506 value modified at most once by the evaluation of an expression.
3507 Furthermore, the prior value shall be read only to determine the value
3508 to be stored.''. If a program breaks these rules, the results on any
3509 particular implementation are entirely unpredictable.
3511 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3512 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3513 diagnosed by this option, and it may give an occasional false positive
3514 result, but in general it has been found fairly effective at detecting
3515 this sort of problem in programs.
3517 The standard is worded confusingly, therefore there is some debate
3518 over the precise meaning of the sequence point rules in subtle cases.
3519 Links to discussions of the problem, including proposed formal
3520 definitions, may be found on the GCC readings page, at
3521 @uref{http://gcc.gnu.org/@/readings.html}.
3523 This warning is enabled by @option{-Wall} for C and C++.
3526 @opindex Wreturn-type
3527 @opindex Wno-return-type
3528 Warn whenever a function is defined with a return-type that defaults
3529 to @code{int}. Also warn about any @code{return} statement with no
3530 return-value in a function whose return-type is not @code{void}
3531 (falling off the end of the function body is considered returning
3532 without a value), and about a @code{return} statement with an
3533 expression in a function whose return-type is @code{void}.
3535 For C++, a function without return type always produces a diagnostic
3536 message, even when @option{-Wno-return-type} is specified. The only
3537 exceptions are @samp{main} and functions defined in system headers.
3539 This warning is enabled by @option{-Wall}.
3544 Warn whenever a @code{switch} statement has an index of enumerated type
3545 and lacks a @code{case} for one or more of the named codes of that
3546 enumeration. (The presence of a @code{default} label prevents this
3547 warning.) @code{case} labels outside the enumeration range also
3548 provoke warnings when this option is used (even if there is a
3549 @code{default} label).
3550 This warning is enabled by @option{-Wall}.
3552 @item -Wswitch-default
3553 @opindex Wswitch-default
3554 @opindex Wno-switch-default
3555 Warn whenever a @code{switch} statement does not have a @code{default}
3559 @opindex Wswitch-enum
3560 @opindex Wno-switch-enum
3561 Warn whenever a @code{switch} statement has an index of enumerated type
3562 and lacks a @code{case} for one or more of the named codes of that
3563 enumeration. @code{case} labels outside the enumeration range also
3564 provoke warnings when this option is used. The only difference
3565 between @option{-Wswitch} and this option is that this option gives a
3566 warning about an omitted enumeration code even if there is a
3567 @code{default} label.
3569 @item -Wsync-nand @r{(C and C++ only)}
3571 @opindex Wno-sync-nand
3572 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3573 built-in functions are used. These functions changed semantics in GCC 4.4.
3577 @opindex Wno-trigraphs
3578 Warn if any trigraphs are encountered that might change the meaning of
3579 the program (trigraphs within comments are not warned about).
3580 This warning is enabled by @option{-Wall}.
3582 @item -Wunused-but-set-parameter
3583 @opindex Wunused-but-set-parameter
3584 @opindex Wno-unused-but-set-parameter
3585 Warn whenever a function parameter is assigned to, but otherwise unused
3586 (aside from its declaration).
3588 To suppress this warning use the @samp{unused} attribute
3589 (@pxref{Variable Attributes}).
3591 This warning is also enabled by @option{-Wunused} together with
3594 @item -Wunused-but-set-variable
3595 @opindex Wunused-but-set-variable
3596 @opindex Wno-unused-but-set-variable
3597 Warn whenever a local variable is assigned to, but otherwise unused
3598 (aside from its declaration).
3599 This warning is enabled by @option{-Wall}.
3601 To suppress this warning use the @samp{unused} attribute
3602 (@pxref{Variable Attributes}).
3604 This warning is also enabled by @option{-Wunused}, which is enabled
3607 @item -Wunused-function
3608 @opindex Wunused-function
3609 @opindex Wno-unused-function
3610 Warn whenever a static function is declared but not defined or a
3611 non-inline static function is unused.
3612 This warning is enabled by @option{-Wall}.
3614 @item -Wunused-label
3615 @opindex Wunused-label
3616 @opindex Wno-unused-label
3617 Warn whenever a label is declared but not used.
3618 This warning is enabled by @option{-Wall}.
3620 To suppress this warning use the @samp{unused} attribute
3621 (@pxref{Variable Attributes}).
3623 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3624 @opindex Wunused-local-typedefs
3625 Warn when a typedef locally defined in a function is not used.
3626 This warning is enabled by @option{-Wall}.
3628 @item -Wunused-parameter
3629 @opindex Wunused-parameter
3630 @opindex Wno-unused-parameter
3631 Warn whenever a function parameter is unused aside from its declaration.
3633 To suppress this warning use the @samp{unused} attribute
3634 (@pxref{Variable Attributes}).
3636 @item -Wno-unused-result
3637 @opindex Wunused-result
3638 @opindex Wno-unused-result
3639 Do not warn if a caller of a function marked with attribute
3640 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3641 its return value. The default is @option{-Wunused-result}.
3643 @item -Wunused-variable
3644 @opindex Wunused-variable
3645 @opindex Wno-unused-variable
3646 Warn whenever a local variable or non-constant static variable is unused
3647 aside from its declaration.
3648 This warning is enabled by @option{-Wall}.
3650 To suppress this warning use the @samp{unused} attribute
3651 (@pxref{Variable Attributes}).
3653 @item -Wunused-value
3654 @opindex Wunused-value
3655 @opindex Wno-unused-value
3656 Warn whenever a statement computes a result that is explicitly not
3657 used. To suppress this warning cast the unused expression to
3658 @samp{void}. This includes an expression-statement or the left-hand
3659 side of a comma expression that contains no side effects. For example,
3660 an expression such as @samp{x[i,j]} causes a warning, while
3661 @samp{x[(void)i,j]} does not.
3663 This warning is enabled by @option{-Wall}.
3668 All the above @option{-Wunused} options combined.
3670 In order to get a warning about an unused function parameter, you must
3671 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
3672 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
3674 @item -Wuninitialized
3675 @opindex Wuninitialized
3676 @opindex Wno-uninitialized
3677 Warn if an automatic variable is used without first being initialized
3678 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3679 warn if a non-static reference or non-static @samp{const} member
3680 appears in a class without constructors.
3682 If you want to warn about code that uses the uninitialized value of the
3683 variable in its own initializer, use the @option{-Winit-self} option.
3685 These warnings occur for individual uninitialized or clobbered
3686 elements of structure, union or array variables as well as for
3687 variables that are uninitialized or clobbered as a whole. They do
3688 not occur for variables or elements declared @code{volatile}. Because
3689 these warnings depend on optimization, the exact variables or elements
3690 for which there are warnings depends on the precise optimization
3691 options and version of GCC used.
3693 Note that there may be no warning about a variable that is used only
3694 to compute a value that itself is never used, because such
3695 computations may be deleted by data flow analysis before the warnings
3698 @item -Wmaybe-uninitialized
3699 @opindex Wmaybe-uninitialized
3700 @opindex Wno-maybe-uninitialized
3701 For an automatic variable, if there exists a path from the function
3702 entry to a use of the variable that is initialized, but there exist
3703 some other paths for which the variable is not initialized, the compiler
3704 emits a warning if it cannot prove the uninitialized paths are not
3705 executed at run time. These warnings are made optional because GCC is
3706 not smart enough to see all the reasons why the code might be correct
3707 in spite of appearing to have an error. Here is one example of how
3728 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3729 always initialized, but GCC doesn't know this. To suppress the
3730 warning, you need to provide a default case with assert(0) or
3733 @cindex @code{longjmp} warnings
3734 This option also warns when a non-volatile automatic variable might be
3735 changed by a call to @code{longjmp}. These warnings as well are possible
3736 only in optimizing compilation.
3738 The compiler sees only the calls to @code{setjmp}. It cannot know
3739 where @code{longjmp} will be called; in fact, a signal handler could
3740 call it at any point in the code. As a result, you may get a warning
3741 even when there is in fact no problem because @code{longjmp} cannot
3742 in fact be called at the place that would cause a problem.
3744 Some spurious warnings can be avoided if you declare all the functions
3745 you use that never return as @code{noreturn}. @xref{Function
3748 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3750 @item -Wunknown-pragmas
3751 @opindex Wunknown-pragmas
3752 @opindex Wno-unknown-pragmas
3753 @cindex warning for unknown pragmas
3754 @cindex unknown pragmas, warning
3755 @cindex pragmas, warning of unknown
3756 Warn when a @code{#pragma} directive is encountered that is not understood by
3757 GCC@. If this command-line option is used, warnings are even issued
3758 for unknown pragmas in system header files. This is not the case if
3759 the warnings are only enabled by the @option{-Wall} command-line option.
3762 @opindex Wno-pragmas
3764 Do not warn about misuses of pragmas, such as incorrect parameters,
3765 invalid syntax, or conflicts between pragmas. See also
3766 @option{-Wunknown-pragmas}.
3768 @item -Wstrict-aliasing
3769 @opindex Wstrict-aliasing
3770 @opindex Wno-strict-aliasing
3771 This option is only active when @option{-fstrict-aliasing} is active.
3772 It warns about code that might break the strict aliasing rules that the
3773 compiler is using for optimization. The warning does not catch all
3774 cases, but does attempt to catch the more common pitfalls. It is
3775 included in @option{-Wall}.
3776 It is equivalent to @option{-Wstrict-aliasing=3}
3778 @item -Wstrict-aliasing=n
3779 @opindex Wstrict-aliasing=n
3780 @opindex Wno-strict-aliasing=n
3781 This option is only active when @option{-fstrict-aliasing} is active.
3782 It warns about code that might break the strict aliasing rules that the
3783 compiler is using for optimization.
3784 Higher levels correspond to higher accuracy (fewer false positives).
3785 Higher levels also correspond to more effort, similar to the way @option{-O}
3787 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
3789 Level 1: Most aggressive, quick, least accurate.
3790 Possibly useful when higher levels
3791 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
3792 false negatives. However, it has many false positives.
3793 Warns for all pointer conversions between possibly incompatible types,
3794 even if never dereferenced. Runs in the front end only.
3796 Level 2: Aggressive, quick, not too precise.
3797 May still have many false positives (not as many as level 1 though),
3798 and few false negatives (but possibly more than level 1).
3799 Unlike level 1, it only warns when an address is taken. Warns about
3800 incomplete types. Runs in the front end only.
3802 Level 3 (default for @option{-Wstrict-aliasing}):
3803 Should have very few false positives and few false
3804 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3805 Takes care of the common pun+dereference pattern in the front end:
3806 @code{*(int*)&some_float}.
3807 If optimization is enabled, it also runs in the back end, where it deals
3808 with multiple statement cases using flow-sensitive points-to information.
3809 Only warns when the converted pointer is dereferenced.
3810 Does not warn about incomplete types.
3812 @item -Wstrict-overflow
3813 @itemx -Wstrict-overflow=@var{n}
3814 @opindex Wstrict-overflow
3815 @opindex Wno-strict-overflow
3816 This option is only active when @option{-fstrict-overflow} is active.
3817 It warns about cases where the compiler optimizes based on the
3818 assumption that signed overflow does not occur. Note that it does not
3819 warn about all cases where the code might overflow: it only warns
3820 about cases where the compiler implements some optimization. Thus
3821 this warning depends on the optimization level.
3823 An optimization that assumes that signed overflow does not occur is
3824 perfectly safe if the values of the variables involved are such that
3825 overflow never does, in fact, occur. Therefore this warning can
3826 easily give a false positive: a warning about code that is not
3827 actually a problem. To help focus on important issues, several
3828 warning levels are defined. No warnings are issued for the use of
3829 undefined signed overflow when estimating how many iterations a loop
3830 requires, in particular when determining whether a loop will be
3834 @item -Wstrict-overflow=1
3835 Warn about cases that are both questionable and easy to avoid. For
3836 example, with @option{-fstrict-overflow}, the compiler simplifies
3837 @code{x + 1 > x} to @code{1}. This level of
3838 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3839 are not, and must be explicitly requested.
3841 @item -Wstrict-overflow=2
3842 Also warn about other cases where a comparison is simplified to a
3843 constant. For example: @code{abs (x) >= 0}. This can only be
3844 simplified when @option{-fstrict-overflow} is in effect, because
3845 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3846 zero. @option{-Wstrict-overflow} (with no level) is the same as
3847 @option{-Wstrict-overflow=2}.
3849 @item -Wstrict-overflow=3
3850 Also warn about other cases where a comparison is simplified. For
3851 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
3853 @item -Wstrict-overflow=4
3854 Also warn about other simplifications not covered by the above cases.
3855 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
3857 @item -Wstrict-overflow=5
3858 Also warn about cases where the compiler reduces the magnitude of a
3859 constant involved in a comparison. For example: @code{x + 2 > y} is
3860 simplified to @code{x + 1 >= y}. This is reported only at the
3861 highest warning level because this simplification applies to many
3862 comparisons, so this warning level gives a very large number of
3866 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
3867 @opindex Wsuggest-attribute=
3868 @opindex Wno-suggest-attribute=
3869 Warn for cases where adding an attribute may be beneficial. The
3870 attributes currently supported are listed below.
3873 @item -Wsuggest-attribute=pure
3874 @itemx -Wsuggest-attribute=const
3875 @itemx -Wsuggest-attribute=noreturn
3876 @opindex Wsuggest-attribute=pure
3877 @opindex Wno-suggest-attribute=pure
3878 @opindex Wsuggest-attribute=const
3879 @opindex Wno-suggest-attribute=const
3880 @opindex Wsuggest-attribute=noreturn
3881 @opindex Wno-suggest-attribute=noreturn
3883 Warn about functions that might be candidates for attributes
3884 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3885 functions visible in other compilation units or (in the case of @code{pure} and
3886 @code{const}) if it cannot prove that the function returns normally. A function
3887 returns normally if it doesn't contain an infinite loop nor returns abnormally
3888 by throwing, calling @code{abort()} or trapping. This analysis requires option
3889 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3890 higher. Higher optimization levels improve the accuracy of the analysis.
3892 @item -Wsuggest-attribute=format
3893 @itemx -Wmissing-format-attribute
3894 @opindex Wsuggest-attribute=format
3895 @opindex Wmissing-format-attribute
3896 @opindex Wno-suggest-attribute=format
3897 @opindex Wno-missing-format-attribute
3901 Warn about function pointers that might be candidates for @code{format}
3902 attributes. Note these are only possible candidates, not absolute ones.
3903 GCC guesses that function pointers with @code{format} attributes that
3904 are used in assignment, initialization, parameter passing or return
3905 statements should have a corresponding @code{format} attribute in the
3906 resulting type. I.e.@: the left-hand side of the assignment or
3907 initialization, the type of the parameter variable, or the return type
3908 of the containing function respectively should also have a @code{format}
3909 attribute to avoid the warning.
3911 GCC also warns about function definitions that might be
3912 candidates for @code{format} attributes. Again, these are only
3913 possible candidates. GCC guesses that @code{format} attributes
3914 might be appropriate for any function that calls a function like
3915 @code{vprintf} or @code{vscanf}, but this might not always be the
3916 case, and some functions for which @code{format} attributes are
3917 appropriate may not be detected.
3920 @item -Warray-bounds
3921 @opindex Wno-array-bounds
3922 @opindex Warray-bounds
3923 This option is only active when @option{-ftree-vrp} is active
3924 (default for @option{-O2} and above). It warns about subscripts to arrays
3925 that are always out of bounds. This warning is enabled by @option{-Wall}.
3927 @item -Wno-div-by-zero
3928 @opindex Wno-div-by-zero
3929 @opindex Wdiv-by-zero
3930 Do not warn about compile-time integer division by zero. Floating-point
3931 division by zero is not warned about, as it can be a legitimate way of
3932 obtaining infinities and NaNs.
3934 @item -Wsystem-headers
3935 @opindex Wsystem-headers
3936 @opindex Wno-system-headers
3937 @cindex warnings from system headers
3938 @cindex system headers, warnings from
3939 Print warning messages for constructs found in system header files.
3940 Warnings from system headers are normally suppressed, on the assumption
3941 that they usually do not indicate real problems and would only make the
3942 compiler output harder to read. Using this command-line option tells
3943 GCC to emit warnings from system headers as if they occurred in user
3944 code. However, note that using @option{-Wall} in conjunction with this
3945 option does @emph{not} warn about unknown pragmas in system
3946 headers---for that, @option{-Wunknown-pragmas} must also be used.
3949 @opindex Wtrampolines
3950 @opindex Wno-trampolines
3951 Warn about trampolines generated for pointers to nested functions.
3953 A trampoline is a small piece of data or code that is created at run
3954 time on the stack when the address of a nested function is taken, and
3955 is used to call the nested function indirectly. For some targets, it
3956 is made up of data only and thus requires no special treatment. But,
3957 for most targets, it is made up of code and thus requires the stack
3958 to be made executable in order for the program to work properly.
3961 @opindex Wfloat-equal
3962 @opindex Wno-float-equal
3963 Warn if floating-point values are used in equality comparisons.
3965 The idea behind this is that sometimes it is convenient (for the
3966 programmer) to consider floating-point values as approximations to
3967 infinitely precise real numbers. If you are doing this, then you need
3968 to compute (by analyzing the code, or in some other way) the maximum or
3969 likely maximum error that the computation introduces, and allow for it
3970 when performing comparisons (and when producing output, but that's a
3971 different problem). In particular, instead of testing for equality, you
3972 should check to see whether the two values have ranges that overlap; and
3973 this is done with the relational operators, so equality comparisons are
3976 @item -Wtraditional @r{(C and Objective-C only)}
3977 @opindex Wtraditional
3978 @opindex Wno-traditional
3979 Warn about certain constructs that behave differently in traditional and
3980 ISO C@. Also warn about ISO C constructs that have no traditional C
3981 equivalent, and/or problematic constructs that should be avoided.
3985 Macro parameters that appear within string literals in the macro body.
3986 In traditional C macro replacement takes place within string literals,
3987 but does not in ISO C@.
3990 In traditional C, some preprocessor directives did not exist.
3991 Traditional preprocessors only considered a line to be a directive
3992 if the @samp{#} appeared in column 1 on the line. Therefore
3993 @option{-Wtraditional} warns about directives that traditional C
3994 understands but ignores because the @samp{#} does not appear as the
3995 first character on the line. It also suggests you hide directives like
3996 @samp{#pragma} not understood by traditional C by indenting them. Some
3997 traditional implementations do not recognize @samp{#elif}, so this option
3998 suggests avoiding it altogether.
4001 A function-like macro that appears without arguments.
4004 The unary plus operator.
4007 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4008 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4009 constants.) Note, these suffixes appear in macros defined in the system
4010 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4011 Use of these macros in user code might normally lead to spurious
4012 warnings, however GCC's integrated preprocessor has enough context to
4013 avoid warning in these cases.
4016 A function declared external in one block and then used after the end of
4020 A @code{switch} statement has an operand of type @code{long}.
4023 A non-@code{static} function declaration follows a @code{static} one.
4024 This construct is not accepted by some traditional C compilers.
4027 The ISO type of an integer constant has a different width or
4028 signedness from its traditional type. This warning is only issued if
4029 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4030 typically represent bit patterns, are not warned about.
4033 Usage of ISO string concatenation is detected.
4036 Initialization of automatic aggregates.
4039 Identifier conflicts with labels. Traditional C lacks a separate
4040 namespace for labels.
4043 Initialization of unions. If the initializer is zero, the warning is
4044 omitted. This is done under the assumption that the zero initializer in
4045 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4046 initializer warnings and relies on default initialization to zero in the
4050 Conversions by prototypes between fixed/floating-point values and vice
4051 versa. The absence of these prototypes when compiling with traditional
4052 C causes serious problems. This is a subset of the possible
4053 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4056 Use of ISO C style function definitions. This warning intentionally is
4057 @emph{not} issued for prototype declarations or variadic functions
4058 because these ISO C features appear in your code when using
4059 libiberty's traditional C compatibility macros, @code{PARAMS} and
4060 @code{VPARAMS}. This warning is also bypassed for nested functions
4061 because that feature is already a GCC extension and thus not relevant to
4062 traditional C compatibility.
4065 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4066 @opindex Wtraditional-conversion
4067 @opindex Wno-traditional-conversion
4068 Warn if a prototype causes a type conversion that is different from what
4069 would happen to the same argument in the absence of a prototype. This
4070 includes conversions of fixed point to floating and vice versa, and
4071 conversions changing the width or signedness of a fixed-point argument
4072 except when the same as the default promotion.
4074 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4075 @opindex Wdeclaration-after-statement
4076 @opindex Wno-declaration-after-statement
4077 Warn when a declaration is found after a statement in a block. This
4078 construct, known from C++, was introduced with ISO C99 and is by default
4079 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4080 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4085 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4087 @item -Wno-endif-labels
4088 @opindex Wno-endif-labels
4089 @opindex Wendif-labels
4090 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4095 Warn whenever a local variable or type declaration shadows another variable,
4096 parameter, type, or class member (in C++), or whenever a built-in function
4097 is shadowed. Note that in C++, the compiler warns if a local variable
4098 shadows an explicit typedef, but not if it shadows a struct/class/enum.
4100 @item -Wlarger-than=@var{len}
4101 @opindex Wlarger-than=@var{len}
4102 @opindex Wlarger-than-@var{len}
4103 Warn whenever an object of larger than @var{len} bytes is defined.
4105 @item -Wframe-larger-than=@var{len}
4106 @opindex Wframe-larger-than
4107 Warn if the size of a function frame is larger than @var{len} bytes.
4108 The computation done to determine the stack frame size is approximate
4109 and not conservative.
4110 The actual requirements may be somewhat greater than @var{len}
4111 even if you do not get a warning. In addition, any space allocated
4112 via @code{alloca}, variable-length arrays, or related constructs
4113 is not included by the compiler when determining
4114 whether or not to issue a warning.
4116 @item -Wno-free-nonheap-object
4117 @opindex Wno-free-nonheap-object
4118 @opindex Wfree-nonheap-object
4119 Do not warn when attempting to free an object that was not allocated
4122 @item -Wstack-usage=@var{len}
4123 @opindex Wstack-usage
4124 Warn if the stack usage of a function might be larger than @var{len} bytes.
4125 The computation done to determine the stack usage is conservative.
4126 Any space allocated via @code{alloca}, variable-length arrays, or related
4127 constructs is included by the compiler when determining whether or not to
4130 The message is in keeping with the output of @option{-fstack-usage}.
4134 If the stack usage is fully static but exceeds the specified amount, it's:
4137 warning: stack usage is 1120 bytes
4140 If the stack usage is (partly) dynamic but bounded, it's:
4143 warning: stack usage might be 1648 bytes
4146 If the stack usage is (partly) dynamic and not bounded, it's:
4149 warning: stack usage might be unbounded
4153 @item -Wunsafe-loop-optimizations
4154 @opindex Wunsafe-loop-optimizations
4155 @opindex Wno-unsafe-loop-optimizations
4156 Warn if the loop cannot be optimized because the compiler cannot
4157 assume anything on the bounds of the loop indices. With
4158 @option{-funsafe-loop-optimizations} warn if the compiler makes
4161 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4162 @opindex Wno-pedantic-ms-format
4163 @opindex Wpedantic-ms-format
4164 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4165 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4166 depending on the MS runtime, when you are using the options @option{-Wformat}
4167 and @option{-Wpedantic} without gnu-extensions.
4169 @item -Wpointer-arith
4170 @opindex Wpointer-arith
4171 @opindex Wno-pointer-arith
4172 Warn about anything that depends on the ``size of'' a function type or
4173 of @code{void}. GNU C assigns these types a size of 1, for
4174 convenience in calculations with @code{void *} pointers and pointers
4175 to functions. In C++, warn also when an arithmetic operation involves
4176 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4179 @opindex Wtype-limits
4180 @opindex Wno-type-limits
4181 Warn if a comparison is always true or always false due to the limited
4182 range of the data type, but do not warn for constant expressions. For
4183 example, warn if an unsigned variable is compared against zero with
4184 @samp{<} or @samp{>=}. This warning is also enabled by
4187 @item -Wbad-function-cast @r{(C and Objective-C only)}
4188 @opindex Wbad-function-cast
4189 @opindex Wno-bad-function-cast
4190 Warn whenever a function call is cast to a non-matching type.
4191 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4193 @item -Wc++-compat @r{(C and Objective-C only)}
4194 Warn about ISO C constructs that are outside of the common subset of
4195 ISO C and ISO C++, e.g.@: request for implicit conversion from
4196 @code{void *} to a pointer to non-@code{void} type.
4198 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4199 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4200 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4201 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4202 enabled by @option{-Wall}.
4206 @opindex Wno-cast-qual
4207 Warn whenever a pointer is cast so as to remove a type qualifier from
4208 the target type. For example, warn if a @code{const char *} is cast
4209 to an ordinary @code{char *}.
4211 Also warn when making a cast that introduces a type qualifier in an
4212 unsafe way. For example, casting @code{char **} to @code{const char **}
4213 is unsafe, as in this example:
4216 /* p is char ** value. */
4217 const char **q = (const char **) p;
4218 /* Assignment of readonly string to const char * is OK. */
4220 /* Now char** pointer points to read-only memory. */
4225 @opindex Wcast-align
4226 @opindex Wno-cast-align
4227 Warn whenever a pointer is cast such that the required alignment of the
4228 target is increased. For example, warn if a @code{char *} is cast to
4229 an @code{int *} on machines where integers can only be accessed at
4230 two- or four-byte boundaries.
4232 @item -Wwrite-strings
4233 @opindex Wwrite-strings
4234 @opindex Wno-write-strings
4235 When compiling C, give string constants the type @code{const
4236 char[@var{length}]} so that copying the address of one into a
4237 non-@code{const} @code{char *} pointer produces a warning. These
4238 warnings help you find at compile time code that can try to write
4239 into a string constant, but only if you have been very careful about
4240 using @code{const} in declarations and prototypes. Otherwise, it is
4241 just a nuisance. This is why we did not make @option{-Wall} request
4244 When compiling C++, warn about the deprecated conversion from string
4245 literals to @code{char *}. This warning is enabled by default for C++
4250 @opindex Wno-clobbered
4251 Warn for variables that might be changed by @samp{longjmp} or
4252 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4255 @opindex Wconversion
4256 @opindex Wno-conversion
4257 Warn for implicit conversions that may alter a value. This includes
4258 conversions between real and integer, like @code{abs (x)} when
4259 @code{x} is @code{double}; conversions between signed and unsigned,
4260 like @code{unsigned ui = -1}; and conversions to smaller types, like
4261 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4262 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4263 changed by the conversion like in @code{abs (2.0)}. Warnings about
4264 conversions between signed and unsigned integers can be disabled by
4265 using @option{-Wno-sign-conversion}.
4267 For C++, also warn for confusing overload resolution for user-defined
4268 conversions; and conversions that never use a type conversion
4269 operator: conversions to @code{void}, the same type, a base class or a
4270 reference to them. Warnings about conversions between signed and
4271 unsigned integers are disabled by default in C++ unless
4272 @option{-Wsign-conversion} is explicitly enabled.
4274 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4275 @opindex Wconversion-null
4276 @opindex Wno-conversion-null
4277 Do not warn for conversions between @code{NULL} and non-pointer
4278 types. @option{-Wconversion-null} is enabled by default.
4280 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4281 @opindex Wzero-as-null-pointer-constant
4282 @opindex Wno-zero-as-null-pointer-constant
4283 Warn when a literal '0' is used as null pointer constant. This can
4284 be useful to facilitate the conversion to @code{nullptr} in C++11.
4286 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4287 @opindex Wuseless-cast
4288 @opindex Wno-useless-cast
4289 Warn when an expression is casted to its own type.
4292 @opindex Wempty-body
4293 @opindex Wno-empty-body
4294 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4295 while} statement. This warning is also enabled by @option{-Wextra}.
4297 @item -Wenum-compare
4298 @opindex Wenum-compare
4299 @opindex Wno-enum-compare
4300 Warn about a comparison between values of different enumerated types.
4301 In C++ enumeral mismatches in conditional expressions are also
4302 diagnosed and the warning is enabled by default. In C this warning is
4303 enabled by @option{-Wall}.
4305 @item -Wjump-misses-init @r{(C, Objective-C only)}
4306 @opindex Wjump-misses-init
4307 @opindex Wno-jump-misses-init
4308 Warn if a @code{goto} statement or a @code{switch} statement jumps
4309 forward across the initialization of a variable, or jumps backward to a
4310 label after the variable has been initialized. This only warns about
4311 variables that are initialized when they are declared. This warning is
4312 only supported for C and Objective-C; in C++ this sort of branch is an
4315 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4316 can be disabled with the @option{-Wno-jump-misses-init} option.
4318 @item -Wsign-compare
4319 @opindex Wsign-compare
4320 @opindex Wno-sign-compare
4321 @cindex warning for comparison of signed and unsigned values
4322 @cindex comparison of signed and unsigned values, warning
4323 @cindex signed and unsigned values, comparison warning
4324 Warn when a comparison between signed and unsigned values could produce
4325 an incorrect result when the signed value is converted to unsigned.
4326 This warning is also enabled by @option{-Wextra}; to get the other warnings
4327 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4329 @item -Wsign-conversion
4330 @opindex Wsign-conversion
4331 @opindex Wno-sign-conversion
4332 Warn for implicit conversions that may change the sign of an integer
4333 value, like assigning a signed integer expression to an unsigned
4334 integer variable. An explicit cast silences the warning. In C, this
4335 option is enabled also by @option{-Wconversion}.
4339 @opindex Wno-address
4340 Warn about suspicious uses of memory addresses. These include using
4341 the address of a function in a conditional expression, such as
4342 @code{void func(void); if (func)}, and comparisons against the memory
4343 address of a string literal, such as @code{if (x == "abc")}. Such
4344 uses typically indicate a programmer error: the address of a function
4345 always evaluates to true, so their use in a conditional usually
4346 indicate that the programmer forgot the parentheses in a function
4347 call; and comparisons against string literals result in unspecified
4348 behavior and are not portable in C, so they usually indicate that the
4349 programmer intended to use @code{strcmp}. This warning is enabled by
4353 @opindex Wlogical-op
4354 @opindex Wno-logical-op
4355 Warn about suspicious uses of logical operators in expressions.
4356 This includes using logical operators in contexts where a
4357 bit-wise operator is likely to be expected.
4359 @item -Waggregate-return
4360 @opindex Waggregate-return
4361 @opindex Wno-aggregate-return
4362 Warn if any functions that return structures or unions are defined or
4363 called. (In languages where you can return an array, this also elicits
4366 @item -Wno-attributes
4367 @opindex Wno-attributes
4368 @opindex Wattributes
4369 Do not warn if an unexpected @code{__attribute__} is used, such as
4370 unrecognized attributes, function attributes applied to variables,
4371 etc. This does not stop errors for incorrect use of supported
4374 @item -Wno-builtin-macro-redefined
4375 @opindex Wno-builtin-macro-redefined
4376 @opindex Wbuiltin-macro-redefined
4377 Do not warn if certain built-in macros are redefined. This suppresses
4378 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4379 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4381 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4382 @opindex Wstrict-prototypes
4383 @opindex Wno-strict-prototypes
4384 Warn if a function is declared or defined without specifying the
4385 argument types. (An old-style function definition is permitted without
4386 a warning if preceded by a declaration that specifies the argument
4389 @item -Wold-style-declaration @r{(C and Objective-C only)}
4390 @opindex Wold-style-declaration
4391 @opindex Wno-old-style-declaration
4392 Warn for obsolescent usages, according to the C Standard, in a
4393 declaration. For example, warn if storage-class specifiers like
4394 @code{static} are not the first things in a declaration. This warning
4395 is also enabled by @option{-Wextra}.
4397 @item -Wold-style-definition @r{(C and Objective-C only)}
4398 @opindex Wold-style-definition
4399 @opindex Wno-old-style-definition
4400 Warn if an old-style function definition is used. A warning is given
4401 even if there is a previous prototype.
4403 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4404 @opindex Wmissing-parameter-type
4405 @opindex Wno-missing-parameter-type
4406 A function parameter is declared without a type specifier in K&R-style
4413 This warning is also enabled by @option{-Wextra}.
4415 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4416 @opindex Wmissing-prototypes
4417 @opindex Wno-missing-prototypes
4418 Warn if a global function is defined without a previous prototype
4419 declaration. This warning is issued even if the definition itself
4420 provides a prototype. Use this option to detect global functions
4421 that do not have a matching prototype declaration in a header file.
4422 This option is not valid for C++ because all function declarations
4423 provide prototypes and a non-matching declaration will declare an
4424 overload rather than conflict with an earlier declaration.
4425 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
4427 @item -Wmissing-declarations
4428 @opindex Wmissing-declarations
4429 @opindex Wno-missing-declarations
4430 Warn if a global function is defined without a previous declaration.
4431 Do so even if the definition itself provides a prototype.
4432 Use this option to detect global functions that are not declared in
4433 header files. In C, no warnings are issued for functions with previous
4434 non-prototype declarations; use @option{-Wmissing-prototype} to detect
4435 missing prototypes. In C++, no warnings are issued for function templates,
4436 or for inline functions, or for functions in anonymous namespaces.
4438 @item -Wmissing-field-initializers
4439 @opindex Wmissing-field-initializers
4440 @opindex Wno-missing-field-initializers
4444 Warn if a structure's initializer has some fields missing. For
4445 example, the following code causes such a warning, because
4446 @code{x.h} is implicitly zero:
4449 struct s @{ int f, g, h; @};
4450 struct s x = @{ 3, 4 @};
4453 This option does not warn about designated initializers, so the following
4454 modification does not trigger a warning:
4457 struct s @{ int f, g, h; @};
4458 struct s x = @{ .f = 3, .g = 4 @};
4461 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4462 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
4464 @item -Wno-multichar
4465 @opindex Wno-multichar
4467 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4468 Usually they indicate a typo in the user's code, as they have
4469 implementation-defined values, and should not be used in portable code.
4471 @item -Wnormalized=<none|id|nfc|nfkc>
4472 @opindex Wnormalized=
4475 @cindex character set, input normalization
4476 In ISO C and ISO C++, two identifiers are different if they are
4477 different sequences of characters. However, sometimes when characters
4478 outside the basic ASCII character set are used, you can have two
4479 different character sequences that look the same. To avoid confusion,
4480 the ISO 10646 standard sets out some @dfn{normalization rules} which
4481 when applied ensure that two sequences that look the same are turned into
4482 the same sequence. GCC can warn you if you are using identifiers that
4483 have not been normalized; this option controls that warning.
4485 There are four levels of warning supported by GCC@. The default is
4486 @option{-Wnormalized=nfc}, which warns about any identifier that is
4487 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4488 recommended form for most uses.
4490 Unfortunately, there are some characters allowed in identifiers by
4491 ISO C and ISO C++ that, when turned into NFC, are not allowed in
4492 identifiers. That is, there's no way to use these symbols in portable
4493 ISO C or C++ and have all your identifiers in NFC@.
4494 @option{-Wnormalized=id} suppresses the warning for these characters.
4495 It is hoped that future versions of the standards involved will correct
4496 this, which is why this option is not the default.
4498 You can switch the warning off for all characters by writing
4499 @option{-Wnormalized=none}. You should only do this if you
4500 are using some other normalization scheme (like ``D''), because
4501 otherwise you can easily create bugs that are literally impossible to see.
4503 Some characters in ISO 10646 have distinct meanings but look identical
4504 in some fonts or display methodologies, especially once formatting has
4505 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4506 LETTER N'', displays just like a regular @code{n} that has been
4507 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4508 normalization scheme to convert all these into a standard form as
4509 well, and GCC warns if your code is not in NFKC if you use
4510 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4511 about every identifier that contains the letter O because it might be
4512 confused with the digit 0, and so is not the default, but may be
4513 useful as a local coding convention if the programming environment
4514 cannot be fixed to display these characters distinctly.
4516 @item -Wno-deprecated
4517 @opindex Wno-deprecated
4518 @opindex Wdeprecated
4519 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4521 @item -Wno-deprecated-declarations
4522 @opindex Wno-deprecated-declarations
4523 @opindex Wdeprecated-declarations
4524 Do not warn about uses of functions (@pxref{Function Attributes}),
4525 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4526 Attributes}) marked as deprecated by using the @code{deprecated}
4530 @opindex Wno-overflow
4532 Do not warn about compile-time overflow in constant expressions.
4534 @item -Woverride-init @r{(C and Objective-C only)}
4535 @opindex Woverride-init
4536 @opindex Wno-override-init
4540 Warn if an initialized field without side effects is overridden when
4541 using designated initializers (@pxref{Designated Inits, , Designated
4544 This warning is included in @option{-Wextra}. To get other
4545 @option{-Wextra} warnings without this one, use @option{-Wextra
4546 -Wno-override-init}.
4551 Warn if a structure is given the packed attribute, but the packed
4552 attribute has no effect on the layout or size of the structure.
4553 Such structures may be mis-aligned for little benefit. For
4554 instance, in this code, the variable @code{f.x} in @code{struct bar}
4555 is misaligned even though @code{struct bar} does not itself
4556 have the packed attribute:
4563 @} __attribute__((packed));
4571 @item -Wpacked-bitfield-compat
4572 @opindex Wpacked-bitfield-compat
4573 @opindex Wno-packed-bitfield-compat
4574 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4575 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4576 the change can lead to differences in the structure layout. GCC
4577 informs you when the offset of such a field has changed in GCC 4.4.
4578 For example there is no longer a 4-bit padding between field @code{a}
4579 and @code{b} in this structure:
4586 @} __attribute__ ((packed));
4589 This warning is enabled by default. Use
4590 @option{-Wno-packed-bitfield-compat} to disable this warning.
4595 Warn if padding is included in a structure, either to align an element
4596 of the structure or to align the whole structure. Sometimes when this
4597 happens it is possible to rearrange the fields of the structure to
4598 reduce the padding and so make the structure smaller.
4600 @item -Wredundant-decls
4601 @opindex Wredundant-decls
4602 @opindex Wno-redundant-decls
4603 Warn if anything is declared more than once in the same scope, even in
4604 cases where multiple declaration is valid and changes nothing.
4606 @item -Wnested-externs @r{(C and Objective-C only)}
4607 @opindex Wnested-externs
4608 @opindex Wno-nested-externs
4609 Warn if an @code{extern} declaration is encountered within a function.
4614 Warn if a function that is declared as inline cannot be inlined.
4615 Even with this option, the compiler does not warn about failures to
4616 inline functions declared in system headers.
4618 The compiler uses a variety of heuristics to determine whether or not
4619 to inline a function. For example, the compiler takes into account
4620 the size of the function being inlined and the amount of inlining
4621 that has already been done in the current function. Therefore,
4622 seemingly insignificant changes in the source program can cause the
4623 warnings produced by @option{-Winline} to appear or disappear.
4625 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4626 @opindex Wno-invalid-offsetof
4627 @opindex Winvalid-offsetof
4628 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4629 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4630 to a non-POD type is undefined. In existing C++ implementations,
4631 however, @samp{offsetof} typically gives meaningful results even when
4632 applied to certain kinds of non-POD types. (Such as a simple
4633 @samp{struct} that fails to be a POD type only by virtue of having a
4634 constructor.) This flag is for users who are aware that they are
4635 writing nonportable code and who have deliberately chosen to ignore the
4638 The restrictions on @samp{offsetof} may be relaxed in a future version
4639 of the C++ standard.
4641 @item -Wno-int-to-pointer-cast
4642 @opindex Wno-int-to-pointer-cast
4643 @opindex Wint-to-pointer-cast
4644 Suppress warnings from casts to pointer type of an integer of a
4645 different size. In C++, casting to a pointer type of smaller size is
4646 an error. @option{Wint-to-pointer-cast} is enabled by default.
4649 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4650 @opindex Wno-pointer-to-int-cast
4651 @opindex Wpointer-to-int-cast
4652 Suppress warnings from casts from a pointer to an integer type of a
4656 @opindex Winvalid-pch
4657 @opindex Wno-invalid-pch
4658 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4659 the search path but can't be used.
4663 @opindex Wno-long-long
4664 Warn if @samp{long long} type is used. This is enabled by either
4665 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
4666 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4668 @item -Wvariadic-macros
4669 @opindex Wvariadic-macros
4670 @opindex Wno-variadic-macros
4671 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4672 alternate syntax when in pedantic ISO C99 mode. This is default.
4673 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4677 @opindex Wno-varargs
4678 Warn upon questionable usage of the macros used to handle variable
4679 arguments like @samp{va_start}. This is default. To inhibit the
4680 warning messages, use @option{-Wno-varargs}.
4682 @item -Wvector-operation-performance
4683 @opindex Wvector-operation-performance
4684 @opindex Wno-vector-operation-performance
4685 Warn if vector operation is not implemented via SIMD capabilities of the
4686 architecture. Mainly useful for the performance tuning.
4687 Vector operation can be implemented @code{piecewise}, which means that the
4688 scalar operation is performed on every vector element;
4689 @code{in parallel}, which means that the vector operation is implemented
4690 using scalars of wider type, which normally is more performance efficient;
4691 and @code{as a single scalar}, which means that vector fits into a
4697 Warn if variable length array is used in the code.
4698 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
4699 the variable length array.
4701 @item -Wvolatile-register-var
4702 @opindex Wvolatile-register-var
4703 @opindex Wno-volatile-register-var
4704 Warn if a register variable is declared volatile. The volatile
4705 modifier does not inhibit all optimizations that may eliminate reads
4706 and/or writes to register variables. This warning is enabled by
4709 @item -Wdisabled-optimization
4710 @opindex Wdisabled-optimization
4711 @opindex Wno-disabled-optimization
4712 Warn if a requested optimization pass is disabled. This warning does
4713 not generally indicate that there is anything wrong with your code; it
4714 merely indicates that GCC's optimizers are unable to handle the code
4715 effectively. Often, the problem is that your code is too big or too
4716 complex; GCC refuses to optimize programs when the optimization
4717 itself is likely to take inordinate amounts of time.
4719 @item -Wpointer-sign @r{(C and Objective-C only)}
4720 @opindex Wpointer-sign
4721 @opindex Wno-pointer-sign
4722 Warn for pointer argument passing or assignment with different signedness.
4723 This option is only supported for C and Objective-C@. It is implied by
4724 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
4725 @option{-Wno-pointer-sign}.
4727 @item -Wstack-protector
4728 @opindex Wstack-protector
4729 @opindex Wno-stack-protector
4730 This option is only active when @option{-fstack-protector} is active. It
4731 warns about functions that are not protected against stack smashing.
4734 @opindex Wno-mudflap
4735 Suppress warnings about constructs that cannot be instrumented by
4738 @item -Woverlength-strings
4739 @opindex Woverlength-strings
4740 @opindex Wno-overlength-strings
4741 Warn about string constants that are longer than the ``minimum
4742 maximum'' length specified in the C standard. Modern compilers
4743 generally allow string constants that are much longer than the
4744 standard's minimum limit, but very portable programs should avoid
4745 using longer strings.
4747 The limit applies @emph{after} string constant concatenation, and does
4748 not count the trailing NUL@. In C90, the limit was 509 characters; in
4749 C99, it was raised to 4095. C++98 does not specify a normative
4750 minimum maximum, so we do not diagnose overlength strings in C++@.
4752 This option is implied by @option{-Wpedantic}, and can be disabled with
4753 @option{-Wno-overlength-strings}.
4755 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4756 @opindex Wunsuffixed-float-constants
4758 Issue a warning for any floating constant that does not have
4759 a suffix. When used together with @option{-Wsystem-headers} it
4760 warns about such constants in system header files. This can be useful
4761 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4762 from the decimal floating-point extension to C99.
4765 @node Debugging Options
4766 @section Options for Debugging Your Program or GCC
4767 @cindex options, debugging
4768 @cindex debugging information options
4770 GCC has various special options that are used for debugging
4771 either your program or GCC:
4776 Produce debugging information in the operating system's native format
4777 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4780 On most systems that use stabs format, @option{-g} enables use of extra
4781 debugging information that only GDB can use; this extra information
4782 makes debugging work better in GDB but probably makes other debuggers
4784 refuse to read the program. If you want to control for certain whether
4785 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4786 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4788 GCC allows you to use @option{-g} with
4789 @option{-O}. The shortcuts taken by optimized code may occasionally
4790 produce surprising results: some variables you declared may not exist
4791 at all; flow of control may briefly move where you did not expect it;
4792 some statements may not be executed because they compute constant
4793 results or their values are already at hand; some statements may
4794 execute in different places because they have been moved out of loops.
4796 Nevertheless it proves possible to debug optimized output. This makes
4797 it reasonable to use the optimizer for programs that might have bugs.
4799 The following options are useful when GCC is generated with the
4800 capability for more than one debugging format.
4804 Produce debugging information for use by GDB@. This means to use the
4805 most expressive format available (DWARF 2, stabs, or the native format
4806 if neither of those are supported), including GDB extensions if at all
4811 Produce debugging information in stabs format (if that is supported),
4812 without GDB extensions. This is the format used by DBX on most BSD
4813 systems. On MIPS, Alpha and System V Release 4 systems this option
4814 produces stabs debugging output that is not understood by DBX or SDB@.
4815 On System V Release 4 systems this option requires the GNU assembler.
4817 @item -feliminate-unused-debug-symbols
4818 @opindex feliminate-unused-debug-symbols
4819 Produce debugging information in stabs format (if that is supported),
4820 for only symbols that are actually used.
4822 @item -femit-class-debug-always
4823 Instead of emitting debugging information for a C++ class in only one
4824 object file, emit it in all object files using the class. This option
4825 should be used only with debuggers that are unable to handle the way GCC
4826 normally emits debugging information for classes because using this
4827 option increases the size of debugging information by as much as a
4830 @item -fdebug-types-section
4831 @opindex fdebug-types-section
4832 @opindex fno-debug-types-section
4833 When using DWARF Version 4 or higher, type DIEs can be put into
4834 their own @code{.debug_types} section instead of making them part of the
4835 @code{.debug_info} section. It is more efficient to put them in a separate
4836 comdat sections since the linker can then remove duplicates.
4837 But not all DWARF consumers support @code{.debug_types} sections yet
4838 and on some objects @code{.debug_types} produces larger instead of smaller
4839 debugging information.
4843 Produce debugging information in stabs format (if that is supported),
4844 using GNU extensions understood only by the GNU debugger (GDB)@. The
4845 use of these extensions is likely to make other debuggers crash or
4846 refuse to read the program.
4850 Produce debugging information in COFF format (if that is supported).
4851 This is the format used by SDB on most System V systems prior to
4856 Produce debugging information in XCOFF format (if that is supported).
4857 This is the format used by the DBX debugger on IBM RS/6000 systems.
4861 Produce debugging information in XCOFF format (if that is supported),
4862 using GNU extensions understood only by the GNU debugger (GDB)@. The
4863 use of these extensions is likely to make other debuggers crash or
4864 refuse to read the program, and may cause assemblers other than the GNU
4865 assembler (GAS) to fail with an error.
4867 @item -gdwarf-@var{version}
4868 @opindex gdwarf-@var{version}
4869 Produce debugging information in DWARF format (if that is supported).
4870 The value of @var{version} may be either 2, 3 or 4; the default version
4871 for most targets is 4.
4873 Note that with DWARF Version 2, some ports require and always
4874 use some non-conflicting DWARF 3 extensions in the unwind tables.
4876 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4877 for maximum benefit.
4879 @item -grecord-gcc-switches
4880 @opindex grecord-gcc-switches
4881 This switch causes the command-line options used to invoke the
4882 compiler that may affect code generation to be appended to the
4883 DW_AT_producer attribute in DWARF debugging information. The options
4884 are concatenated with spaces separating them from each other and from
4885 the compiler version. See also @option{-frecord-gcc-switches} for another
4886 way of storing compiler options into the object file. This is the default.
4888 @item -gno-record-gcc-switches
4889 @opindex gno-record-gcc-switches
4890 Disallow appending command-line options to the DW_AT_producer attribute
4891 in DWARF debugging information.
4893 @item -gstrict-dwarf
4894 @opindex gstrict-dwarf
4895 Disallow using extensions of later DWARF standard version than selected
4896 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4897 DWARF extensions from later standard versions is allowed.
4899 @item -gno-strict-dwarf
4900 @opindex gno-strict-dwarf
4901 Allow using extensions of later DWARF standard version than selected with
4902 @option{-gdwarf-@var{version}}.
4906 Produce debugging information in Alpha/VMS debug format (if that is
4907 supported). This is the format used by DEBUG on Alpha/VMS systems.
4910 @itemx -ggdb@var{level}
4911 @itemx -gstabs@var{level}
4912 @itemx -gcoff@var{level}
4913 @itemx -gxcoff@var{level}
4914 @itemx -gvms@var{level}
4915 Request debugging information and also use @var{level} to specify how
4916 much information. The default level is 2.
4918 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4921 Level 1 produces minimal information, enough for making backtraces in
4922 parts of the program that you don't plan to debug. This includes
4923 descriptions of functions and external variables, but no information
4924 about local variables and no line numbers.
4926 Level 3 includes extra information, such as all the macro definitions
4927 present in the program. Some debuggers support macro expansion when
4928 you use @option{-g3}.
4930 @option{-gdwarf-2} does not accept a concatenated debug level, because
4931 GCC used to support an option @option{-gdwarf} that meant to generate
4932 debug information in version 1 of the DWARF format (which is very
4933 different from version 2), and it would have been too confusing. That
4934 debug format is long obsolete, but the option cannot be changed now.
4935 Instead use an additional @option{-g@var{level}} option to change the
4936 debug level for DWARF.
4940 Turn off generation of debug info, if leaving out this option
4941 generates it, or turn it on at level 2 otherwise. The position of this
4942 argument in the command line does not matter; it takes effect after all
4943 other options are processed, and it does so only once, no matter how
4944 many times it is given. This is mainly intended to be used with
4945 @option{-fcompare-debug}.
4947 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4948 @opindex fdump-final-insns
4949 Dump the final internal representation (RTL) to @var{file}. If the
4950 optional argument is omitted (or if @var{file} is @code{.}), the name
4951 of the dump file is determined by appending @code{.gkd} to the
4952 compilation output file name.
4954 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4955 @opindex fcompare-debug
4956 @opindex fno-compare-debug
4957 If no error occurs during compilation, run the compiler a second time,
4958 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4959 passed to the second compilation. Dump the final internal
4960 representation in both compilations, and print an error if they differ.
4962 If the equal sign is omitted, the default @option{-gtoggle} is used.
4964 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4965 and nonzero, implicitly enables @option{-fcompare-debug}. If
4966 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4967 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4970 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4971 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4972 of the final representation and the second compilation, preventing even
4973 @env{GCC_COMPARE_DEBUG} from taking effect.
4975 To verify full coverage during @option{-fcompare-debug} testing, set
4976 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4977 which GCC rejects as an invalid option in any actual compilation
4978 (rather than preprocessing, assembly or linking). To get just a
4979 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4980 not overridden} will do.
4982 @item -fcompare-debug-second
4983 @opindex fcompare-debug-second
4984 This option is implicitly passed to the compiler for the second
4985 compilation requested by @option{-fcompare-debug}, along with options to
4986 silence warnings, and omitting other options that would cause
4987 side-effect compiler outputs to files or to the standard output. Dump
4988 files and preserved temporary files are renamed so as to contain the
4989 @code{.gk} additional extension during the second compilation, to avoid
4990 overwriting those generated by the first.
4992 When this option is passed to the compiler driver, it causes the
4993 @emph{first} compilation to be skipped, which makes it useful for little
4994 other than debugging the compiler proper.
4996 @item -feliminate-dwarf2-dups
4997 @opindex feliminate-dwarf2-dups
4998 Compress DWARF 2 debugging information by eliminating duplicated
4999 information about each symbol. This option only makes sense when
5000 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5002 @item -femit-struct-debug-baseonly
5003 Emit debug information for struct-like types
5004 only when the base name of the compilation source file
5005 matches the base name of file in which the struct is defined.
5007 This option substantially reduces the size of debugging information,
5008 but at significant potential loss in type information to the debugger.
5009 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5010 See @option{-femit-struct-debug-detailed} for more detailed control.
5012 This option works only with DWARF 2.
5014 @item -femit-struct-debug-reduced
5015 Emit debug information for struct-like types
5016 only when the base name of the compilation source file
5017 matches the base name of file in which the type is defined,
5018 unless the struct is a template or defined in a system header.
5020 This option significantly reduces the size of debugging information,
5021 with some potential loss in type information to the debugger.
5022 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5023 See @option{-femit-struct-debug-detailed} for more detailed control.
5025 This option works only with DWARF 2.
5027 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5028 Specify the struct-like types
5029 for which the compiler generates debug information.
5030 The intent is to reduce duplicate struct debug information
5031 between different object files within the same program.
5033 This option is a detailed version of
5034 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5035 which serves for most needs.
5037 A specification has the syntax@*
5038 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5040 The optional first word limits the specification to
5041 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5042 A struct type is used directly when it is the type of a variable, member.
5043 Indirect uses arise through pointers to structs.
5044 That is, when use of an incomplete struct is valid, the use is indirect.
5046 @samp{struct one direct; struct two * indirect;}.
5048 The optional second word limits the specification to
5049 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5050 Generic structs are a bit complicated to explain.
5051 For C++, these are non-explicit specializations of template classes,
5052 or non-template classes within the above.
5053 Other programming languages have generics,
5054 but @option{-femit-struct-debug-detailed} does not yet implement them.
5056 The third word specifies the source files for those
5057 structs for which the compiler should emit debug information.
5058 The values @samp{none} and @samp{any} have the normal meaning.
5059 The value @samp{base} means that
5060 the base of name of the file in which the type declaration appears
5061 must match the base of the name of the main compilation file.
5062 In practice, this means that when compiling @file{foo.c}, debug information
5063 is generated for types declared in that file and @file{foo.h},
5064 but not other header files.
5065 The value @samp{sys} means those types satisfying @samp{base}
5066 or declared in system or compiler headers.
5068 You may need to experiment to determine the best settings for your application.
5070 The default is @option{-femit-struct-debug-detailed=all}.
5072 This option works only with DWARF 2.
5074 @item -fno-merge-debug-strings
5075 @opindex fmerge-debug-strings
5076 @opindex fno-merge-debug-strings
5077 Direct the linker to not merge together strings in the debugging
5078 information that are identical in different object files. Merging is
5079 not supported by all assemblers or linkers. Merging decreases the size
5080 of the debug information in the output file at the cost of increasing
5081 link processing time. Merging is enabled by default.
5083 @item -fdebug-prefix-map=@var{old}=@var{new}
5084 @opindex fdebug-prefix-map
5085 When compiling files in directory @file{@var{old}}, record debugging
5086 information describing them as in @file{@var{new}} instead.
5088 @item -fno-dwarf2-cfi-asm
5089 @opindex fdwarf2-cfi-asm
5090 @opindex fno-dwarf2-cfi-asm
5091 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5092 instead of using GAS @code{.cfi_*} directives.
5094 @cindex @command{prof}
5097 Generate extra code to write profile information suitable for the
5098 analysis program @command{prof}. You must use this option when compiling
5099 the source files you want data about, and you must also use it when
5102 @cindex @command{gprof}
5105 Generate extra code to write profile information suitable for the
5106 analysis program @command{gprof}. You must use this option when compiling
5107 the source files you want data about, and you must also use it when
5112 Makes the compiler print out each function name as it is compiled, and
5113 print some statistics about each pass when it finishes.
5116 @opindex ftime-report
5117 Makes the compiler print some statistics about the time consumed by each
5118 pass when it finishes.
5121 @opindex fmem-report
5122 Makes the compiler print some statistics about permanent memory
5123 allocation when it finishes.
5125 @item -fpre-ipa-mem-report
5126 @opindex fpre-ipa-mem-report
5127 @item -fpost-ipa-mem-report
5128 @opindex fpost-ipa-mem-report
5129 Makes the compiler print some statistics about permanent memory
5130 allocation before or after interprocedural optimization.
5133 @opindex fstack-usage
5134 Makes the compiler output stack usage information for the program, on a
5135 per-function basis. The filename for the dump is made by appending
5136 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5137 the output file, if explicitly specified and it is not an executable,
5138 otherwise it is the basename of the source file. An entry is made up
5143 The name of the function.
5147 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5150 The qualifier @code{static} means that the function manipulates the stack
5151 statically: a fixed number of bytes are allocated for the frame on function
5152 entry and released on function exit; no stack adjustments are otherwise made
5153 in the function. The second field is this fixed number of bytes.
5155 The qualifier @code{dynamic} means that the function manipulates the stack
5156 dynamically: in addition to the static allocation described above, stack
5157 adjustments are made in the body of the function, for example to push/pop
5158 arguments around function calls. If the qualifier @code{bounded} is also
5159 present, the amount of these adjustments is bounded at compile time and
5160 the second field is an upper bound of the total amount of stack used by
5161 the function. If it is not present, the amount of these adjustments is
5162 not bounded at compile time and the second field only represents the
5165 @item -fprofile-arcs
5166 @opindex fprofile-arcs
5167 Add code so that program flow @dfn{arcs} are instrumented. During
5168 execution the program records how many times each branch and call is
5169 executed and how many times it is taken or returns. When the compiled
5170 program exits it saves this data to a file called
5171 @file{@var{auxname}.gcda} for each source file. The data may be used for
5172 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5173 test coverage analysis (@option{-ftest-coverage}). Each object file's
5174 @var{auxname} is generated from the name of the output file, if
5175 explicitly specified and it is not the final executable, otherwise it is
5176 the basename of the source file. In both cases any suffix is removed
5177 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5178 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5179 @xref{Cross-profiling}.
5181 @cindex @command{gcov}
5185 This option is used to compile and link code instrumented for coverage
5186 analysis. The option is a synonym for @option{-fprofile-arcs}
5187 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5188 linking). See the documentation for those options for more details.
5193 Compile the source files with @option{-fprofile-arcs} plus optimization
5194 and code generation options. For test coverage analysis, use the
5195 additional @option{-ftest-coverage} option. You do not need to profile
5196 every source file in a program.
5199 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5200 (the latter implies the former).
5203 Run the program on a representative workload to generate the arc profile
5204 information. This may be repeated any number of times. You can run
5205 concurrent instances of your program, and provided that the file system
5206 supports locking, the data files will be correctly updated. Also
5207 @code{fork} calls are detected and correctly handled (double counting
5211 For profile-directed optimizations, compile the source files again with
5212 the same optimization and code generation options plus
5213 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5214 Control Optimization}).
5217 For test coverage analysis, use @command{gcov} to produce human readable
5218 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5219 @command{gcov} documentation for further information.
5223 With @option{-fprofile-arcs}, for each function of your program GCC
5224 creates a program flow graph, then finds a spanning tree for the graph.
5225 Only arcs that are not on the spanning tree have to be instrumented: the
5226 compiler adds code to count the number of times that these arcs are
5227 executed. When an arc is the only exit or only entrance to a block, the
5228 instrumentation code can be added to the block; otherwise, a new basic
5229 block must be created to hold the instrumentation code.
5232 @item -ftest-coverage
5233 @opindex ftest-coverage
5234 Produce a notes file that the @command{gcov} code-coverage utility
5235 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5236 show program coverage. Each source file's note file is called
5237 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5238 above for a description of @var{auxname} and instructions on how to
5239 generate test coverage data. Coverage data matches the source files
5240 more closely if you do not optimize.
5242 @item -fdbg-cnt-list
5243 @opindex fdbg-cnt-list
5244 Print the name and the counter upper bound for all debug counters.
5247 @item -fdbg-cnt=@var{counter-value-list}
5249 Set the internal debug counter upper bound. @var{counter-value-list}
5250 is a comma-separated list of @var{name}:@var{value} pairs
5251 which sets the upper bound of each debug counter @var{name} to @var{value}.
5252 All debug counters have the initial upper bound of @code{UINT_MAX};
5253 thus @code{dbg_cnt()} returns true always unless the upper bound
5254 is set by this option.
5255 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
5256 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
5258 @itemx -fenable-@var{kind}-@var{pass}
5259 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5263 This is a set of options that are used to explicitly disable/enable
5264 optimization passes. These options are intended for use for debugging GCC.
5265 Compiler users should use regular options for enabling/disabling
5270 @item -fdisable-ipa-@var{pass}
5271 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
5272 statically invoked in the compiler multiple times, the pass name should be
5273 appended with a sequential number starting from 1.
5275 @item -fdisable-rtl-@var{pass}
5276 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
5277 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
5278 statically invoked in the compiler multiple times, the pass name should be
5279 appended with a sequential number starting from 1. @var{range-list} is a
5280 comma-separated list of function ranges or assembler names. Each range is a number
5281 pair separated by a colon. The range is inclusive in both ends. If the range
5282 is trivial, the number pair can be simplified as a single number. If the
5283 function's cgraph node's @var{uid} falls within one of the specified ranges,
5284 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5285 function header of a dump file, and the pass names can be dumped by using
5286 option @option{-fdump-passes}.
5288 @item -fdisable-tree-@var{pass}
5289 @itemx -fdisable-tree-@var{pass}=@var{range-list}
5290 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5293 @item -fenable-ipa-@var{pass}
5294 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
5295 statically invoked in the compiler multiple times, the pass name should be
5296 appended with a sequential number starting from 1.
5298 @item -fenable-rtl-@var{pass}
5299 @itemx -fenable-rtl-@var{pass}=@var{range-list}
5300 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
5301 description and examples.
5303 @item -fenable-tree-@var{pass}
5304 @itemx -fenable-tree-@var{pass}=@var{range-list}
5305 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5306 of option arguments.
5310 Here are some examples showing uses of these options.
5314 # disable ccp1 for all functions
5316 # disable complete unroll for function whose cgraph node uid is 1
5317 -fenable-tree-cunroll=1
5318 # disable gcse2 for functions at the following ranges [1,1],
5319 # [300,400], and [400,1000]
5320 # disable gcse2 for functions foo and foo2
5321 -fdisable-rtl-gcse2=foo,foo2
5322 # disable early inlining
5323 -fdisable-tree-einline
5324 # disable ipa inlining
5325 -fdisable-ipa-inline
5326 # enable tree full unroll
5327 -fenable-tree-unroll
5331 @item -d@var{letters}
5332 @itemx -fdump-rtl-@var{pass}
5334 Says to make debugging dumps during compilation at times specified by
5335 @var{letters}. This is used for debugging the RTL-based passes of the
5336 compiler. The file names for most of the dumps are made by appending
5337 a pass number and a word to the @var{dumpname}, and the files are
5338 created in the directory of the output file. Note that the pass
5339 number is computed statically as passes get registered into the pass
5340 manager. Thus the numbering is not related to the dynamic order of
5341 execution of passes. In particular, a pass installed by a plugin
5342 could have a number over 200 even if it executed quite early.
5343 @var{dumpname} is generated from the name of the output file, if
5344 explicitly specified and it is not an executable, otherwise it is the
5345 basename of the source file. These switches may have different effects
5346 when @option{-E} is used for preprocessing.
5348 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5349 @option{-d} option @var{letters}. Here are the possible
5350 letters for use in @var{pass} and @var{letters}, and their meanings:
5354 @item -fdump-rtl-alignments
5355 @opindex fdump-rtl-alignments
5356 Dump after branch alignments have been computed.
5358 @item -fdump-rtl-asmcons
5359 @opindex fdump-rtl-asmcons
5360 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5362 @item -fdump-rtl-auto_inc_dec
5363 @opindex fdump-rtl-auto_inc_dec
5364 Dump after auto-inc-dec discovery. This pass is only run on
5365 architectures that have auto inc or auto dec instructions.
5367 @item -fdump-rtl-barriers
5368 @opindex fdump-rtl-barriers
5369 Dump after cleaning up the barrier instructions.
5371 @item -fdump-rtl-bbpart
5372 @opindex fdump-rtl-bbpart
5373 Dump after partitioning hot and cold basic blocks.
5375 @item -fdump-rtl-bbro
5376 @opindex fdump-rtl-bbro
5377 Dump after block reordering.
5379 @item -fdump-rtl-btl1
5380 @itemx -fdump-rtl-btl2
5381 @opindex fdump-rtl-btl2
5382 @opindex fdump-rtl-btl2
5383 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5384 after the two branch
5385 target load optimization passes.
5387 @item -fdump-rtl-bypass
5388 @opindex fdump-rtl-bypass
5389 Dump after jump bypassing and control flow optimizations.
5391 @item -fdump-rtl-combine
5392 @opindex fdump-rtl-combine
5393 Dump after the RTL instruction combination pass.
5395 @item -fdump-rtl-compgotos
5396 @opindex fdump-rtl-compgotos
5397 Dump after duplicating the computed gotos.
5399 @item -fdump-rtl-ce1
5400 @itemx -fdump-rtl-ce2
5401 @itemx -fdump-rtl-ce3
5402 @opindex fdump-rtl-ce1
5403 @opindex fdump-rtl-ce2
5404 @opindex fdump-rtl-ce3
5405 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5406 @option{-fdump-rtl-ce3} enable dumping after the three
5407 if conversion passes.
5409 @itemx -fdump-rtl-cprop_hardreg
5410 @opindex fdump-rtl-cprop_hardreg
5411 Dump after hard register copy propagation.
5413 @itemx -fdump-rtl-csa
5414 @opindex fdump-rtl-csa
5415 Dump after combining stack adjustments.
5417 @item -fdump-rtl-cse1
5418 @itemx -fdump-rtl-cse2
5419 @opindex fdump-rtl-cse1
5420 @opindex fdump-rtl-cse2
5421 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5422 the two common sub-expression elimination passes.
5424 @itemx -fdump-rtl-dce
5425 @opindex fdump-rtl-dce
5426 Dump after the standalone dead code elimination passes.
5428 @itemx -fdump-rtl-dbr
5429 @opindex fdump-rtl-dbr
5430 Dump after delayed branch scheduling.
5432 @item -fdump-rtl-dce1
5433 @itemx -fdump-rtl-dce2
5434 @opindex fdump-rtl-dce1
5435 @opindex fdump-rtl-dce2
5436 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5437 the two dead store elimination passes.
5440 @opindex fdump-rtl-eh
5441 Dump after finalization of EH handling code.
5443 @item -fdump-rtl-eh_ranges
5444 @opindex fdump-rtl-eh_ranges
5445 Dump after conversion of EH handling range regions.
5447 @item -fdump-rtl-expand
5448 @opindex fdump-rtl-expand
5449 Dump after RTL generation.
5451 @item -fdump-rtl-fwprop1
5452 @itemx -fdump-rtl-fwprop2
5453 @opindex fdump-rtl-fwprop1
5454 @opindex fdump-rtl-fwprop2
5455 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5456 dumping after the two forward propagation passes.
5458 @item -fdump-rtl-gcse1
5459 @itemx -fdump-rtl-gcse2
5460 @opindex fdump-rtl-gcse1
5461 @opindex fdump-rtl-gcse2
5462 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5463 after global common subexpression elimination.
5465 @item -fdump-rtl-init-regs
5466 @opindex fdump-rtl-init-regs
5467 Dump after the initialization of the registers.
5469 @item -fdump-rtl-initvals
5470 @opindex fdump-rtl-initvals
5471 Dump after the computation of the initial value sets.
5473 @itemx -fdump-rtl-into_cfglayout
5474 @opindex fdump-rtl-into_cfglayout
5475 Dump after converting to cfglayout mode.
5477 @item -fdump-rtl-ira
5478 @opindex fdump-rtl-ira
5479 Dump after iterated register allocation.
5481 @item -fdump-rtl-jump
5482 @opindex fdump-rtl-jump
5483 Dump after the second jump optimization.
5485 @item -fdump-rtl-loop2
5486 @opindex fdump-rtl-loop2
5487 @option{-fdump-rtl-loop2} enables dumping after the rtl
5488 loop optimization passes.
5490 @item -fdump-rtl-mach
5491 @opindex fdump-rtl-mach
5492 Dump after performing the machine dependent reorganization pass, if that
5495 @item -fdump-rtl-mode_sw
5496 @opindex fdump-rtl-mode_sw
5497 Dump after removing redundant mode switches.
5499 @item -fdump-rtl-rnreg
5500 @opindex fdump-rtl-rnreg
5501 Dump after register renumbering.
5503 @itemx -fdump-rtl-outof_cfglayout
5504 @opindex fdump-rtl-outof_cfglayout
5505 Dump after converting from cfglayout mode.
5507 @item -fdump-rtl-peephole2
5508 @opindex fdump-rtl-peephole2
5509 Dump after the peephole pass.
5511 @item -fdump-rtl-postreload
5512 @opindex fdump-rtl-postreload
5513 Dump after post-reload optimizations.
5515 @itemx -fdump-rtl-pro_and_epilogue
5516 @opindex fdump-rtl-pro_and_epilogue
5517 Dump after generating the function prologues and epilogues.
5519 @item -fdump-rtl-regmove
5520 @opindex fdump-rtl-regmove
5521 Dump after the register move pass.
5523 @item -fdump-rtl-sched1
5524 @itemx -fdump-rtl-sched2
5525 @opindex fdump-rtl-sched1
5526 @opindex fdump-rtl-sched2
5527 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5528 after the basic block scheduling passes.
5530 @item -fdump-rtl-see
5531 @opindex fdump-rtl-see
5532 Dump after sign extension elimination.
5534 @item -fdump-rtl-seqabstr
5535 @opindex fdump-rtl-seqabstr
5536 Dump after common sequence discovery.
5538 @item -fdump-rtl-shorten
5539 @opindex fdump-rtl-shorten
5540 Dump after shortening branches.
5542 @item -fdump-rtl-sibling
5543 @opindex fdump-rtl-sibling
5544 Dump after sibling call optimizations.
5546 @item -fdump-rtl-split1
5547 @itemx -fdump-rtl-split2
5548 @itemx -fdump-rtl-split3
5549 @itemx -fdump-rtl-split4
5550 @itemx -fdump-rtl-split5
5551 @opindex fdump-rtl-split1
5552 @opindex fdump-rtl-split2
5553 @opindex fdump-rtl-split3
5554 @opindex fdump-rtl-split4
5555 @opindex fdump-rtl-split5
5556 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5557 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5558 @option{-fdump-rtl-split5} enable dumping after five rounds of
5559 instruction splitting.
5561 @item -fdump-rtl-sms
5562 @opindex fdump-rtl-sms
5563 Dump after modulo scheduling. This pass is only run on some
5566 @item -fdump-rtl-stack
5567 @opindex fdump-rtl-stack
5568 Dump after conversion from GCC's ``flat register file'' registers to the
5569 x87's stack-like registers. This pass is only run on x86 variants.
5571 @item -fdump-rtl-subreg1
5572 @itemx -fdump-rtl-subreg2
5573 @opindex fdump-rtl-subreg1
5574 @opindex fdump-rtl-subreg2
5575 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5576 the two subreg expansion passes.
5578 @item -fdump-rtl-unshare
5579 @opindex fdump-rtl-unshare
5580 Dump after all rtl has been unshared.
5582 @item -fdump-rtl-vartrack
5583 @opindex fdump-rtl-vartrack
5584 Dump after variable tracking.
5586 @item -fdump-rtl-vregs
5587 @opindex fdump-rtl-vregs
5588 Dump after converting virtual registers to hard registers.
5590 @item -fdump-rtl-web
5591 @opindex fdump-rtl-web
5592 Dump after live range splitting.
5594 @item -fdump-rtl-regclass
5595 @itemx -fdump-rtl-subregs_of_mode_init
5596 @itemx -fdump-rtl-subregs_of_mode_finish
5597 @itemx -fdump-rtl-dfinit
5598 @itemx -fdump-rtl-dfinish
5599 @opindex fdump-rtl-regclass
5600 @opindex fdump-rtl-subregs_of_mode_init
5601 @opindex fdump-rtl-subregs_of_mode_finish
5602 @opindex fdump-rtl-dfinit
5603 @opindex fdump-rtl-dfinish
5604 These dumps are defined but always produce empty files.
5606 @item -fdump-rtl-all
5607 @opindex fdump-rtl-all
5608 Produce all the dumps listed above.
5612 Annotate the assembler output with miscellaneous debugging information.
5616 Dump all macro definitions, at the end of preprocessing, in addition to
5621 Produce a core dump whenever an error occurs.
5625 Print statistics on memory usage, at the end of the run, to
5630 Annotate the assembler output with a comment indicating which
5631 pattern and alternative is used. The length of each instruction is
5636 Dump the RTL in the assembler output as a comment before each instruction.
5637 Also turns on @option{-dp} annotation.
5641 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5642 dump a representation of the control flow graph suitable for viewing with VCG
5643 to @file{@var{file}.@var{pass}.vcg}.
5647 Just generate RTL for a function instead of compiling it. Usually used
5648 with @option{-fdump-rtl-expand}.
5652 @opindex fdump-noaddr
5653 When doing debugging dumps, suppress address output. This makes it more
5654 feasible to use diff on debugging dumps for compiler invocations with
5655 different compiler binaries and/or different
5656 text / bss / data / heap / stack / dso start locations.
5658 @item -fdump-unnumbered
5659 @opindex fdump-unnumbered
5660 When doing debugging dumps, suppress instruction numbers and address output.
5661 This makes it more feasible to use diff on debugging dumps for compiler
5662 invocations with different options, in particular with and without
5665 @item -fdump-unnumbered-links
5666 @opindex fdump-unnumbered-links
5667 When doing debugging dumps (see @option{-d} option above), suppress
5668 instruction numbers for the links to the previous and next instructions
5671 @item -fdump-translation-unit @r{(C++ only)}
5672 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5673 @opindex fdump-translation-unit
5674 Dump a representation of the tree structure for the entire translation
5675 unit to a file. The file name is made by appending @file{.tu} to the
5676 source file name, and the file is created in the same directory as the
5677 output file. If the @samp{-@var{options}} form is used, @var{options}
5678 controls the details of the dump as described for the
5679 @option{-fdump-tree} options.
5681 @item -fdump-class-hierarchy @r{(C++ only)}
5682 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5683 @opindex fdump-class-hierarchy
5684 Dump a representation of each class's hierarchy and virtual function
5685 table layout to a file. The file name is made by appending
5686 @file{.class} to the source file name, and the file is created in the
5687 same directory as the output file. If the @samp{-@var{options}} form
5688 is used, @var{options} controls the details of the dump as described
5689 for the @option{-fdump-tree} options.
5691 @item -fdump-ipa-@var{switch}
5693 Control the dumping at various stages of inter-procedural analysis
5694 language tree to a file. The file name is generated by appending a
5695 switch specific suffix to the source file name, and the file is created
5696 in the same directory as the output file. The following dumps are
5701 Enables all inter-procedural analysis dumps.
5704 Dumps information about call-graph optimization, unused function removal,
5705 and inlining decisions.
5708 Dump after function inlining.
5713 @opindex fdump-passes
5714 Dump the list of optimization passes that are turned on and off by
5715 the current command-line options.
5717 @item -fdump-statistics-@var{option}
5718 @opindex fdump-statistics
5719 Enable and control dumping of pass statistics in a separate file. The
5720 file name is generated by appending a suffix ending in
5721 @samp{.statistics} to the source file name, and the file is created in
5722 the same directory as the output file. If the @samp{-@var{option}}
5723 form is used, @samp{-stats} causes counters to be summed over the
5724 whole compilation unit while @samp{-details} dumps every event as
5725 the passes generate them. The default with no option is to sum
5726 counters for each function compiled.
5728 @item -fdump-tree-@var{switch}
5729 @itemx -fdump-tree-@var{switch}-@var{options}
5731 Control the dumping at various stages of processing the intermediate
5732 language tree to a file. The file name is generated by appending a
5733 switch specific suffix to the source file name, and the file is
5734 created in the same directory as the output file. If the
5735 @samp{-@var{options}} form is used, @var{options} is a list of
5736 @samp{-} separated options which control the details of the dump. Not
5737 all options are applicable to all dumps; those that are not
5738 meaningful are ignored. The following options are available
5742 Print the address of each node. Usually this is not meaningful as it
5743 changes according to the environment and source file. Its primary use
5744 is for tying up a dump file with a debug environment.
5746 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5747 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5748 use working backward from mangled names in the assembly file.
5750 Inhibit dumping of members of a scope or body of a function merely
5751 because that scope has been reached. Only dump such items when they
5752 are directly reachable by some other path. When dumping pretty-printed
5753 trees, this option inhibits dumping the bodies of control structures.
5755 Print a raw representation of the tree. By default, trees are
5756 pretty-printed into a C-like representation.
5758 Enable more detailed dumps (not honored by every dump option).
5760 Enable dumping various statistics about the pass (not honored by every dump
5763 Enable showing basic block boundaries (disabled in raw dumps).
5765 Enable showing virtual operands for every statement.
5767 Enable showing line numbers for statements.
5769 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5771 Enable showing the tree dump for each statement.
5773 Enable showing the EH region number holding each statement.
5775 Enable showing scalar evolution analysis details.
5777 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5778 and @option{lineno}.
5781 The following tree dumps are possible:
5785 @opindex fdump-tree-original
5786 Dump before any tree based optimization, to @file{@var{file}.original}.
5789 @opindex fdump-tree-optimized
5790 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5793 @opindex fdump-tree-gimple
5794 Dump each function before and after the gimplification pass to a file. The
5795 file name is made by appending @file{.gimple} to the source file name.
5798 @opindex fdump-tree-cfg
5799 Dump the control flow graph of each function to a file. The file name is
5800 made by appending @file{.cfg} to the source file name.
5803 @opindex fdump-tree-vcg
5804 Dump the control flow graph of each function to a file in VCG format. The
5805 file name is made by appending @file{.vcg} to the source file name. Note
5806 that if the file contains more than one function, the generated file cannot
5807 be used directly by VCG@. You must cut and paste each function's
5808 graph into its own separate file first.
5811 @opindex fdump-tree-ch
5812 Dump each function after copying loop headers. The file name is made by
5813 appending @file{.ch} to the source file name.
5816 @opindex fdump-tree-ssa
5817 Dump SSA related information to a file. The file name is made by appending
5818 @file{.ssa} to the source file name.
5821 @opindex fdump-tree-alias
5822 Dump aliasing information for each function. The file name is made by
5823 appending @file{.alias} to the source file name.
5826 @opindex fdump-tree-ccp
5827 Dump each function after CCP@. The file name is made by appending
5828 @file{.ccp} to the source file name.
5831 @opindex fdump-tree-storeccp
5832 Dump each function after STORE-CCP@. The file name is made by appending
5833 @file{.storeccp} to the source file name.
5836 @opindex fdump-tree-pre
5837 Dump trees after partial redundancy elimination. The file name is made
5838 by appending @file{.pre} to the source file name.
5841 @opindex fdump-tree-fre
5842 Dump trees after full redundancy elimination. The file name is made
5843 by appending @file{.fre} to the source file name.
5846 @opindex fdump-tree-copyprop
5847 Dump trees after copy propagation. The file name is made
5848 by appending @file{.copyprop} to the source file name.
5850 @item store_copyprop
5851 @opindex fdump-tree-store_copyprop
5852 Dump trees after store copy-propagation. The file name is made
5853 by appending @file{.store_copyprop} to the source file name.
5856 @opindex fdump-tree-dce
5857 Dump each function after dead code elimination. The file name is made by
5858 appending @file{.dce} to the source file name.
5861 @opindex fdump-tree-mudflap
5862 Dump each function after adding mudflap instrumentation. The file name is
5863 made by appending @file{.mudflap} to the source file name.
5866 @opindex fdump-tree-sra
5867 Dump each function after performing scalar replacement of aggregates. The
5868 file name is made by appending @file{.sra} to the source file name.
5871 @opindex fdump-tree-sink
5872 Dump each function after performing code sinking. The file name is made
5873 by appending @file{.sink} to the source file name.
5876 @opindex fdump-tree-dom
5877 Dump each function after applying dominator tree optimizations. The file
5878 name is made by appending @file{.dom} to the source file name.
5881 @opindex fdump-tree-dse
5882 Dump each function after applying dead store elimination. The file
5883 name is made by appending @file{.dse} to the source file name.
5886 @opindex fdump-tree-phiopt
5887 Dump each function after optimizing PHI nodes into straightline code. The file
5888 name is made by appending @file{.phiopt} to the source file name.
5891 @opindex fdump-tree-forwprop
5892 Dump each function after forward propagating single use variables. The file
5893 name is made by appending @file{.forwprop} to the source file name.
5896 @opindex fdump-tree-copyrename
5897 Dump each function after applying the copy rename optimization. The file
5898 name is made by appending @file{.copyrename} to the source file name.
5901 @opindex fdump-tree-nrv
5902 Dump each function after applying the named return value optimization on
5903 generic trees. The file name is made by appending @file{.nrv} to the source
5907 @opindex fdump-tree-vect
5908 Dump each function after applying vectorization of loops. The file name is
5909 made by appending @file{.vect} to the source file name.
5912 @opindex fdump-tree-slp
5913 Dump each function after applying vectorization of basic blocks. The file name
5914 is made by appending @file{.slp} to the source file name.
5917 @opindex fdump-tree-vrp
5918 Dump each function after Value Range Propagation (VRP). The file name
5919 is made by appending @file{.vrp} to the source file name.
5922 @opindex fdump-tree-all
5923 Enable all the available tree dumps with the flags provided in this option.
5926 @item -ftree-vectorizer-verbose=@var{n}
5927 @opindex ftree-vectorizer-verbose
5928 This option controls the amount of debugging output the vectorizer prints.
5929 This information is written to standard error, unless
5930 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5931 in which case it is output to the usual dump listing file, @file{.vect}.
5932 For @var{n}=0 no diagnostic information is reported.
5933 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5934 and the total number of loops that got vectorized.
5935 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5936 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5937 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5938 level that @option{-fdump-tree-vect-stats} uses.
5939 Higher verbosity levels mean either more information dumped for each
5940 reported loop, or same amount of information reported for more loops:
5941 if @var{n}=3, vectorizer cost model information is reported.
5942 If @var{n}=4, alignment related information is added to the reports.
5943 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5944 memory access-patterns) is added to the reports.
5945 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5946 that did not pass the first analysis phase (i.e., may not be countable, or
5947 may have complicated control-flow).
5948 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5949 If @var{n}=8, SLP related information is added to the reports.
5950 For @var{n}=9, all the information the vectorizer generates during its
5951 analysis and transformation is reported. This is the same verbosity level
5952 that @option{-fdump-tree-vect-details} uses.
5954 @item -frandom-seed=@var{string}
5955 @opindex frandom-seed
5956 This option provides a seed that GCC uses in place of
5957 random numbers in generating certain symbol names
5958 that have to be different in every compiled file. It is also used to
5959 place unique stamps in coverage data files and the object files that
5960 produce them. You can use the @option{-frandom-seed} option to produce
5961 reproducibly identical object files.
5963 The @var{string} should be different for every file you compile.
5965 @item -fsched-verbose=@var{n}
5966 @opindex fsched-verbose
5967 On targets that use instruction scheduling, this option controls the
5968 amount of debugging output the scheduler prints. This information is
5969 written to standard error, unless @option{-fdump-rtl-sched1} or
5970 @option{-fdump-rtl-sched2} is specified, in which case it is output
5971 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5972 respectively. However for @var{n} greater than nine, the output is
5973 always printed to standard error.
5975 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5976 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5977 For @var{n} greater than one, it also output basic block probabilities,
5978 detailed ready list information and unit/insn info. For @var{n} greater
5979 than two, it includes RTL at abort point, control-flow and regions info.
5980 And for @var{n} over four, @option{-fsched-verbose} also includes
5984 @itemx -save-temps=cwd
5986 Store the usual ``temporary'' intermediate files permanently; place them
5987 in the current directory and name them based on the source file. Thus,
5988 compiling @file{foo.c} with @option{-c -save-temps} produces files
5989 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5990 preprocessed @file{foo.i} output file even though the compiler now
5991 normally uses an integrated preprocessor.
5993 When used in combination with the @option{-x} command-line option,
5994 @option{-save-temps} is sensible enough to avoid over writing an
5995 input source file with the same extension as an intermediate file.
5996 The corresponding intermediate file may be obtained by renaming the
5997 source file before using @option{-save-temps}.
5999 If you invoke GCC in parallel, compiling several different source
6000 files that share a common base name in different subdirectories or the
6001 same source file compiled for multiple output destinations, it is
6002 likely that the different parallel compilers will interfere with each
6003 other, and overwrite the temporary files. For instance:
6006 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
6007 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
6010 may result in @file{foo.i} and @file{foo.o} being written to
6011 simultaneously by both compilers.
6013 @item -save-temps=obj
6014 @opindex save-temps=obj
6015 Store the usual ``temporary'' intermediate files permanently. If the
6016 @option{-o} option is used, the temporary files are based on the
6017 object file. If the @option{-o} option is not used, the
6018 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
6023 gcc -save-temps=obj -c foo.c
6024 gcc -save-temps=obj -c bar.c -o dir/xbar.o
6025 gcc -save-temps=obj foobar.c -o dir2/yfoobar
6029 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
6030 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
6031 @file{dir2/yfoobar.o}.
6033 @item -time@r{[}=@var{file}@r{]}
6035 Report the CPU time taken by each subprocess in the compilation
6036 sequence. For C source files, this is the compiler proper and assembler
6037 (plus the linker if linking is done).
6039 Without the specification of an output file, the output looks like this:
6046 The first number on each line is the ``user time'', that is time spent
6047 executing the program itself. The second number is ``system time'',
6048 time spent executing operating system routines on behalf of the program.
6049 Both numbers are in seconds.
6051 With the specification of an output file, the output is appended to the
6052 named file, and it looks like this:
6055 0.12 0.01 cc1 @var{options}
6056 0.00 0.01 as @var{options}
6059 The ``user time'' and the ``system time'' are moved before the program
6060 name, and the options passed to the program are displayed, so that one
6061 can later tell what file was being compiled, and with which options.
6063 @item -fvar-tracking
6064 @opindex fvar-tracking
6065 Run variable tracking pass. It computes where variables are stored at each
6066 position in code. Better debugging information is then generated
6067 (if the debugging information format supports this information).
6069 It is enabled by default when compiling with optimization (@option{-Os},
6070 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6071 the debug info format supports it.
6073 @item -fvar-tracking-assignments
6074 @opindex fvar-tracking-assignments
6075 @opindex fno-var-tracking-assignments
6076 Annotate assignments to user variables early in the compilation and
6077 attempt to carry the annotations over throughout the compilation all the
6078 way to the end, in an attempt to improve debug information while
6079 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6081 It can be enabled even if var-tracking is disabled, in which case
6082 annotations are created and maintained, but discarded at the end.
6084 @item -fvar-tracking-assignments-toggle
6085 @opindex fvar-tracking-assignments-toggle
6086 @opindex fno-var-tracking-assignments-toggle
6087 Toggle @option{-fvar-tracking-assignments}, in the same way that
6088 @option{-gtoggle} toggles @option{-g}.
6090 @item -print-file-name=@var{library}
6091 @opindex print-file-name
6092 Print the full absolute name of the library file @var{library} that
6093 would be used when linking---and don't do anything else. With this
6094 option, GCC does not compile or link anything; it just prints the
6097 @item -print-multi-directory
6098 @opindex print-multi-directory
6099 Print the directory name corresponding to the multilib selected by any
6100 other switches present in the command line. This directory is supposed
6101 to exist in @env{GCC_EXEC_PREFIX}.
6103 @item -print-multi-lib
6104 @opindex print-multi-lib
6105 Print the mapping from multilib directory names to compiler switches
6106 that enable them. The directory name is separated from the switches by
6107 @samp{;}, and each switch starts with an @samp{@@} instead of the
6108 @samp{-}, without spaces between multiple switches. This is supposed to
6109 ease shell-processing.
6111 @item -print-multi-os-directory
6112 @opindex print-multi-os-directory
6113 Print the path to OS libraries for the selected
6114 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6115 present in the @file{lib} subdirectory and no multilibs are used, this is
6116 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6117 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6118 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6119 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6121 @item -print-prog-name=@var{program}
6122 @opindex print-prog-name
6123 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6125 @item -print-libgcc-file-name
6126 @opindex print-libgcc-file-name
6127 Same as @option{-print-file-name=libgcc.a}.
6129 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6130 but you do want to link with @file{libgcc.a}. You can do
6133 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6136 @item -print-search-dirs
6137 @opindex print-search-dirs
6138 Print the name of the configured installation directory and a list of
6139 program and library directories @command{gcc} searches---and don't do anything else.
6141 This is useful when @command{gcc} prints the error message
6142 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6143 To resolve this you either need to put @file{cpp0} and the other compiler
6144 components where @command{gcc} expects to find them, or you can set the environment
6145 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6146 Don't forget the trailing @samp{/}.
6147 @xref{Environment Variables}.
6149 @item -print-sysroot
6150 @opindex print-sysroot
6151 Print the target sysroot directory that is used during
6152 compilation. This is the target sysroot specified either at configure
6153 time or using the @option{--sysroot} option, possibly with an extra
6154 suffix that depends on compilation options. If no target sysroot is
6155 specified, the option prints nothing.
6157 @item -print-sysroot-headers-suffix
6158 @opindex print-sysroot-headers-suffix
6159 Print the suffix added to the target sysroot when searching for
6160 headers, or give an error if the compiler is not configured with such
6161 a suffix---and don't do anything else.
6164 @opindex dumpmachine
6165 Print the compiler's target machine (for example,
6166 @samp{i686-pc-linux-gnu})---and don't do anything else.
6169 @opindex dumpversion
6170 Print the compiler version (for example, @samp{3.0})---and don't do
6175 Print the compiler's built-in specs---and don't do anything else. (This
6176 is used when GCC itself is being built.) @xref{Spec Files}.
6178 @item -feliminate-unused-debug-types
6179 @opindex feliminate-unused-debug-types
6180 Normally, when producing DWARF 2 output, GCC emits debugging
6181 information for all types declared in a compilation
6182 unit, regardless of whether or not they are actually used
6183 in that compilation unit. Sometimes this is useful, such as
6184 if, in the debugger, you want to cast a value to a type that is
6185 not actually used in your program (but is declared). More often,
6186 however, this results in a significant amount of wasted space.
6187 With this option, GCC avoids producing debug symbol output
6188 for types that are nowhere used in the source file being compiled.
6191 @node Optimize Options
6192 @section Options That Control Optimization
6193 @cindex optimize options
6194 @cindex options, optimization
6196 These options control various sorts of optimizations.
6198 Without any optimization option, the compiler's goal is to reduce the
6199 cost of compilation and to make debugging produce the expected
6200 results. Statements are independent: if you stop the program with a
6201 breakpoint between statements, you can then assign a new value to any
6202 variable or change the program counter to any other statement in the
6203 function and get exactly the results you expect from the source
6206 Turning on optimization flags makes the compiler attempt to improve
6207 the performance and/or code size at the expense of compilation time
6208 and possibly the ability to debug the program.
6210 The compiler performs optimization based on the knowledge it has of the
6211 program. Compiling multiple files at once to a single output file mode allows
6212 the compiler to use information gained from all of the files when compiling
6215 Not all optimizations are controlled directly by a flag. Only
6216 optimizations that have a flag are listed in this section.
6218 Most optimizations are only enabled if an @option{-O} level is set on
6219 the command line. Otherwise they are disabled, even if individual
6220 optimization flags are specified.
6222 Depending on the target and how GCC was configured, a slightly different
6223 set of optimizations may be enabled at each @option{-O} level than
6224 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6225 to find out the exact set of optimizations that are enabled at each level.
6226 @xref{Overall Options}, for examples.
6233 Optimize. Optimizing compilation takes somewhat more time, and a lot
6234 more memory for a large function.
6236 With @option{-O}, the compiler tries to reduce code size and execution
6237 time, without performing any optimizations that take a great deal of
6240 @option{-O} turns on the following optimization flags:
6244 -fcprop-registers @gol
6247 -fdelayed-branch @gol
6249 -fguess-branch-probability @gol
6250 -fif-conversion2 @gol
6251 -fif-conversion @gol
6252 -fipa-pure-const @gol
6254 -fipa-reference @gol
6256 -fsplit-wide-types @gol
6258 -ftree-builtin-call-dce @gol
6261 -ftree-copyrename @gol
6263 -ftree-dominator-opts @gol
6265 -ftree-forwprop @gol
6273 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6274 where doing so does not interfere with debugging.
6278 Optimize even more. GCC performs nearly all supported optimizations
6279 that do not involve a space-speed tradeoff.
6280 As compared to @option{-O}, this option increases both compilation time
6281 and the performance of the generated code.
6283 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6284 also turns on the following optimization flags:
6285 @gccoptlist{-fthread-jumps @gol
6286 -falign-functions -falign-jumps @gol
6287 -falign-loops -falign-labels @gol
6290 -fcse-follow-jumps -fcse-skip-blocks @gol
6291 -fdelete-null-pointer-checks @gol
6293 -fexpensive-optimizations @gol
6294 -fgcse -fgcse-lm @gol
6295 -finline-small-functions @gol
6296 -findirect-inlining @gol
6298 -foptimize-sibling-calls @gol
6299 -fpartial-inlining @gol
6302 -freorder-blocks -freorder-functions @gol
6303 -frerun-cse-after-loop @gol
6304 -fsched-interblock -fsched-spec @gol
6305 -fschedule-insns -fschedule-insns2 @gol
6306 -fstrict-aliasing -fstrict-overflow @gol
6307 -ftree-switch-conversion -ftree-tail-merge @gol
6311 Please note the warning under @option{-fgcse} about
6312 invoking @option{-O2} on programs that use computed gotos.
6316 Optimize yet more. @option{-O3} turns on all optimizations specified
6317 by @option{-O2} and also turns on the @option{-finline-functions},
6318 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6319 @option{-fgcse-after-reload}, @option{-ftree-vectorize},
6320 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
6324 Reduce compilation time and make debugging produce the expected
6325 results. This is the default.
6329 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6330 do not typically increase code size. It also performs further
6331 optimizations designed to reduce code size.
6333 @option{-Os} disables the following optimization flags:
6334 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6335 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6336 -fprefetch-loop-arrays -ftree-vect-loop-version}
6340 Disregard strict standards compliance. @option{-Ofast} enables all
6341 @option{-O3} optimizations. It also enables optimizations that are not
6342 valid for all standard compliant programs.
6343 It turns on @option{-ffast-math} and the Fortran-specific
6344 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6346 If you use multiple @option{-O} options, with or without level numbers,
6347 the last such option is the one that is effective.
6350 Options of the form @option{-f@var{flag}} specify machine-independent
6351 flags. Most flags have both positive and negative forms; the negative
6352 form of @option{-ffoo} is @option{-fno-foo}. In the table
6353 below, only one of the forms is listed---the one you typically
6354 use. You can figure out the other form by either removing @samp{no-}
6357 The following options control specific optimizations. They are either
6358 activated by @option{-O} options or are related to ones that are. You
6359 can use the following flags in the rare cases when ``fine-tuning'' of
6360 optimizations to be performed is desired.
6363 @item -fno-default-inline
6364 @opindex fno-default-inline
6365 Do not make member functions inline by default merely because they are
6366 defined inside the class scope (C++ only). Otherwise, when you specify
6367 @w{@option{-O}}, member functions defined inside class scope are compiled
6368 inline by default; i.e., you don't need to add @samp{inline} in front of
6369 the member function name.
6371 @item -fno-defer-pop
6372 @opindex fno-defer-pop
6373 Always pop the arguments to each function call as soon as that function
6374 returns. For machines that must pop arguments after a function call,
6375 the compiler normally lets arguments accumulate on the stack for several
6376 function calls and pops them all at once.
6378 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6380 @item -fforward-propagate
6381 @opindex fforward-propagate
6382 Perform a forward propagation pass on RTL@. The pass tries to combine two
6383 instructions and checks if the result can be simplified. If loop unrolling
6384 is active, two passes are performed and the second is scheduled after
6387 This option is enabled by default at optimization levels @option{-O},
6388 @option{-O2}, @option{-O3}, @option{-Os}.
6390 @item -ffp-contract=@var{style}
6391 @opindex ffp-contract
6392 @option{-ffp-contract=off} disables floating-point expression contraction.
6393 @option{-ffp-contract=fast} enables floating-point expression contraction
6394 such as forming of fused multiply-add operations if the target has
6395 native support for them.
6396 @option{-ffp-contract=on} enables floating-point expression contraction
6397 if allowed by the language standard. This is currently not implemented
6398 and treated equal to @option{-ffp-contract=off}.
6400 The default is @option{-ffp-contract=fast}.
6402 @item -fomit-frame-pointer
6403 @opindex fomit-frame-pointer
6404 Don't keep the frame pointer in a register for functions that
6405 don't need one. This avoids the instructions to save, set up and
6406 restore frame pointers; it also makes an extra register available
6407 in many functions. @strong{It also makes debugging impossible on
6410 On some machines, such as the VAX, this flag has no effect, because
6411 the standard calling sequence automatically handles the frame pointer
6412 and nothing is saved by pretending it doesn't exist. The
6413 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6414 whether a target machine supports this flag. @xref{Registers,,Register
6415 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6417 Starting with GCC version 4.6, the default setting (when not optimizing for
6418 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
6419 @option{-fomit-frame-pointer}. The default can be reverted to
6420 @option{-fno-omit-frame-pointer} by configuring GCC with the
6421 @option{--enable-frame-pointer} configure option.
6423 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6425 @item -foptimize-sibling-calls
6426 @opindex foptimize-sibling-calls
6427 Optimize sibling and tail recursive calls.
6429 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6433 Do not expand any functions inline apart from those marked with
6434 the @code{always_inline} attribute. This is the default when not
6437 Single functions can be exempted from inlining by marking them
6438 with the @code{noinline} attribute.
6440 @item -finline-small-functions
6441 @opindex finline-small-functions
6442 Integrate functions into their callers when their body is smaller than expected
6443 function call code (so overall size of program gets smaller). The compiler
6444 heuristically decides which functions are simple enough to be worth integrating
6445 in this way. This inlining applies to all functions, even those not declared
6448 Enabled at level @option{-O2}.
6450 @item -findirect-inlining
6451 @opindex findirect-inlining
6452 Inline also indirect calls that are discovered to be known at compile
6453 time thanks to previous inlining. This option has any effect only
6454 when inlining itself is turned on by the @option{-finline-functions}
6455 or @option{-finline-small-functions} options.
6457 Enabled at level @option{-O2}.
6459 @item -finline-functions
6460 @opindex finline-functions
6461 Consider all functions for inlining, even if they are not declared inline.
6462 The compiler heuristically decides which functions are worth integrating
6465 If all calls to a given function are integrated, and the function is
6466 declared @code{static}, then the function is normally not output as
6467 assembler code in its own right.
6469 Enabled at level @option{-O3}.
6471 @item -finline-functions-called-once
6472 @opindex finline-functions-called-once
6473 Consider all @code{static} functions called once for inlining into their
6474 caller even if they are not marked @code{inline}. If a call to a given
6475 function is integrated, then the function is not output as assembler code
6478 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6480 @item -fearly-inlining
6481 @opindex fearly-inlining
6482 Inline functions marked by @code{always_inline} and functions whose body seems
6483 smaller than the function call overhead early before doing
6484 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6485 makes profiling significantly cheaper and usually inlining faster on programs
6486 having large chains of nested wrapper functions.
6492 Perform interprocedural scalar replacement of aggregates, removal of
6493 unused parameters and replacement of parameters passed by reference
6494 by parameters passed by value.
6496 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6498 @item -finline-limit=@var{n}
6499 @opindex finline-limit
6500 By default, GCC limits the size of functions that can be inlined. This flag
6501 allows coarse control of this limit. @var{n} is the size of functions that
6502 can be inlined in number of pseudo instructions.
6504 Inlining is actually controlled by a number of parameters, which may be
6505 specified individually by using @option{--param @var{name}=@var{value}}.
6506 The @option{-finline-limit=@var{n}} option sets some of these parameters
6510 @item max-inline-insns-single
6511 is set to @var{n}/2.
6512 @item max-inline-insns-auto
6513 is set to @var{n}/2.
6516 See below for a documentation of the individual
6517 parameters controlling inlining and for the defaults of these parameters.
6519 @emph{Note:} there may be no value to @option{-finline-limit} that results
6520 in default behavior.
6522 @emph{Note:} pseudo instruction represents, in this particular context, an
6523 abstract measurement of function's size. In no way does it represent a count
6524 of assembly instructions and as such its exact meaning might change from one
6525 release to an another.
6527 @item -fno-keep-inline-dllexport
6528 @opindex -fno-keep-inline-dllexport
6529 This is a more fine-grained version of @option{-fkeep-inline-functions},
6530 which applies only to functions that are declared using the @code{dllexport}
6531 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6534 @item -fkeep-inline-functions
6535 @opindex fkeep-inline-functions
6536 In C, emit @code{static} functions that are declared @code{inline}
6537 into the object file, even if the function has been inlined into all
6538 of its callers. This switch does not affect functions using the
6539 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6540 inline functions into the object file.
6542 @item -fkeep-static-consts
6543 @opindex fkeep-static-consts
6544 Emit variables declared @code{static const} when optimization isn't turned
6545 on, even if the variables aren't referenced.
6547 GCC enables this option by default. If you want to force the compiler to
6548 check if a variable is referenced, regardless of whether or not
6549 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6551 @item -fmerge-constants
6552 @opindex fmerge-constants
6553 Attempt to merge identical constants (string constants and floating-point
6554 constants) across compilation units.
6556 This option is the default for optimized compilation if the assembler and
6557 linker support it. Use @option{-fno-merge-constants} to inhibit this
6560 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6562 @item -fmerge-all-constants
6563 @opindex fmerge-all-constants
6564 Attempt to merge identical constants and identical variables.
6566 This option implies @option{-fmerge-constants}. In addition to
6567 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6568 arrays or initialized constant variables with integral or floating-point
6569 types. Languages like C or C++ require each variable, including multiple
6570 instances of the same variable in recursive calls, to have distinct locations,
6571 so using this option results in non-conforming
6574 @item -fmodulo-sched
6575 @opindex fmodulo-sched
6576 Perform swing modulo scheduling immediately before the first scheduling
6577 pass. This pass looks at innermost loops and reorders their
6578 instructions by overlapping different iterations.
6580 @item -fmodulo-sched-allow-regmoves
6581 @opindex fmodulo-sched-allow-regmoves
6582 Perform more aggressive SMS-based modulo scheduling with register moves
6583 allowed. By setting this flag certain anti-dependences edges are
6584 deleted, which triggers the generation of reg-moves based on the
6585 life-range analysis. This option is effective only with
6586 @option{-fmodulo-sched} enabled.
6588 @item -fno-branch-count-reg
6589 @opindex fno-branch-count-reg
6590 Do not use ``decrement and branch'' instructions on a count register,
6591 but instead generate a sequence of instructions that decrement a
6592 register, compare it against zero, then branch based upon the result.
6593 This option is only meaningful on architectures that support such
6594 instructions, which include x86, PowerPC, IA-64 and S/390.
6596 The default is @option{-fbranch-count-reg}.
6598 @item -fno-function-cse
6599 @opindex fno-function-cse
6600 Do not put function addresses in registers; make each instruction that
6601 calls a constant function contain the function's address explicitly.
6603 This option results in less efficient code, but some strange hacks
6604 that alter the assembler output may be confused by the optimizations
6605 performed when this option is not used.
6607 The default is @option{-ffunction-cse}
6609 @item -fno-zero-initialized-in-bss
6610 @opindex fno-zero-initialized-in-bss
6611 If the target supports a BSS section, GCC by default puts variables that
6612 are initialized to zero into BSS@. This can save space in the resulting
6615 This option turns off this behavior because some programs explicitly
6616 rely on variables going to the data section. E.g., so that the
6617 resulting executable can find the beginning of that section and/or make
6618 assumptions based on that.
6620 The default is @option{-fzero-initialized-in-bss}.
6622 @item -fmudflap -fmudflapth -fmudflapir
6626 @cindex bounds checking
6628 For front-ends that support it (C and C++), instrument all risky
6629 pointer/array dereferencing operations, some standard library
6630 string/heap functions, and some other associated constructs with
6631 range/validity tests. Modules so instrumented should be immune to
6632 buffer overflows, invalid heap use, and some other classes of C/C++
6633 programming errors. The instrumentation relies on a separate runtime
6634 library (@file{libmudflap}), which is linked into a program if
6635 @option{-fmudflap} is given at link time. Run-time behavior of the
6636 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6637 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6640 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6641 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6642 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6643 instrumentation should ignore pointer reads. This produces less
6644 instrumentation (and therefore faster execution) and still provides
6645 some protection against outright memory corrupting writes, but allows
6646 erroneously read data to propagate within a program.
6648 @item -fthread-jumps
6649 @opindex fthread-jumps
6650 Perform optimizations where we check to see if a jump branches to a
6651 location where another comparison subsumed by the first is found. If
6652 so, the first branch is redirected to either the destination of the
6653 second branch or a point immediately following it, depending on whether
6654 the condition is known to be true or false.
6656 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6658 @item -fsplit-wide-types
6659 @opindex fsplit-wide-types
6660 When using a type that occupies multiple registers, such as @code{long
6661 long} on a 32-bit system, split the registers apart and allocate them
6662 independently. This normally generates better code for those types,
6663 but may make debugging more difficult.
6665 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6668 @item -fcse-follow-jumps
6669 @opindex fcse-follow-jumps
6670 In common subexpression elimination (CSE), scan through jump instructions
6671 when the target of the jump is not reached by any other path. For
6672 example, when CSE encounters an @code{if} statement with an
6673 @code{else} clause, CSE follows the jump when the condition
6676 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6678 @item -fcse-skip-blocks
6679 @opindex fcse-skip-blocks
6680 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6681 follow jumps that conditionally skip over blocks. When CSE
6682 encounters a simple @code{if} statement with no else clause,
6683 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6684 body of the @code{if}.
6686 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6688 @item -frerun-cse-after-loop
6689 @opindex frerun-cse-after-loop
6690 Re-run common subexpression elimination after loop optimizations are
6693 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6697 Perform a global common subexpression elimination pass.
6698 This pass also performs global constant and copy propagation.
6700 @emph{Note:} When compiling a program using computed gotos, a GCC
6701 extension, you may get better run-time performance if you disable
6702 the global common subexpression elimination pass by adding
6703 @option{-fno-gcse} to the command line.
6705 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6709 When @option{-fgcse-lm} is enabled, global common subexpression elimination
6710 attempts to move loads that are only killed by stores into themselves. This
6711 allows a loop containing a load/store sequence to be changed to a load outside
6712 the loop, and a copy/store within the loop.
6714 Enabled by default when @option{-fgcse} is enabled.
6718 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6719 global common subexpression elimination. This pass attempts to move
6720 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6721 loops containing a load/store sequence can be changed to a load before
6722 the loop and a store after the loop.
6724 Not enabled at any optimization level.
6728 When @option{-fgcse-las} is enabled, the global common subexpression
6729 elimination pass eliminates redundant loads that come after stores to the
6730 same memory location (both partial and full redundancies).
6732 Not enabled at any optimization level.
6734 @item -fgcse-after-reload
6735 @opindex fgcse-after-reload
6736 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6737 pass is performed after reload. The purpose of this pass is to cleanup
6740 @item -funsafe-loop-optimizations
6741 @opindex funsafe-loop-optimizations
6742 If given, the loop optimizer assumes that loop indices do not
6743 overflow, and that the loops with nontrivial exit condition are not
6744 infinite. This enables a wider range of loop optimizations even if
6745 the loop optimizer itself cannot prove that these assumptions are valid.
6746 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
6747 if it finds this kind of loop.
6749 @item -fcrossjumping
6750 @opindex fcrossjumping
6751 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6752 resulting code may or may not perform better than without cross-jumping.
6754 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6756 @item -fauto-inc-dec
6757 @opindex fauto-inc-dec
6758 Combine increments or decrements of addresses with memory accesses.
6759 This pass is always skipped on architectures that do not have
6760 instructions to support this. Enabled by default at @option{-O} and
6761 higher on architectures that support this.
6765 Perform dead code elimination (DCE) on RTL@.
6766 Enabled by default at @option{-O} and higher.
6770 Perform dead store elimination (DSE) on RTL@.
6771 Enabled by default at @option{-O} and higher.
6773 @item -fif-conversion
6774 @opindex fif-conversion
6775 Attempt to transform conditional jumps into branch-less equivalents. This
6776 include use of conditional moves, min, max, set flags and abs instructions, and
6777 some tricks doable by standard arithmetics. The use of conditional execution
6778 on chips where it is available is controlled by @code{if-conversion2}.
6780 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6782 @item -fif-conversion2
6783 @opindex fif-conversion2
6784 Use conditional execution (where available) to transform conditional jumps into
6785 branch-less equivalents.
6787 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6789 @item -fdelete-null-pointer-checks
6790 @opindex fdelete-null-pointer-checks
6791 Assume that programs cannot safely dereference null pointers, and that
6792 no code or data element resides there. This enables simple constant
6793 folding optimizations at all optimization levels. In addition, other
6794 optimization passes in GCC use this flag to control global dataflow
6795 analyses that eliminate useless checks for null pointers; these assume
6796 that if a pointer is checked after it has already been dereferenced,
6799 Note however that in some environments this assumption is not true.
6800 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6801 for programs that depend on that behavior.
6803 Some targets, especially embedded ones, disable this option at all levels.
6804 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6805 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6806 are enabled independently at different optimization levels.
6808 @item -fdevirtualize
6809 @opindex fdevirtualize
6810 Attempt to convert calls to virtual functions to direct calls. This
6811 is done both within a procedure and interprocedurally as part of
6812 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6813 propagation (@option{-fipa-cp}).
6814 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6816 @item -fexpensive-optimizations
6817 @opindex fexpensive-optimizations
6818 Perform a number of minor optimizations that are relatively expensive.
6820 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6824 Attempt to remove redundant extension instructions. This is especially
6825 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
6826 registers after writing to their lower 32-bit half.
6828 Enabled for x86 at levels @option{-O2}, @option{-O3}.
6830 @item -foptimize-register-move
6832 @opindex foptimize-register-move
6834 Attempt to reassign register numbers in move instructions and as
6835 operands of other simple instructions in order to maximize the amount of
6836 register tying. This is especially helpful on machines with two-operand
6839 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6842 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6844 @item -fira-algorithm=@var{algorithm}
6845 Use the specified coloring algorithm for the integrated register
6846 allocator. The @var{algorithm} argument can be @samp{priority}, which
6847 specifies Chow's priority coloring, or @samp{CB}, which specifies
6848 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
6849 for all architectures, but for those targets that do support it, it is
6850 the default because it generates better code.
6852 @item -fira-region=@var{region}
6853 Use specified regions for the integrated register allocator. The
6854 @var{region} argument should be one of the following:
6859 Use all loops as register allocation regions.
6860 This can give the best results for machines with a small and/or
6861 irregular register set.
6864 Use all loops except for loops with small register pressure
6865 as the regions. This value usually gives
6866 the best results in most cases and for most architectures,
6867 and is enabled by default when compiling with optimization for speed
6868 (@option{-O}, @option{-O2}, @dots{}).
6871 Use all functions as a single region.
6872 This typically results in the smallest code size, and is enabled by default for
6873 @option{-Os} or @option{-O0}.
6877 @item -fira-loop-pressure
6878 @opindex fira-loop-pressure
6879 Use IRA to evaluate register pressure in loops for decisions to move
6880 loop invariants. This option usually results in generation
6881 of faster and smaller code on machines with large register files (>= 32
6882 registers), but it can slow the compiler down.
6884 This option is enabled at level @option{-O3} for some targets.
6886 @item -fno-ira-share-save-slots
6887 @opindex fno-ira-share-save-slots
6888 Disable sharing of stack slots used for saving call-used hard
6889 registers living through a call. Each hard register gets a
6890 separate stack slot, and as a result function stack frames are
6893 @item -fno-ira-share-spill-slots
6894 @opindex fno-ira-share-spill-slots
6895 Disable sharing of stack slots allocated for pseudo-registers. Each
6896 pseudo-register that does not get a hard register gets a separate
6897 stack slot, and as a result function stack frames are larger.
6899 @item -fira-verbose=@var{n}
6900 @opindex fira-verbose
6901 Control the verbosity of the dump file for the integrated register allocator.
6902 The default value is 5. If the value @var{n} is greater or equal to 10,
6903 the dump output is sent to stderr using the same format as @var{n} minus 10.
6905 @item -fdelayed-branch
6906 @opindex fdelayed-branch
6907 If supported for the target machine, attempt to reorder instructions
6908 to exploit instruction slots available after delayed branch
6911 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6913 @item -fschedule-insns
6914 @opindex fschedule-insns
6915 If supported for the target machine, attempt to reorder instructions to
6916 eliminate execution stalls due to required data being unavailable. This
6917 helps machines that have slow floating point or memory load instructions
6918 by allowing other instructions to be issued until the result of the load
6919 or floating-point instruction is required.
6921 Enabled at levels @option{-O2}, @option{-O3}.
6923 @item -fschedule-insns2
6924 @opindex fschedule-insns2
6925 Similar to @option{-fschedule-insns}, but requests an additional pass of
6926 instruction scheduling after register allocation has been done. This is
6927 especially useful on machines with a relatively small number of
6928 registers and where memory load instructions take more than one cycle.
6930 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6932 @item -fno-sched-interblock
6933 @opindex fno-sched-interblock
6934 Don't schedule instructions across basic blocks. This is normally
6935 enabled by default when scheduling before register allocation, i.e.@:
6936 with @option{-fschedule-insns} or at @option{-O2} or higher.
6938 @item -fno-sched-spec
6939 @opindex fno-sched-spec
6940 Don't allow speculative motion of non-load instructions. This is normally
6941 enabled by default when scheduling before register allocation, i.e.@:
6942 with @option{-fschedule-insns} or at @option{-O2} or higher.
6944 @item -fsched-pressure
6945 @opindex fsched-pressure
6946 Enable register pressure sensitive insn scheduling before the register
6947 allocation. This only makes sense when scheduling before register
6948 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6949 @option{-O2} or higher. Usage of this option can improve the
6950 generated code and decrease its size by preventing register pressure
6951 increase above the number of available hard registers and as a
6952 consequence register spills in the register allocation.
6954 @item -fsched-spec-load
6955 @opindex fsched-spec-load
6956 Allow speculative motion of some load instructions. This only makes
6957 sense when scheduling before register allocation, i.e.@: with
6958 @option{-fschedule-insns} or at @option{-O2} or higher.
6960 @item -fsched-spec-load-dangerous
6961 @opindex fsched-spec-load-dangerous
6962 Allow speculative motion of more load instructions. This only makes
6963 sense when scheduling before register allocation, i.e.@: with
6964 @option{-fschedule-insns} or at @option{-O2} or higher.
6966 @item -fsched-stalled-insns
6967 @itemx -fsched-stalled-insns=@var{n}
6968 @opindex fsched-stalled-insns
6969 Define how many insns (if any) can be moved prematurely from the queue
6970 of stalled insns into the ready list during the second scheduling pass.
6971 @option{-fno-sched-stalled-insns} means that no insns are moved
6972 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6973 on how many queued insns can be moved prematurely.
6974 @option{-fsched-stalled-insns} without a value is equivalent to
6975 @option{-fsched-stalled-insns=1}.
6977 @item -fsched-stalled-insns-dep
6978 @itemx -fsched-stalled-insns-dep=@var{n}
6979 @opindex fsched-stalled-insns-dep
6980 Define how many insn groups (cycles) are examined for a dependency
6981 on a stalled insn that is a candidate for premature removal from the queue
6982 of stalled insns. This has an effect only during the second scheduling pass,
6983 and only if @option{-fsched-stalled-insns} is used.
6984 @option{-fno-sched-stalled-insns-dep} is equivalent to
6985 @option{-fsched-stalled-insns-dep=0}.
6986 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6987 @option{-fsched-stalled-insns-dep=1}.
6989 @item -fsched2-use-superblocks
6990 @opindex fsched2-use-superblocks
6991 When scheduling after register allocation, do use superblock scheduling
6992 algorithm. Superblock scheduling allows motion across basic block boundaries
6993 resulting on faster schedules. This option is experimental, as not all machine
6994 descriptions used by GCC model the CPU closely enough to avoid unreliable
6995 results from the algorithm.
6997 This only makes sense when scheduling after register allocation, i.e.@: with
6998 @option{-fschedule-insns2} or at @option{-O2} or higher.
7000 @item -fsched-group-heuristic
7001 @opindex fsched-group-heuristic
7002 Enable the group heuristic in the scheduler. This heuristic favors
7003 the instruction that belongs to a schedule group. This is enabled
7004 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7005 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7007 @item -fsched-critical-path-heuristic
7008 @opindex fsched-critical-path-heuristic
7009 Enable the critical-path heuristic in the scheduler. This heuristic favors
7010 instructions on the critical path. This is enabled by default when
7011 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7012 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7014 @item -fsched-spec-insn-heuristic
7015 @opindex fsched-spec-insn-heuristic
7016 Enable the speculative instruction heuristic in the scheduler. This
7017 heuristic favors speculative instructions with greater dependency weakness.
7018 This is enabled by default when scheduling is enabled, i.e.@:
7019 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7020 or at @option{-O2} or higher.
7022 @item -fsched-rank-heuristic
7023 @opindex fsched-rank-heuristic
7024 Enable the rank heuristic in the scheduler. This heuristic favors
7025 the instruction belonging to a basic block with greater size or frequency.
7026 This is enabled by default when scheduling is enabled, i.e.@:
7027 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7028 at @option{-O2} or higher.
7030 @item -fsched-last-insn-heuristic
7031 @opindex fsched-last-insn-heuristic
7032 Enable the last-instruction heuristic in the scheduler. This heuristic
7033 favors the instruction that is less dependent on the last instruction
7034 scheduled. This is enabled by default when scheduling is enabled,
7035 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7036 at @option{-O2} or higher.
7038 @item -fsched-dep-count-heuristic
7039 @opindex fsched-dep-count-heuristic
7040 Enable the dependent-count heuristic in the scheduler. This heuristic
7041 favors the instruction that has more instructions depending on it.
7042 This is enabled by default when scheduling is enabled, i.e.@:
7043 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7044 at @option{-O2} or higher.
7046 @item -freschedule-modulo-scheduled-loops
7047 @opindex freschedule-modulo-scheduled-loops
7048 The modulo scheduling comes before the traditional scheduling. If a loop
7049 is modulo scheduled you may want to prevent the later scheduling passes
7050 from changing its schedule; use this option to control that.
7052 @item -fselective-scheduling
7053 @opindex fselective-scheduling
7054 Schedule instructions using selective scheduling algorithm. Selective
7055 scheduling runs instead of the first scheduler pass.
7057 @item -fselective-scheduling2
7058 @opindex fselective-scheduling2
7059 Schedule instructions using selective scheduling algorithm. Selective
7060 scheduling runs instead of the second scheduler pass.
7062 @item -fsel-sched-pipelining
7063 @opindex fsel-sched-pipelining
7064 Enable software pipelining of innermost loops during selective scheduling.
7065 This option has no effect until one of @option{-fselective-scheduling} or
7066 @option{-fselective-scheduling2} is turned on.
7068 @item -fsel-sched-pipelining-outer-loops
7069 @opindex fsel-sched-pipelining-outer-loops
7070 When pipelining loops during selective scheduling, also pipeline outer loops.
7071 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
7074 @opindex fshrink-wrap
7075 Emit function prologues only before parts of the function that need it,
7076 rather than at the top of the function. This flag is enabled by default at
7077 @option{-O} and higher.
7079 @item -fcaller-saves
7080 @opindex fcaller-saves
7081 Enable allocation of values to registers that are clobbered by
7082 function calls, by emitting extra instructions to save and restore the
7083 registers around such calls. Such allocation is done only when it
7084 seems to result in better code.
7086 This option is always enabled by default on certain machines, usually
7087 those which have no call-preserved registers to use instead.
7089 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7091 @item -fcombine-stack-adjustments
7092 @opindex fcombine-stack-adjustments
7093 Tracks stack adjustments (pushes and pops) and stack memory references
7094 and then tries to find ways to combine them.
7096 Enabled by default at @option{-O1} and higher.
7098 @item -fconserve-stack
7099 @opindex fconserve-stack
7100 Attempt to minimize stack usage. The compiler attempts to use less
7101 stack space, even if that makes the program slower. This option
7102 implies setting the @option{large-stack-frame} parameter to 100
7103 and the @option{large-stack-frame-growth} parameter to 400.
7105 @item -ftree-reassoc
7106 @opindex ftree-reassoc
7107 Perform reassociation on trees. This flag is enabled by default
7108 at @option{-O} and higher.
7112 Perform partial redundancy elimination (PRE) on trees. This flag is
7113 enabled by default at @option{-O2} and @option{-O3}.
7115 @item -ftree-partial-pre
7116 @opindex ftree-partial-pre
7117 Make partial redundancy elimination (PRE) more aggressive. This flag is
7118 enabled by default at @option{-O3}.
7120 @item -ftree-forwprop
7121 @opindex ftree-forwprop
7122 Perform forward propagation on trees. This flag is enabled by default
7123 at @option{-O} and higher.
7127 Perform full redundancy elimination (FRE) on trees. The difference
7128 between FRE and PRE is that FRE only considers expressions
7129 that are computed on all paths leading to the redundant computation.
7130 This analysis is faster than PRE, though it exposes fewer redundancies.
7131 This flag is enabled by default at @option{-O} and higher.
7133 @item -ftree-phiprop
7134 @opindex ftree-phiprop
7135 Perform hoisting of loads from conditional pointers on trees. This
7136 pass is enabled by default at @option{-O} and higher.
7138 @item -ftree-copy-prop
7139 @opindex ftree-copy-prop
7140 Perform copy propagation on trees. This pass eliminates unnecessary
7141 copy operations. This flag is enabled by default at @option{-O} and
7144 @item -fipa-pure-const
7145 @opindex fipa-pure-const
7146 Discover which functions are pure or constant.
7147 Enabled by default at @option{-O} and higher.
7149 @item -fipa-reference
7150 @opindex fipa-reference
7151 Discover which static variables do not escape cannot escape the
7153 Enabled by default at @option{-O} and higher.
7157 Perform interprocedural pointer analysis and interprocedural modification
7158 and reference analysis. This option can cause excessive memory and
7159 compile-time usage on large compilation units. It is not enabled by
7160 default at any optimization level.
7163 @opindex fipa-profile
7164 Perform interprocedural profile propagation. The functions called only from
7165 cold functions are marked as cold. Also functions executed once (such as
7166 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7167 functions and loop less parts of functions executed once are then optimized for
7169 Enabled by default at @option{-O} and higher.
7173 Perform interprocedural constant propagation.
7174 This optimization analyzes the program to determine when values passed
7175 to functions are constants and then optimizes accordingly.
7176 This optimization can substantially increase performance
7177 if the application has constants passed to functions.
7178 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7180 @item -fipa-cp-clone
7181 @opindex fipa-cp-clone
7182 Perform function cloning to make interprocedural constant propagation stronger.
7183 When enabled, interprocedural constant propagation performs function cloning
7184 when externally visible function can be called with constant arguments.
7185 Because this optimization can create multiple copies of functions,
7186 it may significantly increase code size
7187 (see @option{--param ipcp-unit-growth=@var{value}}).
7188 This flag is enabled by default at @option{-O3}.
7190 @item -fipa-matrix-reorg
7191 @opindex fipa-matrix-reorg
7192 Perform matrix flattening and transposing.
7193 Matrix flattening tries to replace an @math{m}-dimensional matrix
7194 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7195 This reduces the level of indirection needed for accessing the elements
7196 of the matrix. The second optimization is matrix transposing, which
7197 attempts to change the order of the matrix's dimensions in order to
7198 improve cache locality.
7199 Both optimizations need the @option{-fwhole-program} flag.
7200 Transposing is enabled only if profiling information is available.
7204 Perform forward store motion on trees. This flag is
7205 enabled by default at @option{-O} and higher.
7207 @item -ftree-bit-ccp
7208 @opindex ftree-bit-ccp
7209 Perform sparse conditional bit constant propagation on trees and propagate
7210 pointer alignment information.
7211 This pass only operates on local scalar variables and is enabled by default
7212 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7216 Perform sparse conditional constant propagation (CCP) on trees. This
7217 pass only operates on local scalar variables and is enabled by default
7218 at @option{-O} and higher.
7220 @item -ftree-switch-conversion
7221 Perform conversion of simple initializations in a switch to
7222 initializations from a scalar array. This flag is enabled by default
7223 at @option{-O2} and higher.
7225 @item -ftree-tail-merge
7226 Look for identical code sequences. When found, replace one with a jump to the
7227 other. This optimization is known as tail merging or cross jumping. This flag
7228 is enabled by default at @option{-O2} and higher. The compilation time
7230 be limited using @option{max-tail-merge-comparisons} parameter and
7231 @option{max-tail-merge-iterations} parameter.
7235 Perform dead code elimination (DCE) on trees. This flag is enabled by
7236 default at @option{-O} and higher.
7238 @item -ftree-builtin-call-dce
7239 @opindex ftree-builtin-call-dce
7240 Perform conditional dead code elimination (DCE) for calls to builtin functions
7241 that may set @code{errno} but are otherwise side-effect free. This flag is
7242 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7245 @item -ftree-dominator-opts
7246 @opindex ftree-dominator-opts
7247 Perform a variety of simple scalar cleanups (constant/copy
7248 propagation, redundancy elimination, range propagation and expression
7249 simplification) based on a dominator tree traversal. This also
7250 performs jump threading (to reduce jumps to jumps). This flag is
7251 enabled by default at @option{-O} and higher.
7255 Perform dead store elimination (DSE) on trees. A dead store is a store into
7256 a memory location that is later overwritten by another store without
7257 any intervening loads. In this case the earlier store can be deleted. This
7258 flag is enabled by default at @option{-O} and higher.
7262 Perform loop header copying on trees. This is beneficial since it increases
7263 effectiveness of code motion optimizations. It also saves one jump. This flag
7264 is enabled by default at @option{-O} and higher. It is not enabled
7265 for @option{-Os}, since it usually increases code size.
7267 @item -ftree-loop-optimize
7268 @opindex ftree-loop-optimize
7269 Perform loop optimizations on trees. This flag is enabled by default
7270 at @option{-O} and higher.
7272 @item -ftree-loop-linear
7273 @opindex ftree-loop-linear
7274 Perform loop interchange transformations on tree. Same as
7275 @option{-floop-interchange}. To use this code transformation, GCC has
7276 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7277 enable the Graphite loop transformation infrastructure.
7279 @item -floop-interchange
7280 @opindex floop-interchange
7281 Perform loop interchange transformations on loops. Interchanging two
7282 nested loops switches the inner and outer loops. For example, given a
7287 A(J, I) = A(J, I) * C
7291 loop interchange transforms the loop as if it were written:
7295 A(J, I) = A(J, I) * C
7299 which can be beneficial when @code{N} is larger than the caches,
7300 because in Fortran, the elements of an array are stored in memory
7301 contiguously by column, and the original loop iterates over rows,
7302 potentially creating at each access a cache miss. This optimization
7303 applies to all the languages supported by GCC and is not limited to
7304 Fortran. To use this code transformation, GCC has to be configured
7305 with @option{--with-ppl} and @option{--with-cloog} to enable the
7306 Graphite loop transformation infrastructure.
7308 @item -floop-strip-mine
7309 @opindex floop-strip-mine
7310 Perform loop strip mining transformations on loops. Strip mining
7311 splits a loop into two nested loops. The outer loop has strides
7312 equal to the strip size and the inner loop has strides of the
7313 original loop within a strip. The strip length can be changed
7314 using the @option{loop-block-tile-size} parameter. For example,
7321 loop strip mining transforms the loop as if it were written:
7324 DO I = II, min (II + 50, N)
7329 This optimization applies to all the languages supported by GCC and is
7330 not limited to Fortran. To use this code transformation, GCC has to
7331 be configured with @option{--with-ppl} and @option{--with-cloog} to
7332 enable the Graphite loop transformation infrastructure.
7335 @opindex floop-block
7336 Perform loop blocking transformations on loops. Blocking strip mines
7337 each loop in the loop nest such that the memory accesses of the
7338 element loops fit inside caches. The strip length can be changed
7339 using the @option{loop-block-tile-size} parameter. For example, given
7344 A(J, I) = B(I) + C(J)
7348 loop blocking transforms the loop as if it were written:
7352 DO I = II, min (II + 50, N)
7353 DO J = JJ, min (JJ + 50, M)
7354 A(J, I) = B(I) + C(J)
7360 which can be beneficial when @code{M} is larger than the caches,
7361 because the innermost loop iterates over a smaller amount of data
7362 which can be kept in the caches. This optimization applies to all the
7363 languages supported by GCC and is not limited to Fortran. To use this
7364 code transformation, GCC has to be configured with @option{--with-ppl}
7365 and @option{--with-cloog} to enable the Graphite loop transformation
7368 @item -fgraphite-identity
7369 @opindex fgraphite-identity
7370 Enable the identity transformation for graphite. For every SCoP we generate
7371 the polyhedral representation and transform it back to gimple. Using
7372 @option{-fgraphite-identity} we can check the costs or benefits of the
7373 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7374 are also performed by the code generator CLooG, like index splitting and
7375 dead code elimination in loops.
7377 @item -floop-parallelize-all
7378 @opindex floop-parallelize-all
7379 Use the Graphite data dependence analysis to identify loops that can
7380 be parallelized. Parallelize all the loops that can be analyzed to
7381 not contain loop carried dependences without checking that it is
7382 profitable to parallelize the loops.
7384 @item -fcheck-data-deps
7385 @opindex fcheck-data-deps
7386 Compare the results of several data dependence analyzers. This option
7387 is used for debugging the data dependence analyzers.
7389 @item -ftree-loop-if-convert
7390 Attempt to transform conditional jumps in the innermost loops to
7391 branch-less equivalents. The intent is to remove control-flow from
7392 the innermost loops in order to improve the ability of the
7393 vectorization pass to handle these loops. This is enabled by default
7394 if vectorization is enabled.
7396 @item -ftree-loop-if-convert-stores
7397 Attempt to also if-convert conditional jumps containing memory writes.
7398 This transformation can be unsafe for multi-threaded programs as it
7399 transforms conditional memory writes into unconditional memory writes.
7402 for (i = 0; i < N; i++)
7408 for (i = 0; i < N; i++)
7409 A[i] = cond ? expr : A[i];
7411 potentially producing data races.
7413 @item -ftree-loop-distribution
7414 Perform loop distribution. This flag can improve cache performance on
7415 big loop bodies and allow further loop optimizations, like
7416 parallelization or vectorization, to take place. For example, the loop
7433 @item -ftree-loop-distribute-patterns
7434 Perform loop distribution of patterns that can be code generated with
7435 calls to a library. This flag is enabled by default at @option{-O3}.
7437 This pass distributes the initialization loops and generates a call to
7438 memset zero. For example, the loop
7454 and the initialization loop is transformed into a call to memset zero.
7456 @item -ftree-loop-im
7457 @opindex ftree-loop-im
7458 Perform loop invariant motion on trees. This pass moves only invariants that
7459 are hard to handle at RTL level (function calls, operations that expand to
7460 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7461 operands of conditions that are invariant out of the loop, so that we can use
7462 just trivial invariantness analysis in loop unswitching. The pass also includes
7465 @item -ftree-loop-ivcanon
7466 @opindex ftree-loop-ivcanon
7467 Create a canonical counter for number of iterations in loops for which
7468 determining number of iterations requires complicated analysis. Later
7469 optimizations then may determine the number easily. Useful especially
7470 in connection with unrolling.
7474 Perform induction variable optimizations (strength reduction, induction
7475 variable merging and induction variable elimination) on trees.
7477 @item -ftree-parallelize-loops=n
7478 @opindex ftree-parallelize-loops
7479 Parallelize loops, i.e., split their iteration space to run in n threads.
7480 This is only possible for loops whose iterations are independent
7481 and can be arbitrarily reordered. The optimization is only
7482 profitable on multiprocessor machines, for loops that are CPU-intensive,
7483 rather than constrained e.g.@: by memory bandwidth. This option
7484 implies @option{-pthread}, and thus is only supported on targets
7485 that have support for @option{-pthread}.
7489 Perform function-local points-to analysis on trees. This flag is
7490 enabled by default at @option{-O} and higher.
7494 Perform scalar replacement of aggregates. This pass replaces structure
7495 references with scalars to prevent committing structures to memory too
7496 early. This flag is enabled by default at @option{-O} and higher.
7498 @item -ftree-copyrename
7499 @opindex ftree-copyrename
7500 Perform copy renaming on trees. This pass attempts to rename compiler
7501 temporaries to other variables at copy locations, usually resulting in
7502 variable names which more closely resemble the original variables. This flag
7503 is enabled by default at @option{-O} and higher.
7507 Perform temporary expression replacement during the SSA->normal phase. Single
7508 use/single def temporaries are replaced at their use location with their
7509 defining expression. This results in non-GIMPLE code, but gives the expanders
7510 much more complex trees to work on resulting in better RTL generation. This is
7511 enabled by default at @option{-O} and higher.
7513 @item -ftree-vectorize
7514 @opindex ftree-vectorize
7515 Perform loop vectorization on trees. This flag is enabled by default at
7518 @item -ftree-slp-vectorize
7519 @opindex ftree-slp-vectorize
7520 Perform basic block vectorization on trees. This flag is enabled by default at
7521 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7523 @item -ftree-vect-loop-version
7524 @opindex ftree-vect-loop-version
7525 Perform loop versioning when doing loop vectorization on trees. When a loop
7526 appears to be vectorizable except that data alignment or data dependence cannot
7527 be determined at compile time, then vectorized and non-vectorized versions of
7528 the loop are generated along with run-time checks for alignment or dependence
7529 to control which version is executed. This option is enabled by default
7530 except at level @option{-Os} where it is disabled.
7532 @item -fvect-cost-model
7533 @opindex fvect-cost-model
7534 Enable cost model for vectorization.
7538 Perform Value Range Propagation on trees. This is similar to the
7539 constant propagation pass, but instead of values, ranges of values are
7540 propagated. This allows the optimizers to remove unnecessary range
7541 checks like array bound checks and null pointer checks. This is
7542 enabled by default at @option{-O2} and higher. Null pointer check
7543 elimination is only done if @option{-fdelete-null-pointer-checks} is
7548 Perform tail duplication to enlarge superblock size. This transformation
7549 simplifies the control flow of the function allowing other optimizations to do
7552 @item -funroll-loops
7553 @opindex funroll-loops
7554 Unroll loops whose number of iterations can be determined at compile
7555 time or upon entry to the loop. @option{-funroll-loops} implies
7556 @option{-frerun-cse-after-loop}. This option makes code larger,
7557 and may or may not make it run faster.
7559 @item -funroll-all-loops
7560 @opindex funroll-all-loops
7561 Unroll all loops, even if their number of iterations is uncertain when
7562 the loop is entered. This usually makes programs run more slowly.
7563 @option{-funroll-all-loops} implies the same options as
7564 @option{-funroll-loops},
7566 @item -fsplit-ivs-in-unroller
7567 @opindex fsplit-ivs-in-unroller
7568 Enables expressing of values of induction variables in later iterations
7569 of the unrolled loop using the value in the first iteration. This breaks
7570 long dependency chains, thus improving efficiency of the scheduling passes.
7572 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7573 same effect. However in cases the loop body is more complicated than
7574 a single basic block, this is not reliable. It also does not work at all
7575 on some of the architectures due to restrictions in the CSE pass.
7577 This optimization is enabled by default.
7579 @item -fvariable-expansion-in-unroller
7580 @opindex fvariable-expansion-in-unroller
7581 With this option, the compiler creates multiple copies of some
7582 local variables when unrolling a loop, which can result in superior code.
7584 @item -fpartial-inlining
7585 @opindex fpartial-inlining
7586 Inline parts of functions. This option has any effect only
7587 when inlining itself is turned on by the @option{-finline-functions}
7588 or @option{-finline-small-functions} options.
7590 Enabled at level @option{-O2}.
7592 @item -fpredictive-commoning
7593 @opindex fpredictive-commoning
7594 Perform predictive commoning optimization, i.e., reusing computations
7595 (especially memory loads and stores) performed in previous
7596 iterations of loops.
7598 This option is enabled at level @option{-O3}.
7600 @item -fprefetch-loop-arrays
7601 @opindex fprefetch-loop-arrays
7602 If supported by the target machine, generate instructions to prefetch
7603 memory to improve the performance of loops that access large arrays.
7605 This option may generate better or worse code; results are highly
7606 dependent on the structure of loops within the source code.
7608 Disabled at level @option{-Os}.
7611 @itemx -fno-peephole2
7612 @opindex fno-peephole
7613 @opindex fno-peephole2
7614 Disable any machine-specific peephole optimizations. The difference
7615 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7616 are implemented in the compiler; some targets use one, some use the
7617 other, a few use both.
7619 @option{-fpeephole} is enabled by default.
7620 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7622 @item -fno-guess-branch-probability
7623 @opindex fno-guess-branch-probability
7624 Do not guess branch probabilities using heuristics.
7626 GCC uses heuristics to guess branch probabilities if they are
7627 not provided by profiling feedback (@option{-fprofile-arcs}). These
7628 heuristics are based on the control flow graph. If some branch probabilities
7629 are specified by @samp{__builtin_expect}, then the heuristics are
7630 used to guess branch probabilities for the rest of the control flow graph,
7631 taking the @samp{__builtin_expect} info into account. The interactions
7632 between the heuristics and @samp{__builtin_expect} can be complex, and in
7633 some cases, it may be useful to disable the heuristics so that the effects
7634 of @samp{__builtin_expect} are easier to understand.
7636 The default is @option{-fguess-branch-probability} at levels
7637 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7639 @item -freorder-blocks
7640 @opindex freorder-blocks
7641 Reorder basic blocks in the compiled function in order to reduce number of
7642 taken branches and improve code locality.
7644 Enabled at levels @option{-O2}, @option{-O3}.
7646 @item -freorder-blocks-and-partition
7647 @opindex freorder-blocks-and-partition
7648 In addition to reordering basic blocks in the compiled function, in order
7649 to reduce number of taken branches, partitions hot and cold basic blocks
7650 into separate sections of the assembly and .o files, to improve
7651 paging and cache locality performance.
7653 This optimization is automatically turned off in the presence of
7654 exception handling, for linkonce sections, for functions with a user-defined
7655 section attribute and on any architecture that does not support named
7658 @item -freorder-functions
7659 @opindex freorder-functions
7660 Reorder functions in the object file in order to
7661 improve code locality. This is implemented by using special
7662 subsections @code{.text.hot} for most frequently executed functions and
7663 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7664 the linker so object file format must support named sections and linker must
7665 place them in a reasonable way.
7667 Also profile feedback must be available in to make this option effective. See
7668 @option{-fprofile-arcs} for details.
7670 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7672 @item -fstrict-aliasing
7673 @opindex fstrict-aliasing
7674 Allow the compiler to assume the strictest aliasing rules applicable to
7675 the language being compiled. For C (and C++), this activates
7676 optimizations based on the type of expressions. In particular, an
7677 object of one type is assumed never to reside at the same address as an
7678 object of a different type, unless the types are almost the same. For
7679 example, an @code{unsigned int} can alias an @code{int}, but not a
7680 @code{void*} or a @code{double}. A character type may alias any other
7683 @anchor{Type-punning}Pay special attention to code like this:
7696 The practice of reading from a different union member than the one most
7697 recently written to (called ``type-punning'') is common. Even with
7698 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7699 is accessed through the union type. So, the code above works as
7700 expected. @xref{Structures unions enumerations and bit-fields
7701 implementation}. However, this code might not:
7712 Similarly, access by taking the address, casting the resulting pointer
7713 and dereferencing the result has undefined behavior, even if the cast
7714 uses a union type, e.g.:
7718 return ((union a_union *) &d)->i;
7722 The @option{-fstrict-aliasing} option is enabled at levels
7723 @option{-O2}, @option{-O3}, @option{-Os}.
7725 @item -fstrict-overflow
7726 @opindex fstrict-overflow
7727 Allow the compiler to assume strict signed overflow rules, depending
7728 on the language being compiled. For C (and C++) this means that
7729 overflow when doing arithmetic with signed numbers is undefined, which
7730 means that the compiler may assume that it does not happen. This
7731 permits various optimizations. For example, the compiler assumes
7732 that an expression like @code{i + 10 > i} is always true for
7733 signed @code{i}. This assumption is only valid if signed overflow is
7734 undefined, as the expression is false if @code{i + 10} overflows when
7735 using twos complement arithmetic. When this option is in effect any
7736 attempt to determine whether an operation on signed numbers
7737 overflows must be written carefully to not actually involve overflow.
7739 This option also allows the compiler to assume strict pointer
7740 semantics: given a pointer to an object, if adding an offset to that
7741 pointer does not produce a pointer to the same object, the addition is
7742 undefined. This permits the compiler to conclude that @code{p + u >
7743 p} is always true for a pointer @code{p} and unsigned integer
7744 @code{u}. This assumption is only valid because pointer wraparound is
7745 undefined, as the expression is false if @code{p + u} overflows using
7746 twos complement arithmetic.
7748 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7749 that integer signed overflow is fully defined: it wraps. When
7750 @option{-fwrapv} is used, there is no difference between
7751 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7752 integers. With @option{-fwrapv} certain types of overflow are
7753 permitted. For example, if the compiler gets an overflow when doing
7754 arithmetic on constants, the overflowed value can still be used with
7755 @option{-fwrapv}, but not otherwise.
7757 The @option{-fstrict-overflow} option is enabled at levels
7758 @option{-O2}, @option{-O3}, @option{-Os}.
7760 @item -falign-functions
7761 @itemx -falign-functions=@var{n}
7762 @opindex falign-functions
7763 Align the start of functions to the next power-of-two greater than
7764 @var{n}, skipping up to @var{n} bytes. For instance,
7765 @option{-falign-functions=32} aligns functions to the next 32-byte
7766 boundary, but @option{-falign-functions=24} aligns to the next
7767 32-byte boundary only if this can be done by skipping 23 bytes or less.
7769 @option{-fno-align-functions} and @option{-falign-functions=1} are
7770 equivalent and mean that functions are not aligned.
7772 Some assemblers only support this flag when @var{n} is a power of two;
7773 in that case, it is rounded up.
7775 If @var{n} is not specified or is zero, use a machine-dependent default.
7777 Enabled at levels @option{-O2}, @option{-O3}.
7779 @item -falign-labels
7780 @itemx -falign-labels=@var{n}
7781 @opindex falign-labels
7782 Align all branch targets to a power-of-two boundary, skipping up to
7783 @var{n} bytes like @option{-falign-functions}. This option can easily
7784 make code slower, because it must insert dummy operations for when the
7785 branch target is reached in the usual flow of the code.
7787 @option{-fno-align-labels} and @option{-falign-labels=1} are
7788 equivalent and mean that labels are not aligned.
7790 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7791 are greater than this value, then their values are used instead.
7793 If @var{n} is not specified or is zero, use a machine-dependent default
7794 which is very likely to be @samp{1}, meaning no alignment.
7796 Enabled at levels @option{-O2}, @option{-O3}.
7799 @itemx -falign-loops=@var{n}
7800 @opindex falign-loops
7801 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7802 like @option{-falign-functions}. If the loops are
7803 executed many times, this makes up for any execution of the dummy
7806 @option{-fno-align-loops} and @option{-falign-loops=1} are
7807 equivalent and mean that loops are not aligned.
7809 If @var{n} is not specified or is zero, use a machine-dependent default.
7811 Enabled at levels @option{-O2}, @option{-O3}.
7814 @itemx -falign-jumps=@var{n}
7815 @opindex falign-jumps
7816 Align branch targets to a power-of-two boundary, for branch targets
7817 where the targets can only be reached by jumping, skipping up to @var{n}
7818 bytes like @option{-falign-functions}. In this case, no dummy operations
7821 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7822 equivalent and mean that loops are not aligned.
7824 If @var{n} is not specified or is zero, use a machine-dependent default.
7826 Enabled at levels @option{-O2}, @option{-O3}.
7828 @item -funit-at-a-time
7829 @opindex funit-at-a-time
7830 This option is left for compatibility reasons. @option{-funit-at-a-time}
7831 has no effect, while @option{-fno-unit-at-a-time} implies
7832 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7836 @item -fno-toplevel-reorder
7837 @opindex fno-toplevel-reorder
7838 Do not reorder top-level functions, variables, and @code{asm}
7839 statements. Output them in the same order that they appear in the
7840 input file. When this option is used, unreferenced static variables
7841 are not removed. This option is intended to support existing code
7842 that relies on a particular ordering. For new code, it is better to
7845 Enabled at level @option{-O0}. When disabled explicitly, it also implies
7846 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
7851 Constructs webs as commonly used for register allocation purposes and assign
7852 each web individual pseudo register. This allows the register allocation pass
7853 to operate on pseudos directly, but also strengthens several other optimization
7854 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7855 however, make debugging impossible, since variables no longer stay in a
7858 Enabled by default with @option{-funroll-loops}.
7860 @item -fwhole-program
7861 @opindex fwhole-program
7862 Assume that the current compilation unit represents the whole program being
7863 compiled. All public functions and variables with the exception of @code{main}
7864 and those merged by attribute @code{externally_visible} become static functions
7865 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.
7866 While this option is equivalent to proper use of the @code{static} keyword for
7867 programs consisting of a single file, in combination with option
7868 @option{-flto} this flag can be used to
7869 compile many smaller scale programs since the functions and variables become
7870 local for the whole combined compilation unit, not for the single source file
7873 This option implies @option{-fwhole-file} for Fortran programs.
7875 @item -flto[=@var{n}]
7877 This option runs the standard link-time optimizer. When invoked
7878 with source code, it generates GIMPLE (one of GCC's internal
7879 representations) and writes it to special ELF sections in the object
7880 file. When the object files are linked together, all the function
7881 bodies are read from these ELF sections and instantiated as if they
7882 had been part of the same translation unit.
7884 To use the link-time optimizer, @option{-flto} needs to be specified at
7885 compile time and during the final link. For example:
7888 gcc -c -O2 -flto foo.c
7889 gcc -c -O2 -flto bar.c
7890 gcc -o myprog -flto -O2 foo.o bar.o
7893 The first two invocations to GCC save a bytecode representation
7894 of GIMPLE into special ELF sections inside @file{foo.o} and
7895 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
7896 @file{foo.o} and @file{bar.o}, merges the two files into a single
7897 internal image, and compiles the result as usual. Since both
7898 @file{foo.o} and @file{bar.o} are merged into a single image, this
7899 causes all the interprocedural analyses and optimizations in GCC to
7900 work across the two files as if they were a single one. This means,
7901 for example, that the inliner is able to inline functions in
7902 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7904 Another (simpler) way to enable link-time optimization is:
7907 gcc -o myprog -flto -O2 foo.c bar.c
7910 The above generates bytecode for @file{foo.c} and @file{bar.c},
7911 merges them together into a single GIMPLE representation and optimizes
7912 them as usual to produce @file{myprog}.
7914 The only important thing to keep in mind is that to enable link-time
7915 optimizations the @option{-flto} flag needs to be passed to both the
7916 compile and the link commands.
7918 To make whole program optimization effective, it is necessary to make
7919 certain whole program assumptions. The compiler needs to know
7920 what functions and variables can be accessed by libraries and runtime
7921 outside of the link-time optimized unit. When supported by the linker,
7922 the linker plugin (see @option{-fuse-linker-plugin}) passes information
7923 to the compiler about used and externally visible symbols. When
7924 the linker plugin is not available, @option{-fwhole-program} should be
7925 used to allow the compiler to make these assumptions, which leads
7926 to more aggressive optimization decisions.
7928 Note that when a file is compiled with @option{-flto}, the generated
7929 object file is larger than a regular object file because it
7930 contains GIMPLE bytecodes and the usual final code. This means that
7931 object files with LTO information can be linked as normal object
7932 files; if @option{-flto} is not passed to the linker, no
7933 interprocedural optimizations are applied.
7935 Additionally, the optimization flags used to compile individual files
7936 are not necessarily related to those used at link time. For instance,
7939 gcc -c -O0 -flto foo.c
7940 gcc -c -O0 -flto bar.c
7941 gcc -o myprog -flto -O3 foo.o bar.o
7944 This produces individual object files with unoptimized assembler
7945 code, but the resulting binary @file{myprog} is optimized at
7946 @option{-O3}. If, instead, the final binary is generated without
7947 @option{-flto}, then @file{myprog} is not optimized.
7949 When producing the final binary with @option{-flto}, GCC only
7950 applies link-time optimizations to those files that contain bytecode.
7951 Therefore, you can mix and match object files and libraries with
7952 GIMPLE bytecodes and final object code. GCC automatically selects
7953 which files to optimize in LTO mode and which files to link without
7956 There are some code generation flags preserved by GCC when
7957 generating bytecodes, as they need to be used during the final link
7958 stage. Currently, the following options are saved into the GIMPLE
7959 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7960 @option{-m} target flags.
7962 At link time, these options are read in and reapplied. Note that the
7963 current implementation makes no attempt to recognize conflicting
7964 values for these options. If different files have conflicting option
7965 values (e.g., one file is compiled with @option{-fPIC} and another
7966 isn't), the compiler simply uses the last value read from the
7967 bytecode files. It is recommended, then, that you compile all the files
7968 participating in the same link with the same options.
7970 If LTO encounters objects with C linkage declared with incompatible
7971 types in separate translation units to be linked together (undefined
7972 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7973 issued. The behavior is still undefined at run time.
7975 Another feature of LTO is that it is possible to apply interprocedural
7976 optimizations on files written in different languages. This requires
7977 support in the language front end. Currently, the C, C++ and
7978 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7979 something like this should work:
7984 gfortran -c -flto baz.f90
7985 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7988 Notice that the final link is done with @command{g++} to get the C++
7989 runtime libraries and @option{-lgfortran} is added to get the Fortran
7990 runtime libraries. In general, when mixing languages in LTO mode, you
7991 should use the same link command options as when mixing languages in a
7992 regular (non-LTO) compilation; all you need to add is @option{-flto} to
7993 all the compile and link commands.
7995 If object files containing GIMPLE bytecode are stored in a library archive, say
7996 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7997 are using a linker with plugin support. To enable this feature, use
7998 the flag @option{-fuse-linker-plugin} at link time:
8001 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8004 With the linker plugin enabled, the linker extracts the needed
8005 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8006 to make them part of the aggregated GIMPLE image to be optimized.
8008 If you are not using a linker with plugin support and/or do not
8009 enable the linker plugin, then the objects inside @file{libfoo.a}
8010 are extracted and linked as usual, but they do not participate
8011 in the LTO optimization process.
8013 Link-time optimizations do not require the presence of the whole program to
8014 operate. If the program does not require any symbols to be exported, it is
8015 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8016 the interprocedural optimizers to use more aggressive assumptions which may
8017 lead to improved optimization opportunities.
8018 Use of @option{-fwhole-program} is not needed when linker plugin is
8019 active (see @option{-fuse-linker-plugin}).
8021 The current implementation of LTO makes no
8022 attempt to generate bytecode that is portable between different
8023 types of hosts. The bytecode files are versioned and there is a
8024 strict version check, so bytecode files generated in one version of
8025 GCC will not work with an older/newer version of GCC@.
8027 Link-time optimization does not work well with generation of debugging
8028 information. Combining @option{-flto} with
8029 @option{-g} is currently experimental and expected to produce wrong
8032 If you specify the optional @var{n}, the optimization and code
8033 generation done at link time is executed in parallel using @var{n}
8034 parallel jobs by utilizing an installed @command{make} program. The
8035 environment variable @env{MAKE} may be used to override the program
8036 used. The default value for @var{n} is 1.
8038 You can also specify @option{-flto=jobserver} to use GNU make's
8039 job server mode to determine the number of parallel jobs. This
8040 is useful when the Makefile calling GCC is already executing in parallel.
8041 You must prepend a @samp{+} to the command recipe in the parent Makefile
8042 for this to work. This option likely only works if @env{MAKE} is
8045 This option is disabled by default
8047 @item -flto-partition=@var{alg}
8048 @opindex flto-partition
8049 Specify the partitioning algorithm used by the link-time optimizer.
8050 The value is either @code{1to1} to specify a partitioning mirroring
8051 the original source files or @code{balanced} to specify partitioning
8052 into equally sized chunks (whenever possible). Specifying @code{none}
8053 as an algorithm disables partitioning and streaming completely. The
8054 default value is @code{balanced}.
8056 @item -flto-compression-level=@var{n}
8057 This option specifies the level of compression used for intermediate
8058 language written to LTO object files, and is only meaningful in
8059 conjunction with LTO mode (@option{-flto}). Valid
8060 values are 0 (no compression) to 9 (maximum compression). Values
8061 outside this range are clamped to either 0 or 9. If the option is not
8062 given, a default balanced compression setting is used.
8065 Prints a report with internal details on the workings of the link-time
8066 optimizer. The contents of this report vary from version to version.
8067 It is meant to be useful to GCC developers when processing object
8068 files in LTO mode (via @option{-flto}).
8070 Disabled by default.
8072 @item -fuse-linker-plugin
8073 Enables the use of a linker plugin during link-time optimization. This
8074 option relies on plugin support in the linker, which is available in gold
8075 or in GNU ld 2.21 or newer.
8077 This option enables the extraction of object files with GIMPLE bytecode out
8078 of library archives. This improves the quality of optimization by exposing
8079 more code to the link-time optimizer. This information specifies what
8080 symbols can be accessed externally (by non-LTO object or during dynamic
8081 linking). Resulting code quality improvements on binaries (and shared
8082 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
8083 See @option{-flto} for a description of the effect of this flag and how to
8086 This option is enabled by default when LTO support in GCC is enabled
8087 and GCC was configured for use with
8088 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8090 @item -ffat-lto-objects
8091 @opindex ffat-lto-objects
8092 Fat LTO objects are object files that contain both the intermediate language
8093 and the object code. This makes them usable for both LTO linking and normal
8094 linking. This option is effective only when compiling with @option{-flto}
8095 and is ignored at link time.
8097 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8098 requires the complete toolchain to be aware of LTO. It requires a linker with
8099 linker plugin support for basic functionality. Additionally,
8100 @command{nm}, @command{ar} and @command{ranlib}
8101 need to support linker plugins to allow a full-featured build environment
8102 (capable of building static libraries etc).
8104 The default is @option{-ffat-lto-objects} but this default is intended to
8105 change in future releases when linker plugin enabled environments become more
8108 @item -fcompare-elim
8109 @opindex fcompare-elim
8110 After register allocation and post-register allocation instruction splitting,
8111 identify arithmetic instructions that compute processor flags similar to a
8112 comparison operation based on that arithmetic. If possible, eliminate the
8113 explicit comparison operation.
8115 This pass only applies to certain targets that cannot explicitly represent
8116 the comparison operation before register allocation is complete.
8118 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8120 @item -fcprop-registers
8121 @opindex fcprop-registers
8122 After register allocation and post-register allocation instruction splitting,
8123 we perform a copy-propagation pass to try to reduce scheduling dependencies
8124 and occasionally eliminate the copy.
8126 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8128 @item -fprofile-correction
8129 @opindex fprofile-correction
8130 Profiles collected using an instrumented binary for multi-threaded programs may
8131 be inconsistent due to missed counter updates. When this option is specified,
8132 GCC uses heuristics to correct or smooth out such inconsistencies. By
8133 default, GCC emits an error message when an inconsistent profile is detected.
8135 @item -fprofile-dir=@var{path}
8136 @opindex fprofile-dir
8138 Set the directory to search for the profile data files in to @var{path}.
8139 This option affects only the profile data generated by
8140 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8141 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8142 and its related options. Both absolute and relative paths can be used.
8143 By default, GCC uses the current directory as @var{path}, thus the
8144 profile data file appears in the same directory as the object file.
8146 @item -fprofile-generate
8147 @itemx -fprofile-generate=@var{path}
8148 @opindex fprofile-generate
8150 Enable options usually used for instrumenting application to produce
8151 profile useful for later recompilation with profile feedback based
8152 optimization. You must use @option{-fprofile-generate} both when
8153 compiling and when linking your program.
8155 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
8157 If @var{path} is specified, GCC looks at the @var{path} to find
8158 the profile feedback data files. See @option{-fprofile-dir}.
8161 @itemx -fprofile-use=@var{path}
8162 @opindex fprofile-use
8163 Enable profile feedback directed optimizations, and optimizations
8164 generally profitable only with profile feedback available.
8166 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8167 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
8169 By default, GCC emits an error message if the feedback profiles do not
8170 match the source code. This error can be turned into a warning by using
8171 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8174 If @var{path} is specified, GCC looks at the @var{path} to find
8175 the profile feedback data files. See @option{-fprofile-dir}.
8178 The following options control compiler behavior regarding floating-point
8179 arithmetic. These options trade off between speed and
8180 correctness. All must be specifically enabled.
8184 @opindex ffloat-store
8185 Do not store floating-point variables in registers, and inhibit other
8186 options that might change whether a floating-point value is taken from a
8189 @cindex floating-point precision
8190 This option prevents undesirable excess precision on machines such as
8191 the 68000 where the floating registers (of the 68881) keep more
8192 precision than a @code{double} is supposed to have. Similarly for the
8193 x86 architecture. For most programs, the excess precision does only
8194 good, but a few programs rely on the precise definition of IEEE floating
8195 point. Use @option{-ffloat-store} for such programs, after modifying
8196 them to store all pertinent intermediate computations into variables.
8198 @item -fexcess-precision=@var{style}
8199 @opindex fexcess-precision
8200 This option allows further control over excess precision on machines
8201 where floating-point registers have more precision than the IEEE
8202 @code{float} and @code{double} types and the processor does not
8203 support operations rounding to those types. By default,
8204 @option{-fexcess-precision=fast} is in effect; this means that
8205 operations are carried out in the precision of the registers and that
8206 it is unpredictable when rounding to the types specified in the source
8207 code takes place. When compiling C, if
8208 @option{-fexcess-precision=standard} is specified then excess
8209 precision follows the rules specified in ISO C99; in particular,
8210 both casts and assignments cause values to be rounded to their
8211 semantic types (whereas @option{-ffloat-store} only affects
8212 assignments). This option is enabled by default for C if a strict
8213 conformance option such as @option{-std=c99} is used.
8216 @option{-fexcess-precision=standard} is not implemented for languages
8217 other than C, and has no effect if
8218 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8219 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8220 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8221 semantics apply without excess precision, and in the latter, rounding
8226 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8227 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8228 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8230 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8232 This option is not turned on by any @option{-O} option besides
8233 @option{-Ofast} since it can result in incorrect output for programs
8234 that depend on an exact implementation of IEEE or ISO rules/specifications
8235 for math functions. It may, however, yield faster code for programs
8236 that do not require the guarantees of these specifications.
8238 @item -fno-math-errno
8239 @opindex fno-math-errno
8240 Do not set @code{errno} after calling math functions that are executed
8241 with a single instruction, e.g., @code{sqrt}. A program that relies on
8242 IEEE exceptions for math error handling may want to use this flag
8243 for speed while maintaining IEEE arithmetic compatibility.
8245 This option is not turned on by any @option{-O} option since
8246 it can result in incorrect output for programs that depend on
8247 an exact implementation of IEEE or ISO rules/specifications for
8248 math functions. It may, however, yield faster code for programs
8249 that do not require the guarantees of these specifications.
8251 The default is @option{-fmath-errno}.
8253 On Darwin systems, the math library never sets @code{errno}. There is
8254 therefore no reason for the compiler to consider the possibility that
8255 it might, and @option{-fno-math-errno} is the default.
8257 @item -funsafe-math-optimizations
8258 @opindex funsafe-math-optimizations
8260 Allow optimizations for floating-point arithmetic that (a) assume
8261 that arguments and results are valid and (b) may violate IEEE or
8262 ANSI standards. When used at link-time, it may include libraries
8263 or startup files that change the default FPU control word or other
8264 similar optimizations.
8266 This option is not turned on by any @option{-O} option since
8267 it can result in incorrect output for programs that depend on
8268 an exact implementation of IEEE or ISO rules/specifications for
8269 math functions. It may, however, yield faster code for programs
8270 that do not require the guarantees of these specifications.
8271 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8272 @option{-fassociative-math} and @option{-freciprocal-math}.
8274 The default is @option{-fno-unsafe-math-optimizations}.
8276 @item -fassociative-math
8277 @opindex fassociative-math
8279 Allow re-association of operands in series of floating-point operations.
8280 This violates the ISO C and C++ language standard by possibly changing
8281 computation result. NOTE: re-ordering may change the sign of zero as
8282 well as ignore NaNs and inhibit or create underflow or overflow (and
8283 thus cannot be used on code that relies on rounding behavior like
8284 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8285 and thus may not be used when ordered comparisons are required.
8286 This option requires that both @option{-fno-signed-zeros} and
8287 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8288 much sense with @option{-frounding-math}. For Fortran the option
8289 is automatically enabled when both @option{-fno-signed-zeros} and
8290 @option{-fno-trapping-math} are in effect.
8292 The default is @option{-fno-associative-math}.
8294 @item -freciprocal-math
8295 @opindex freciprocal-math
8297 Allow the reciprocal of a value to be used instead of dividing by
8298 the value if this enables optimizations. For example @code{x / y}
8299 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
8300 is subject to common subexpression elimination. Note that this loses
8301 precision and increases the number of flops operating on the value.
8303 The default is @option{-fno-reciprocal-math}.
8305 @item -ffinite-math-only
8306 @opindex ffinite-math-only
8307 Allow optimizations for floating-point arithmetic that assume
8308 that arguments and results are not NaNs or +-Infs.
8310 This option is not turned on by any @option{-O} option since
8311 it can result in incorrect output for programs that depend on
8312 an exact implementation of IEEE or ISO rules/specifications for
8313 math functions. It may, however, yield faster code for programs
8314 that do not require the guarantees of these specifications.
8316 The default is @option{-fno-finite-math-only}.
8318 @item -fno-signed-zeros
8319 @opindex fno-signed-zeros
8320 Allow optimizations for floating-point arithmetic that ignore the
8321 signedness of zero. IEEE arithmetic specifies the behavior of
8322 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8323 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8324 This option implies that the sign of a zero result isn't significant.
8326 The default is @option{-fsigned-zeros}.
8328 @item -fno-trapping-math
8329 @opindex fno-trapping-math
8330 Compile code assuming that floating-point operations cannot generate
8331 user-visible traps. These traps include division by zero, overflow,
8332 underflow, inexact result and invalid operation. This option requires
8333 that @option{-fno-signaling-nans} be in effect. Setting this option may
8334 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8336 This option should never be turned on by any @option{-O} option since
8337 it can result in incorrect output for programs that depend on
8338 an exact implementation of IEEE or ISO rules/specifications for
8341 The default is @option{-ftrapping-math}.
8343 @item -frounding-math
8344 @opindex frounding-math
8345 Disable transformations and optimizations that assume default floating-point
8346 rounding behavior. This is round-to-zero for all floating point
8347 to integer conversions, and round-to-nearest for all other arithmetic
8348 truncations. This option should be specified for programs that change
8349 the FP rounding mode dynamically, or that may be executed with a
8350 non-default rounding mode. This option disables constant folding of
8351 floating-point expressions at compile time (which may be affected by
8352 rounding mode) and arithmetic transformations that are unsafe in the
8353 presence of sign-dependent rounding modes.
8355 The default is @option{-fno-rounding-math}.
8357 This option is experimental and does not currently guarantee to
8358 disable all GCC optimizations that are affected by rounding mode.
8359 Future versions of GCC may provide finer control of this setting
8360 using C99's @code{FENV_ACCESS} pragma. This command-line option
8361 will be used to specify the default state for @code{FENV_ACCESS}.
8363 @item -fsignaling-nans
8364 @opindex fsignaling-nans
8365 Compile code assuming that IEEE signaling NaNs may generate user-visible
8366 traps during floating-point operations. Setting this option disables
8367 optimizations that may change the number of exceptions visible with
8368 signaling NaNs. This option implies @option{-ftrapping-math}.
8370 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8373 The default is @option{-fno-signaling-nans}.
8375 This option is experimental and does not currently guarantee to
8376 disable all GCC optimizations that affect signaling NaN behavior.
8378 @item -fsingle-precision-constant
8379 @opindex fsingle-precision-constant
8380 Treat floating-point constants as single precision instead of
8381 implicitly converting them to double-precision constants.
8383 @item -fcx-limited-range
8384 @opindex fcx-limited-range
8385 When enabled, this option states that a range reduction step is not
8386 needed when performing complex division. Also, there is no checking
8387 whether the result of a complex multiplication or division is @code{NaN
8388 + I*NaN}, with an attempt to rescue the situation in that case. The
8389 default is @option{-fno-cx-limited-range}, but is enabled by
8390 @option{-ffast-math}.
8392 This option controls the default setting of the ISO C99
8393 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8396 @item -fcx-fortran-rules
8397 @opindex fcx-fortran-rules
8398 Complex multiplication and division follow Fortran rules. Range
8399 reduction is done as part of complex division, but there is no checking
8400 whether the result of a complex multiplication or division is @code{NaN
8401 + I*NaN}, with an attempt to rescue the situation in that case.
8403 The default is @option{-fno-cx-fortran-rules}.
8407 The following options control optimizations that may improve
8408 performance, but are not enabled by any @option{-O} options. This
8409 section includes experimental options that may produce broken code.
8412 @item -fbranch-probabilities
8413 @opindex fbranch-probabilities
8414 After running a program compiled with @option{-fprofile-arcs}
8415 (@pxref{Debugging Options,, Options for Debugging Your Program or
8416 @command{gcc}}), you can compile it a second time using
8417 @option{-fbranch-probabilities}, to improve optimizations based on
8418 the number of times each branch was taken. When the program
8419 compiled with @option{-fprofile-arcs} exits it saves arc execution
8420 counts to a file called @file{@var{sourcename}.gcda} for each source
8421 file. The information in this data file is very dependent on the
8422 structure of the generated code, so you must use the same source code
8423 and the same optimization options for both compilations.
8425 With @option{-fbranch-probabilities}, GCC puts a
8426 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8427 These can be used to improve optimization. Currently, they are only
8428 used in one place: in @file{reorg.c}, instead of guessing which path a
8429 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8430 exactly determine which path is taken more often.
8432 @item -fprofile-values
8433 @opindex fprofile-values
8434 If combined with @option{-fprofile-arcs}, it adds code so that some
8435 data about values of expressions in the program is gathered.
8437 With @option{-fbranch-probabilities}, it reads back the data gathered
8438 from profiling values of expressions for usage in optimizations.
8440 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8444 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8445 a code to gather information about values of expressions.
8447 With @option{-fbranch-probabilities}, it reads back the data gathered
8448 and actually performs the optimizations based on them.
8449 Currently the optimizations include specialization of division operation
8450 using the knowledge about the value of the denominator.
8452 @item -frename-registers
8453 @opindex frename-registers
8454 Attempt to avoid false dependencies in scheduled code by making use
8455 of registers left over after register allocation. This optimization
8456 most benefits processors with lots of registers. Depending on the
8457 debug information format adopted by the target, however, it can
8458 make debugging impossible, since variables no longer stay in
8459 a ``home register''.
8461 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8465 Perform tail duplication to enlarge superblock size. This transformation
8466 simplifies the control flow of the function allowing other optimizations to do
8469 Enabled with @option{-fprofile-use}.
8471 @item -funroll-loops
8472 @opindex funroll-loops
8473 Unroll loops whose number of iterations can be determined at compile time or
8474 upon entry to the loop. @option{-funroll-loops} implies
8475 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8476 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8477 small constant number of iterations). This option makes code larger, and may
8478 or may not make it run faster.
8480 Enabled with @option{-fprofile-use}.
8482 @item -funroll-all-loops
8483 @opindex funroll-all-loops
8484 Unroll all loops, even if their number of iterations is uncertain when
8485 the loop is entered. This usually makes programs run more slowly.
8486 @option{-funroll-all-loops} implies the same options as
8487 @option{-funroll-loops}.
8490 @opindex fpeel-loops
8491 Peels loops for which there is enough information that they do not
8492 roll much (from profile feedback). It also turns on complete loop peeling
8493 (i.e.@: complete removal of loops with small constant number of iterations).
8495 Enabled with @option{-fprofile-use}.
8497 @item -fmove-loop-invariants
8498 @opindex fmove-loop-invariants
8499 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8500 at level @option{-O1}
8502 @item -funswitch-loops
8503 @opindex funswitch-loops
8504 Move branches with loop invariant conditions out of the loop, with duplicates
8505 of the loop on both branches (modified according to result of the condition).
8507 @item -ffunction-sections
8508 @itemx -fdata-sections
8509 @opindex ffunction-sections
8510 @opindex fdata-sections
8511 Place each function or data item into its own section in the output
8512 file if the target supports arbitrary sections. The name of the
8513 function or the name of the data item determines the section's name
8516 Use these options on systems where the linker can perform optimizations
8517 to improve locality of reference in the instruction space. Most systems
8518 using the ELF object format and SPARC processors running Solaris 2 have
8519 linkers with such optimizations. AIX may have these optimizations in
8522 Only use these options when there are significant benefits from doing
8523 so. When you specify these options, the assembler and linker
8524 create larger object and executable files and are also slower.
8525 You cannot use @code{gprof} on all systems if you
8526 specify this option, and you may have problems with debugging if
8527 you specify both this option and @option{-g}.
8529 @item -fbranch-target-load-optimize
8530 @opindex fbranch-target-load-optimize
8531 Perform branch target register load optimization before prologue / epilogue
8533 The use of target registers can typically be exposed only during reload,
8534 thus hoisting loads out of loops and doing inter-block scheduling needs
8535 a separate optimization pass.
8537 @item -fbranch-target-load-optimize2
8538 @opindex fbranch-target-load-optimize2
8539 Perform branch target register load optimization after prologue / epilogue
8542 @item -fbtr-bb-exclusive
8543 @opindex fbtr-bb-exclusive
8544 When performing branch target register load optimization, don't reuse
8545 branch target registers in within any basic block.
8547 @item -fstack-protector
8548 @opindex fstack-protector
8549 Emit extra code to check for buffer overflows, such as stack smashing
8550 attacks. This is done by adding a guard variable to functions with
8551 vulnerable objects. This includes functions that call @code{alloca}, and
8552 functions with buffers larger than 8 bytes. The guards are initialized
8553 when a function is entered and then checked when the function exits.
8554 If a guard check fails, an error message is printed and the program exits.
8556 @item -fstack-protector-all
8557 @opindex fstack-protector-all
8558 Like @option{-fstack-protector} except that all functions are protected.
8560 @item -fsection-anchors
8561 @opindex fsection-anchors
8562 Try to reduce the number of symbolic address calculations by using
8563 shared ``anchor'' symbols to address nearby objects. This transformation
8564 can help to reduce the number of GOT entries and GOT accesses on some
8567 For example, the implementation of the following function @code{foo}:
8571 int foo (void) @{ return a + b + c; @}
8575 usually calculates the addresses of all three variables, but if you
8576 compile it with @option{-fsection-anchors}, it accesses the variables
8577 from a common anchor point instead. The effect is similar to the
8578 following pseudocode (which isn't valid C):
8583 register int *xr = &x;
8584 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8588 Not all targets support this option.
8590 @item --param @var{name}=@var{value}
8592 In some places, GCC uses various constants to control the amount of
8593 optimization that is done. For example, GCC does not inline functions
8594 that contain more than a certain number of instructions. You can
8595 control some of these constants on the command line using the
8596 @option{--param} option.
8598 The names of specific parameters, and the meaning of the values, are
8599 tied to the internals of the compiler, and are subject to change
8600 without notice in future releases.
8602 In each case, the @var{value} is an integer. The allowable choices for
8603 @var{name} are given in the following table:
8606 @item predictable-branch-outcome
8607 When branch is predicted to be taken with probability lower than this threshold
8608 (in percent), then it is considered well predictable. The default is 10.
8610 @item max-crossjump-edges
8611 The maximum number of incoming edges to consider for crossjumping.
8612 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8613 the number of edges incoming to each block. Increasing values mean
8614 more aggressive optimization, making the compilation time increase with
8615 probably small improvement in executable size.
8617 @item min-crossjump-insns
8618 The minimum number of instructions that must be matched at the end
8619 of two blocks before crossjumping is performed on them. This
8620 value is ignored in the case where all instructions in the block being
8621 crossjumped from are matched. The default value is 5.
8623 @item max-grow-copy-bb-insns
8624 The maximum code size expansion factor when copying basic blocks
8625 instead of jumping. The expansion is relative to a jump instruction.
8626 The default value is 8.
8628 @item max-goto-duplication-insns
8629 The maximum number of instructions to duplicate to a block that jumps
8630 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8631 passes, GCC factors computed gotos early in the compilation process,
8632 and unfactors them as late as possible. Only computed jumps at the
8633 end of a basic blocks with no more than max-goto-duplication-insns are
8634 unfactored. The default value is 8.
8636 @item max-delay-slot-insn-search
8637 The maximum number of instructions to consider when looking for an
8638 instruction to fill a delay slot. If more than this arbitrary number of
8639 instructions are searched, the time savings from filling the delay slot
8640 are minimal, so stop searching. Increasing values mean more
8641 aggressive optimization, making the compilation time increase with probably
8642 small improvement in execution time.
8644 @item max-delay-slot-live-search
8645 When trying to fill delay slots, the maximum number of instructions to
8646 consider when searching for a block with valid live register
8647 information. Increasing this arbitrarily chosen value means more
8648 aggressive optimization, increasing the compilation time. This parameter
8649 should be removed when the delay slot code is rewritten to maintain the
8652 @item max-gcse-memory
8653 The approximate maximum amount of memory that can be allocated in
8654 order to perform the global common subexpression elimination
8655 optimization. If more memory than specified is required, the
8656 optimization is not done.
8658 @item max-gcse-insertion-ratio
8659 If the ratio of expression insertions to deletions is larger than this value
8660 for any expression, then RTL PRE inserts or removes the expression and thus
8661 leaves partially redundant computations in the instruction stream. The default value is 20.
8663 @item max-pending-list-length
8664 The maximum number of pending dependencies scheduling allows
8665 before flushing the current state and starting over. Large functions
8666 with few branches or calls can create excessively large lists which
8667 needlessly consume memory and resources.
8669 @item max-modulo-backtrack-attempts
8670 The maximum number of backtrack attempts the scheduler should make
8671 when modulo scheduling a loop. Larger values can exponentially increase
8674 @item max-inline-insns-single
8675 Several parameters control the tree inliner used in GCC@.
8676 This number sets the maximum number of instructions (counted in GCC's
8677 internal representation) in a single function that the tree inliner
8678 considers for inlining. This only affects functions declared
8679 inline and methods implemented in a class declaration (C++).
8680 The default value is 400.
8682 @item max-inline-insns-auto
8683 When you use @option{-finline-functions} (included in @option{-O3}),
8684 a lot of functions that would otherwise not be considered for inlining
8685 by the compiler are investigated. To those functions, a different
8686 (more restrictive) limit compared to functions declared inline can
8688 The default value is 40.
8690 @item large-function-insns
8691 The limit specifying really large functions. For functions larger than this
8692 limit after inlining, inlining is constrained by
8693 @option{--param large-function-growth}. This parameter is useful primarily
8694 to avoid extreme compilation time caused by non-linear algorithms used by the
8696 The default value is 2700.
8698 @item large-function-growth
8699 Specifies maximal growth of large function caused by inlining in percents.
8700 The default value is 100 which limits large function growth to 2.0 times
8703 @item large-unit-insns
8704 The limit specifying large translation unit. Growth caused by inlining of
8705 units larger than this limit is limited by @option{--param inline-unit-growth}.
8706 For small units this might be too tight (consider unit consisting of function A
8707 that is inline and B that just calls A three time. If B is small relative to
8708 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8709 large units consisting of small inlineable functions however the overall unit
8710 growth limit is needed to avoid exponential explosion of code size. Thus for
8711 smaller units, the size is increased to @option{--param large-unit-insns}
8712 before applying @option{--param inline-unit-growth}. The default is 10000
8714 @item inline-unit-growth
8715 Specifies maximal overall growth of the compilation unit caused by inlining.
8716 The default value is 30 which limits unit growth to 1.3 times the original
8719 @item ipcp-unit-growth
8720 Specifies maximal overall growth of the compilation unit caused by
8721 interprocedural constant propagation. The default value is 10 which limits
8722 unit growth to 1.1 times the original size.
8724 @item large-stack-frame
8725 The limit specifying large stack frames. While inlining the algorithm is trying
8726 to not grow past this limit too much. Default value is 256 bytes.
8728 @item large-stack-frame-growth
8729 Specifies maximal growth of large stack frames caused by inlining in percents.
8730 The default value is 1000 which limits large stack frame growth to 11 times
8733 @item max-inline-insns-recursive
8734 @itemx max-inline-insns-recursive-auto
8735 Specifies maximum number of instructions out-of-line copy of self recursive inline
8736 function can grow into by performing recursive inlining.
8738 For functions declared inline @option{--param max-inline-insns-recursive} is
8739 taken into account. For function not declared inline, recursive inlining
8740 happens only when @option{-finline-functions} (included in @option{-O3}) is
8741 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8742 default value is 450.
8744 @item max-inline-recursive-depth
8745 @itemx max-inline-recursive-depth-auto
8746 Specifies maximum recursion depth used by the recursive inlining.
8748 For functions declared inline @option{--param max-inline-recursive-depth} is
8749 taken into account. For function not declared inline, recursive inlining
8750 happens only when @option{-finline-functions} (included in @option{-O3}) is
8751 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8754 @item min-inline-recursive-probability
8755 Recursive inlining is profitable only for function having deep recursion
8756 in average and can hurt for function having little recursion depth by
8757 increasing the prologue size or complexity of function body to other
8760 When profile feedback is available (see @option{-fprofile-generate}) the actual
8761 recursion depth can be guessed from probability that function will recurse via
8762 given call expression. This parameter limits inlining only to call expression
8763 whose probability exceeds given threshold (in percents). The default value is
8766 @item early-inlining-insns
8767 Specify growth that early inliner can make. In effect it increases amount of
8768 inlining for code having large abstraction penalty. The default value is 10.
8770 @item max-early-inliner-iterations
8771 @itemx max-early-inliner-iterations
8772 Limit of iterations of early inliner. This basically bounds number of nested
8773 indirect calls early inliner can resolve. Deeper chains are still handled by
8776 @item comdat-sharing-probability
8777 @itemx comdat-sharing-probability
8778 Probability (in percent) that C++ inline function with comdat visibility
8779 are shared across multiple compilation units. The default value is 20.
8781 @item min-vect-loop-bound
8782 The minimum number of iterations under which a loop is not vectorized
8783 when @option{-ftree-vectorize} is used. The number of iterations after
8784 vectorization needs to be greater than the value specified by this option
8785 to allow vectorization. The default value is 0.
8787 @item gcse-cost-distance-ratio
8788 Scaling factor in calculation of maximum distance an expression
8789 can be moved by GCSE optimizations. This is currently supported only in the
8790 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8791 is with simple expressions, i.e., the expressions that have cost
8792 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
8793 hoisting of simple expressions. The default value is 10.
8795 @item gcse-unrestricted-cost
8796 Cost, roughly measured as the cost of a single typical machine
8797 instruction, at which GCSE optimizations do not constrain
8798 the distance an expression can travel. This is currently
8799 supported only in the code hoisting pass. The lesser the cost,
8800 the more aggressive code hoisting is. Specifying 0
8801 allows all expressions to travel unrestricted distances.
8802 The default value is 3.
8804 @item max-hoist-depth
8805 The depth of search in the dominator tree for expressions to hoist.
8806 This is used to avoid quadratic behavior in hoisting algorithm.
8807 The value of 0 does not limit on the search, but may slow down compilation
8808 of huge functions. The default value is 30.
8810 @item max-tail-merge-comparisons
8811 The maximum amount of similar bbs to compare a bb with. This is used to
8812 avoid quadratic behavior in tree tail merging. The default value is 10.
8814 @item max-tail-merge-iterations
8815 The maximum amount of iterations of the pass over the function. This is used to
8816 limit compilation time in tree tail merging. The default value is 2.
8818 @item max-unrolled-insns
8819 The maximum number of instructions that a loop should have if that loop
8820 is unrolled, and if the loop is unrolled, it determines how many times
8821 the loop code is unrolled.
8823 @item max-average-unrolled-insns
8824 The maximum number of instructions biased by probabilities of their execution
8825 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8826 it determines how many times the loop code is unrolled.
8828 @item max-unroll-times
8829 The maximum number of unrollings of a single loop.
8831 @item max-peeled-insns
8832 The maximum number of instructions that a loop should have if that loop
8833 is peeled, and if the loop is peeled, it determines how many times
8834 the loop code is peeled.
8836 @item max-peel-times
8837 The maximum number of peelings of a single loop.
8839 @item max-completely-peeled-insns
8840 The maximum number of insns of a completely peeled loop.
8842 @item max-completely-peel-times
8843 The maximum number of iterations of a loop to be suitable for complete peeling.
8845 @item max-completely-peel-loop-nest-depth
8846 The maximum depth of a loop nest suitable for complete peeling.
8848 @item max-unswitch-insns
8849 The maximum number of insns of an unswitched loop.
8851 @item max-unswitch-level
8852 The maximum number of branches unswitched in a single loop.
8855 The minimum cost of an expensive expression in the loop invariant motion.
8857 @item iv-consider-all-candidates-bound
8858 Bound on number of candidates for induction variables below that
8859 all candidates are considered for each use in induction variable
8860 optimizations. Only the most relevant candidates are considered
8861 if there are more candidates, to avoid quadratic time complexity.
8863 @item iv-max-considered-uses
8864 The induction variable optimizations give up on loops that contain more
8865 induction variable uses.
8867 @item iv-always-prune-cand-set-bound
8868 If number of candidates in the set is smaller than this value,
8869 we always try to remove unnecessary ivs from the set during its
8870 optimization when a new iv is added to the set.
8872 @item scev-max-expr-size
8873 Bound on size of expressions used in the scalar evolutions analyzer.
8874 Large expressions slow the analyzer.
8876 @item scev-max-expr-complexity
8877 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8878 Complex expressions slow the analyzer.
8880 @item omega-max-vars
8881 The maximum number of variables in an Omega constraint system.
8882 The default value is 128.
8884 @item omega-max-geqs
8885 The maximum number of inequalities in an Omega constraint system.
8886 The default value is 256.
8889 The maximum number of equalities in an Omega constraint system.
8890 The default value is 128.
8892 @item omega-max-wild-cards
8893 The maximum number of wildcard variables that the Omega solver is
8894 able to insert. The default value is 18.
8896 @item omega-hash-table-size
8897 The size of the hash table in the Omega solver. The default value is
8900 @item omega-max-keys
8901 The maximal number of keys used by the Omega solver. The default
8904 @item omega-eliminate-redundant-constraints
8905 When set to 1, use expensive methods to eliminate all redundant
8906 constraints. The default value is 0.
8908 @item vect-max-version-for-alignment-checks
8909 The maximum number of run-time checks that can be performed when
8910 doing loop versioning for alignment in the vectorizer. See option
8911 @option{-ftree-vect-loop-version} for more information.
8913 @item vect-max-version-for-alias-checks
8914 The maximum number of run-time checks that can be performed when
8915 doing loop versioning for alias in the vectorizer. See option
8916 @option{-ftree-vect-loop-version} for more information.
8918 @item max-iterations-to-track
8920 The maximum number of iterations of a loop the brute force algorithm
8921 for analysis of # of iterations of the loop tries to evaluate.
8923 @item hot-bb-count-fraction
8924 Select fraction of the maximal count of repetitions of basic block in program
8925 given basic block needs to have to be considered hot.
8927 @item hot-bb-frequency-fraction
8928 Select fraction of the entry block frequency of executions of basic block in
8929 function given basic block needs to have to be considered hot.
8931 @item max-predicted-iterations
8932 The maximum number of loop iterations we predict statically. This is useful
8933 in cases where a function contains a single loop with known bound and
8934 another loop with unknown bound.
8935 The known number of iterations is predicted correctly, while
8936 the unknown number of iterations average to roughly 10. This means that the
8937 loop without bounds appears artificially cold relative to the other one.
8939 @item align-threshold
8941 Select fraction of the maximal frequency of executions of a basic block in
8942 a function to align the basic block.
8944 @item align-loop-iterations
8946 A loop expected to iterate at least the selected number of iterations is
8949 @item tracer-dynamic-coverage
8950 @itemx tracer-dynamic-coverage-feedback
8952 This value is used to limit superblock formation once the given percentage of
8953 executed instructions is covered. This limits unnecessary code size
8956 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8957 feedback is available. The real profiles (as opposed to statically estimated
8958 ones) are much less balanced allowing the threshold to be larger value.
8960 @item tracer-max-code-growth
8961 Stop tail duplication once code growth has reached given percentage. This is
8962 a rather artificial limit, as most of the duplicates are eliminated later in
8963 cross jumping, so it may be set to much higher values than is the desired code
8966 @item tracer-min-branch-ratio
8968 Stop reverse growth when the reverse probability of best edge is less than this
8969 threshold (in percent).
8971 @item tracer-min-branch-ratio
8972 @itemx tracer-min-branch-ratio-feedback
8974 Stop forward growth if the best edge has probability lower than this
8977 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8978 compilation for profile feedback and one for compilation without. The value
8979 for compilation with profile feedback needs to be more conservative (higher) in
8980 order to make tracer effective.
8982 @item max-cse-path-length
8984 Maximum number of basic blocks on path that cse considers. The default is 10.
8987 The maximum instructions CSE process before flushing. The default is 1000.
8989 @item ggc-min-expand
8991 GCC uses a garbage collector to manage its own memory allocation. This
8992 parameter specifies the minimum percentage by which the garbage
8993 collector's heap should be allowed to expand between collections.
8994 Tuning this may improve compilation speed; it has no effect on code
8997 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8998 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
8999 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9000 GCC is not able to calculate RAM on a particular platform, the lower
9001 bound of 30% is used. Setting this parameter and
9002 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9003 every opportunity. This is extremely slow, but can be useful for
9006 @item ggc-min-heapsize
9008 Minimum size of the garbage collector's heap before it begins bothering
9009 to collect garbage. The first collection occurs after the heap expands
9010 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9011 tuning this may improve compilation speed, and has no effect on code
9014 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9015 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9016 with a lower bound of 4096 (four megabytes) and an upper bound of
9017 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9018 particular platform, the lower bound is used. Setting this parameter
9019 very large effectively disables garbage collection. Setting this
9020 parameter and @option{ggc-min-expand} to zero causes a full collection
9021 to occur at every opportunity.
9023 @item max-reload-search-insns
9024 The maximum number of instruction reload should look backward for equivalent
9025 register. Increasing values mean more aggressive optimization, making the
9026 compilation time increase with probably slightly better performance.
9027 The default value is 100.
9029 @item max-cselib-memory-locations
9030 The maximum number of memory locations cselib should take into account.
9031 Increasing values mean more aggressive optimization, making the compilation time
9032 increase with probably slightly better performance. The default value is 500.
9034 @item reorder-blocks-duplicate
9035 @itemx reorder-blocks-duplicate-feedback
9037 Used by basic block reordering pass to decide whether to use unconditional
9038 branch or duplicate the code on its destination. Code is duplicated when its
9039 estimated size is smaller than this value multiplied by the estimated size of
9040 unconditional jump in the hot spots of the program.
9042 The @option{reorder-block-duplicate-feedback} is used only when profile
9043 feedback is available and may be set to higher values than
9044 @option{reorder-block-duplicate} since information about the hot spots is more
9047 @item max-sched-ready-insns
9048 The maximum number of instructions ready to be issued the scheduler should
9049 consider at any given time during the first scheduling pass. Increasing
9050 values mean more thorough searches, making the compilation time increase
9051 with probably little benefit. The default value is 100.
9053 @item max-sched-region-blocks
9054 The maximum number of blocks in a region to be considered for
9055 interblock scheduling. The default value is 10.
9057 @item max-pipeline-region-blocks
9058 The maximum number of blocks in a region to be considered for
9059 pipelining in the selective scheduler. The default value is 15.
9061 @item max-sched-region-insns
9062 The maximum number of insns in a region to be considered for
9063 interblock scheduling. The default value is 100.
9065 @item max-pipeline-region-insns
9066 The maximum number of insns in a region to be considered for
9067 pipelining in the selective scheduler. The default value is 200.
9070 The minimum probability (in percents) of reaching a source block
9071 for interblock speculative scheduling. The default value is 40.
9073 @item max-sched-extend-regions-iters
9074 The maximum number of iterations through CFG to extend regions.
9075 A value of 0 (the default) disables region extensions.
9077 @item max-sched-insn-conflict-delay
9078 The maximum conflict delay for an insn to be considered for speculative motion.
9079 The default value is 3.
9081 @item sched-spec-prob-cutoff
9082 The minimal probability of speculation success (in percents), so that
9083 speculative insns are scheduled.
9084 The default value is 40.
9086 @item sched-mem-true-dep-cost
9087 Minimal distance (in CPU cycles) between store and load targeting same
9088 memory locations. The default value is 1.
9090 @item selsched-max-lookahead
9091 The maximum size of the lookahead window of selective scheduling. It is a
9092 depth of search for available instructions.
9093 The default value is 50.
9095 @item selsched-max-sched-times
9096 The maximum number of times that an instruction is scheduled during
9097 selective scheduling. This is the limit on the number of iterations
9098 through which the instruction may be pipelined. The default value is 2.
9100 @item selsched-max-insns-to-rename
9101 The maximum number of best instructions in the ready list that are considered
9102 for renaming in the selective scheduler. The default value is 2.
9105 The minimum value of stage count that swing modulo scheduler
9106 generates. The default value is 2.
9108 @item max-last-value-rtl
9109 The maximum size measured as number of RTLs that can be recorded in an expression
9110 in combiner for a pseudo register as last known value of that register. The default
9113 @item integer-share-limit
9114 Small integer constants can use a shared data structure, reducing the
9115 compiler's memory usage and increasing its speed. This sets the maximum
9116 value of a shared integer constant. The default value is 256.
9118 @item min-virtual-mappings
9119 Specifies the minimum number of virtual mappings in the incremental
9120 SSA updater that should be registered to trigger the virtual mappings
9121 heuristic defined by virtual-mappings-ratio. The default value is
9124 @item virtual-mappings-ratio
9125 If the number of virtual mappings is virtual-mappings-ratio bigger
9126 than the number of virtual symbols to be updated, then the incremental
9127 SSA updater switches to a full update for those symbols. The default
9130 @item ssp-buffer-size
9131 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
9132 protection when @option{-fstack-protection} is used.
9134 @item max-jump-thread-duplication-stmts
9135 Maximum number of statements allowed in a block that needs to be
9136 duplicated when threading jumps.
9138 @item max-fields-for-field-sensitive
9139 Maximum number of fields in a structure treated in
9140 a field sensitive manner during pointer analysis. The default is zero
9141 for @option{-O0} and @option{-O1},
9142 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
9144 @item prefetch-latency
9145 Estimate on average number of instructions that are executed before
9146 prefetch finishes. The distance we prefetch ahead is proportional
9147 to this constant. Increasing this number may also lead to less
9148 streams being prefetched (see @option{simultaneous-prefetches}).
9150 @item simultaneous-prefetches
9151 Maximum number of prefetches that can run at the same time.
9153 @item l1-cache-line-size
9154 The size of cache line in L1 cache, in bytes.
9157 The size of L1 cache, in kilobytes.
9160 The size of L2 cache, in kilobytes.
9162 @item min-insn-to-prefetch-ratio
9163 The minimum ratio between the number of instructions and the
9164 number of prefetches to enable prefetching in a loop.
9166 @item prefetch-min-insn-to-mem-ratio
9167 The minimum ratio between the number of instructions and the
9168 number of memory references to enable prefetching in a loop.
9170 @item use-canonical-types
9171 Whether the compiler should use the ``canonical'' type system. By
9172 default, this should always be 1, which uses a more efficient internal
9173 mechanism for comparing types in C++ and Objective-C++. However, if
9174 bugs in the canonical type system are causing compilation failures,
9175 set this value to 0 to disable canonical types.
9177 @item switch-conversion-max-branch-ratio
9178 Switch initialization conversion refuses to create arrays that are
9179 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9180 branches in the switch.
9182 @item max-partial-antic-length
9183 Maximum length of the partial antic set computed during the tree
9184 partial redundancy elimination optimization (@option{-ftree-pre}) when
9185 optimizing at @option{-O3} and above. For some sorts of source code
9186 the enhanced partial redundancy elimination optimization can run away,
9187 consuming all of the memory available on the host machine. This
9188 parameter sets a limit on the length of the sets that are computed,
9189 which prevents the runaway behavior. Setting a value of 0 for
9190 this parameter allows an unlimited set length.
9192 @item sccvn-max-scc-size
9193 Maximum size of a strongly connected component (SCC) during SCCVN
9194 processing. If this limit is hit, SCCVN processing for the whole
9195 function is not done and optimizations depending on it are
9196 disabled. The default maximum SCC size is 10000.
9198 @item ira-max-loops-num
9199 IRA uses regional register allocation by default. If a function
9200 contains more loops than the number given by this parameter, only at most
9201 the given number of the most frequently-executed loops form regions
9202 for regional register allocation. The default value of the
9205 @item ira-max-conflict-table-size
9206 Although IRA uses a sophisticated algorithm to compress the conflict
9207 table, the table can still require excessive amounts of memory for
9208 huge functions. If the conflict table for a function could be more
9209 than the size in MB given by this parameter, the register allocator
9210 instead uses a faster, simpler, and lower-quality
9211 algorithm that does not require building a pseudo-register conflict table.
9212 The default value of the parameter is 2000.
9214 @item ira-loop-reserved-regs
9215 IRA can be used to evaluate more accurate register pressure in loops
9216 for decisions to move loop invariants (see @option{-O3}). The number
9217 of available registers reserved for some other purposes is given
9218 by this parameter. The default value of the parameter is 2, which is
9219 the minimal number of registers needed by typical instructions.
9220 This value is the best found from numerous experiments.
9222 @item loop-invariant-max-bbs-in-loop
9223 Loop invariant motion can be very expensive, both in compilation time and
9224 in amount of needed compile-time memory, with very large loops. Loops
9225 with more basic blocks than this parameter won't have loop invariant
9226 motion optimization performed on them. The default value of the
9227 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
9229 @item loop-max-datarefs-for-datadeps
9230 Building data dapendencies is expensive for very large loops. This
9231 parameter limits the number of data references in loops that are
9232 considered for data dependence analysis. These large loops are no
9233 handled by the optimizations using loop data dependencies.
9234 The default value is 1000.
9236 @item max-vartrack-size
9237 Sets a maximum number of hash table slots to use during variable
9238 tracking dataflow analysis of any function. If this limit is exceeded
9239 with variable tracking at assignments enabled, analysis for that
9240 function is retried without it, after removing all debug insns from
9241 the function. If the limit is exceeded even without debug insns, var
9242 tracking analysis is completely disabled for the function. Setting
9243 the parameter to zero makes it unlimited.
9245 @item max-vartrack-expr-depth
9246 Sets a maximum number of recursion levels when attempting to map
9247 variable names or debug temporaries to value expressions. This trades
9248 compilation time for more complete debug information. If this is set too
9249 low, value expressions that are available and could be represented in
9250 debug information may end up not being used; setting this higher may
9251 enable the compiler to find more complex debug expressions, but compile
9252 time and memory use may grow. The default is 12.
9254 @item min-nondebug-insn-uid
9255 Use uids starting at this parameter for nondebug insns. The range below
9256 the parameter is reserved exclusively for debug insns created by
9257 @option{-fvar-tracking-assignments}, but debug insns may get
9258 (non-overlapping) uids above it if the reserved range is exhausted.
9260 @item ipa-sra-ptr-growth-factor
9261 IPA-SRA replaces a pointer to an aggregate with one or more new
9262 parameters only when their cumulative size is less or equal to
9263 @option{ipa-sra-ptr-growth-factor} times the size of the original
9266 @item tm-max-aggregate-size
9267 When making copies of thread-local variables in a transaction, this
9268 parameter specifies the size in bytes after which variables are
9269 saved with the logging functions as opposed to save/restore code
9270 sequence pairs. This option only applies when using
9273 @item graphite-max-nb-scop-params
9274 To avoid exponential effects in the Graphite loop transforms, the
9275 number of parameters in a Static Control Part (SCoP) is bounded. The
9276 default value is 10 parameters. A variable whose value is unknown at
9277 compilation time and defined outside a SCoP is a parameter of the SCoP.
9279 @item graphite-max-bbs-per-function
9280 To avoid exponential effects in the detection of SCoPs, the size of
9281 the functions analyzed by Graphite is bounded. The default value is
9284 @item loop-block-tile-size
9285 Loop blocking or strip mining transforms, enabled with
9286 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9287 loop in the loop nest by a given number of iterations. The strip
9288 length can be changed using the @option{loop-block-tile-size}
9289 parameter. The default value is 51 iterations.
9291 @item ipa-cp-value-list-size
9292 IPA-CP attempts to track all possible values and types passed to a function's
9293 parameter in order to propagate them and perform devirtualization.
9294 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9295 stores per one formal parameter of a function.
9297 @item lto-partitions
9298 Specify desired number of partitions produced during WHOPR compilation.
9299 The number of partitions should exceed the number of CPUs used for compilation.
9300 The default value is 32.
9302 @item lto-minpartition
9303 Size of minimal partition for WHOPR (in estimated instructions).
9304 This prevents expenses of splitting very small programs into too many
9307 @item cxx-max-namespaces-for-diagnostic-help
9308 The maximum number of namespaces to consult for suggestions when C++
9309 name lookup fails for an identifier. The default is 1000.
9311 @item sink-frequency-threshold
9312 The maximum relative execution frequency (in percents) of the target block
9313 relative to a statement's original block to allow statement sinking of a
9314 statement. Larger numbers result in more aggressive statement sinking.
9315 The default value is 75. A small positive adjustment is applied for
9316 statements with memory operands as those are even more profitable so sink.
9318 @item max-stores-to-sink
9319 The maximum number of conditional stores paires that can be sunk. Set to 0
9320 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9321 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9323 @item allow-load-data-races
9324 Allow optimizers to introduce new data races on loads.
9325 Set to 1 to allow, otherwise to 0. This option is enabled by default
9326 unless implicitly set by the @option{-fmemory-model=} option.
9328 @item allow-store-data-races
9329 Allow optimizers to introduce new data races on stores.
9330 Set to 1 to allow, otherwise to 0. This option is enabled by default
9331 unless implicitly set by the @option{-fmemory-model=} option.
9333 @item allow-packed-load-data-races
9334 Allow optimizers to introduce new data races on packed data loads.
9335 Set to 1 to allow, otherwise to 0. This option is enabled by default
9336 unless implicitly set by the @option{-fmemory-model=} option.
9338 @item allow-packed-store-data-races
9339 Allow optimizers to introduce new data races on packed data stores.
9340 Set to 1 to allow, otherwise to 0. This option is enabled by default
9341 unless implicitly set by the @option{-fmemory-model=} option.
9343 @item case-values-threshold
9344 The smallest number of different values for which it is best to use a
9345 jump-table instead of a tree of conditional branches. If the value is
9346 0, use the default for the machine. The default is 0.
9348 @item tree-reassoc-width
9349 Set the maximum number of instructions executed in parallel in
9350 reassociated tree. This parameter overrides target dependent
9351 heuristics used by default if has non zero value.
9353 @item sched-pressure-algorithm
9354 Choose between the two available implementations of
9355 @option{-fsched-pressure}. Algorithm 1 is the original implementation
9356 and is the more likely to prevent instructions from being reordered.
9357 Algorithm 2 was designed to be a compromise between the relatively
9358 conservative approach taken by algorithm 1 and the rather aggressive
9359 approach taken by the default scheduler. It relies more heavily on
9360 having a regular register file and accurate register pressure classes.
9361 See @file{haifa-sched.c} in the GCC sources for more details.
9363 The default choice depends on the target.
9367 @node Preprocessor Options
9368 @section Options Controlling the Preprocessor
9369 @cindex preprocessor options
9370 @cindex options, preprocessor
9372 These options control the C preprocessor, which is run on each C source
9373 file before actual compilation.
9375 If you use the @option{-E} option, nothing is done except preprocessing.
9376 Some of these options make sense only together with @option{-E} because
9377 they cause the preprocessor output to be unsuitable for actual
9381 @item -Wp,@var{option}
9383 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9384 and pass @var{option} directly through to the preprocessor. If
9385 @var{option} contains commas, it is split into multiple options at the
9386 commas. However, many options are modified, translated or interpreted
9387 by the compiler driver before being passed to the preprocessor, and
9388 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9389 interface is undocumented and subject to change, so whenever possible
9390 you should avoid using @option{-Wp} and let the driver handle the
9393 @item -Xpreprocessor @var{option}
9394 @opindex Xpreprocessor
9395 Pass @var{option} as an option to the preprocessor. You can use this to
9396 supply system-specific preprocessor options that GCC does not know how to
9399 If you want to pass an option that takes an argument, you must use
9400 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9402 @item -no-integrated-cpp
9403 @opindex no-integrated-cpp
9404 Perform preprocessing as a separate pass before compilation.
9405 By default, GCC performs preprocessing as an integrated part of
9406 input tokenization and parsing.
9407 If this option is provided, the appropriate language front end
9408 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
9409 and Objective-C, respectively) is instead invoked twice,
9410 once for preprocessing only and once for actual compilation
9411 of the preprocessed input.
9412 This option may be useful in conjunction with the @option{-B} or
9413 @option{-wrapper} options to specify an alternate preprocessor or
9414 perform additional processing of the program source between
9415 normal preprocessing and compilation.
9418 @include cppopts.texi
9420 @node Assembler Options
9421 @section Passing Options to the Assembler
9423 @c prevent bad page break with this line
9424 You can pass options to the assembler.
9427 @item -Wa,@var{option}
9429 Pass @var{option} as an option to the assembler. If @var{option}
9430 contains commas, it is split into multiple options at the commas.
9432 @item -Xassembler @var{option}
9434 Pass @var{option} as an option to the assembler. You can use this to
9435 supply system-specific assembler options that GCC does not know how to
9438 If you want to pass an option that takes an argument, you must use
9439 @option{-Xassembler} twice, once for the option and once for the argument.
9444 @section Options for Linking
9445 @cindex link options
9446 @cindex options, linking
9448 These options come into play when the compiler links object files into
9449 an executable output file. They are meaningless if the compiler is
9450 not doing a link step.
9454 @item @var{object-file-name}
9455 A file name that does not end in a special recognized suffix is
9456 considered to name an object file or library. (Object files are
9457 distinguished from libraries by the linker according to the file
9458 contents.) If linking is done, these object files are used as input
9467 If any of these options is used, then the linker is not run, and
9468 object file names should not be used as arguments. @xref{Overall
9472 @item -l@var{library}
9473 @itemx -l @var{library}
9475 Search the library named @var{library} when linking. (The second
9476 alternative with the library as a separate argument is only for
9477 POSIX compliance and is not recommended.)
9479 It makes a difference where in the command you write this option; the
9480 linker searches and processes libraries and object files in the order they
9481 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9482 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9483 to functions in @samp{z}, those functions may not be loaded.
9485 The linker searches a standard list of directories for the library,
9486 which is actually a file named @file{lib@var{library}.a}. The linker
9487 then uses this file as if it had been specified precisely by name.
9489 The directories searched include several standard system directories
9490 plus any that you specify with @option{-L}.
9492 Normally the files found this way are library files---archive files
9493 whose members are object files. The linker handles an archive file by
9494 scanning through it for members which define symbols that have so far
9495 been referenced but not defined. But if the file that is found is an
9496 ordinary object file, it is linked in the usual fashion. The only
9497 difference between using an @option{-l} option and specifying a file name
9498 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9499 and searches several directories.
9503 You need this special case of the @option{-l} option in order to
9504 link an Objective-C or Objective-C++ program.
9507 @opindex nostartfiles
9508 Do not use the standard system startup files when linking.
9509 The standard system libraries are used normally, unless @option{-nostdlib}
9510 or @option{-nodefaultlibs} is used.
9512 @item -nodefaultlibs
9513 @opindex nodefaultlibs
9514 Do not use the standard system libraries when linking.
9515 Only the libraries you specify are passed to the linker, and options
9516 specifying linkage of the system libraries, such as @code{-static-libgcc}
9517 or @code{-shared-libgcc}, are ignored.
9518 The standard startup files are used normally, unless @option{-nostartfiles}
9519 is used. The compiler may generate calls to @code{memcmp},
9520 @code{memset}, @code{memcpy} and @code{memmove}.
9521 These entries are usually resolved by entries in
9522 libc. These entry points should be supplied through some other
9523 mechanism when this option is specified.
9527 Do not use the standard system startup files or libraries when linking.
9528 No startup files and only the libraries you specify are passed to
9529 the linker, and options specifying linkage of the system libraries, such as
9530 @code{-static-libgcc} or @code{-shared-libgcc}, are ignored.
9532 The compiler may generate calls to @code{memcmp}, @code{memset},
9533 @code{memcpy} and @code{memmove}.
9534 These entries are usually resolved by entries in
9535 libc. These entry points should be supplied through some other
9536 mechanism when this option is specified.
9538 @cindex @option{-lgcc}, use with @option{-nostdlib}
9539 @cindex @option{-nostdlib} and unresolved references
9540 @cindex unresolved references and @option{-nostdlib}
9541 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9542 @cindex @option{-nodefaultlibs} and unresolved references
9543 @cindex unresolved references and @option{-nodefaultlibs}
9544 One of the standard libraries bypassed by @option{-nostdlib} and
9545 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9546 which GCC uses to overcome shortcomings of particular machines, or special
9547 needs for some languages.
9548 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9549 Collection (GCC) Internals},
9550 for more discussion of @file{libgcc.a}.)
9551 In most cases, you need @file{libgcc.a} even when you want to avoid
9552 other standard libraries. In other words, when you specify @option{-nostdlib}
9553 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9554 This ensures that you have no unresolved references to internal GCC
9555 library subroutines.
9556 (An example of such an internal subroutine is @samp{__main}, used to ensure C++
9557 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
9558 GNU Compiler Collection (GCC) Internals}.)
9562 Produce a position independent executable on targets that support it.
9563 For predictable results, you must also specify the same set of options
9564 used for compilation (@option{-fpie}, @option{-fPIE},
9565 or model suboptions) when you specify this linker option.
9569 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9570 that support it. This instructs the linker to add all symbols, not
9571 only used ones, to the dynamic symbol table. This option is needed
9572 for some uses of @code{dlopen} or to allow obtaining backtraces
9573 from within a program.
9577 Remove all symbol table and relocation information from the executable.
9581 On systems that support dynamic linking, this prevents linking with the shared
9582 libraries. On other systems, this option has no effect.
9586 Produce a shared object which can then be linked with other objects to
9587 form an executable. Not all systems support this option. For predictable
9588 results, you must also specify the same set of options used for compilation
9589 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
9590 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
9591 needs to build supplementary stub code for constructors to work. On
9592 multi-libbed systems, @samp{gcc -shared} must select the correct support
9593 libraries to link against. Failing to supply the correct flags may lead
9594 to subtle defects. Supplying them in cases where they are not necessary
9597 @item -shared-libgcc
9598 @itemx -static-libgcc
9599 @opindex shared-libgcc
9600 @opindex static-libgcc
9601 On systems that provide @file{libgcc} as a shared library, these options
9602 force the use of either the shared or static version respectively.
9603 If no shared version of @file{libgcc} was built when the compiler was
9604 configured, these options have no effect.
9606 There are several situations in which an application should use the
9607 shared @file{libgcc} instead of the static version. The most common
9608 of these is when the application wishes to throw and catch exceptions
9609 across different shared libraries. In that case, each of the libraries
9610 as well as the application itself should use the shared @file{libgcc}.
9612 Therefore, the G++ and GCJ drivers automatically add
9613 @option{-shared-libgcc} whenever you build a shared library or a main
9614 executable, because C++ and Java programs typically use exceptions, so
9615 this is the right thing to do.
9617 If, instead, you use the GCC driver to create shared libraries, you may
9618 find that they are not always linked with the shared @file{libgcc}.
9619 If GCC finds, at its configuration time, that you have a non-GNU linker
9620 or a GNU linker that does not support option @option{--eh-frame-hdr},
9621 it links the shared version of @file{libgcc} into shared libraries
9622 by default. Otherwise, it takes advantage of the linker and optimizes
9623 away the linking with the shared version of @file{libgcc}, linking with
9624 the static version of libgcc by default. This allows exceptions to
9625 propagate through such shared libraries, without incurring relocation
9626 costs at library load time.
9628 However, if a library or main executable is supposed to throw or catch
9629 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9630 for the languages used in the program, or using the option
9631 @option{-shared-libgcc}, such that it is linked with the shared
9634 @item -static-libstdc++
9635 When the @command{g++} program is used to link a C++ program, it
9636 normally automatically links against @option{libstdc++}. If
9637 @file{libstdc++} is available as a shared library, and the
9638 @option{-static} option is not used, then this links against the
9639 shared version of @file{libstdc++}. That is normally fine. However, it
9640 is sometimes useful to freeze the version of @file{libstdc++} used by
9641 the program without going all the way to a fully static link. The
9642 @option{-static-libstdc++} option directs the @command{g++} driver to
9643 link @file{libstdc++} statically, without necessarily linking other
9644 libraries statically.
9648 Bind references to global symbols when building a shared object. Warn
9649 about any unresolved references (unless overridden by the link editor
9650 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
9653 @item -T @var{script}
9655 @cindex linker script
9656 Use @var{script} as the linker script. This option is supported by most
9657 systems using the GNU linker. On some targets, such as bare-board
9658 targets without an operating system, the @option{-T} option may be required
9659 when linking to avoid references to undefined symbols.
9661 @item -Xlinker @var{option}
9663 Pass @var{option} as an option to the linker. You can use this to
9664 supply system-specific linker options that GCC does not recognize.
9666 If you want to pass an option that takes a separate argument, you must use
9667 @option{-Xlinker} twice, once for the option and once for the argument.
9668 For example, to pass @option{-assert definitions}, you must write
9669 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
9670 @option{-Xlinker "-assert definitions"}, because this passes the entire
9671 string as a single argument, which is not what the linker expects.
9673 When using the GNU linker, it is usually more convenient to pass
9674 arguments to linker options using the @option{@var{option}=@var{value}}
9675 syntax than as separate arguments. For example, you can specify
9676 @option{-Xlinker -Map=output.map} rather than
9677 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9678 this syntax for command-line options.
9680 @item -Wl,@var{option}
9682 Pass @var{option} as an option to the linker. If @var{option} contains
9683 commas, it is split into multiple options at the commas. You can use this
9684 syntax to pass an argument to the option.
9685 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
9686 linker. When using the GNU linker, you can also get the same effect with
9687 @option{-Wl,-Map=output.map}.
9689 @item -u @var{symbol}
9691 Pretend the symbol @var{symbol} is undefined, to force linking of
9692 library modules to define it. You can use @option{-u} multiple times with
9693 different symbols to force loading of additional library modules.
9696 @node Directory Options
9697 @section Options for Directory Search
9698 @cindex directory options
9699 @cindex options, directory search
9702 These options specify directories to search for header files, for
9703 libraries and for parts of the compiler:
9708 Add the directory @var{dir} to the head of the list of directories to be
9709 searched for header files. This can be used to override a system header
9710 file, substituting your own version, since these directories are
9711 searched before the system header file directories. However, you should
9712 not use this option to add directories that contain vendor-supplied
9713 system header files (use @option{-isystem} for that). If you use more than
9714 one @option{-I} option, the directories are scanned in left-to-right
9715 order; the standard system directories come after.
9717 If a standard system include directory, or a directory specified with
9718 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9719 option is ignored. The directory is still searched but as a
9720 system directory at its normal position in the system include chain.
9721 This is to ensure that GCC's procedure to fix buggy system headers and
9722 the ordering for the @code{include_next} directive are not inadvertently changed.
9723 If you really need to change the search order for system directories,
9724 use the @option{-nostdinc} and/or @option{-isystem} options.
9726 @item -iplugindir=@var{dir}
9727 Set the directory to search for plugins that are passed
9728 by @option{-fplugin=@var{name}} instead of
9729 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9730 to be used by the user, but only passed by the driver.
9732 @item -iquote@var{dir}
9734 Add the directory @var{dir} to the head of the list of directories to
9735 be searched for header files only for the case of @samp{#include
9736 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9737 otherwise just like @option{-I}.
9741 Add directory @var{dir} to the list of directories to be searched
9744 @item -B@var{prefix}
9746 This option specifies where to find the executables, libraries,
9747 include files, and data files of the compiler itself.
9749 The compiler driver program runs one or more of the subprograms
9750 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
9751 @var{prefix} as a prefix for each program it tries to run, both with and
9752 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9754 For each subprogram to be run, the compiler driver first tries the
9755 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9756 is not specified, the driver tries two standard prefixes,
9757 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9758 those results in a file name that is found, the unmodified program
9759 name is searched for using the directories specified in your
9760 @env{PATH} environment variable.
9762 The compiler checks to see if the path provided by the @option{-B}
9763 refers to a directory, and if necessary it adds a directory
9764 separator character at the end of the path.
9766 @option{-B} prefixes that effectively specify directory names also apply
9767 to libraries in the linker, because the compiler translates these
9768 options into @option{-L} options for the linker. They also apply to
9769 includes files in the preprocessor, because the compiler translates these
9770 options into @option{-isystem} options for the preprocessor. In this case,
9771 the compiler appends @samp{include} to the prefix.
9773 The runtime support file @file{libgcc.a} can also be searched for using
9774 the @option{-B} prefix, if needed. If it is not found there, the two
9775 standard prefixes above are tried, and that is all. The file is left
9776 out of the link if it is not found by those means.
9778 Another way to specify a prefix much like the @option{-B} prefix is to use
9779 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9782 As a special kludge, if the path provided by @option{-B} is
9783 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9784 9, then it is replaced by @file{[dir/]include}. This is to help
9785 with boot-strapping the compiler.
9787 @item -specs=@var{file}
9789 Process @var{file} after the compiler reads in the standard @file{specs}
9790 file, in order to override the defaults which the @command{gcc} driver
9791 program uses when determining what switches to pass to @command{cc1},
9792 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
9793 @option{-specs=@var{file}} can be specified on the command line, and they
9794 are processed in order, from left to right.
9796 @item --sysroot=@var{dir}
9798 Use @var{dir} as the logical root directory for headers and libraries.
9799 For example, if the compiler normally searches for headers in
9800 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
9801 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9803 If you use both this option and the @option{-isysroot} option, then
9804 the @option{--sysroot} option applies to libraries, but the
9805 @option{-isysroot} option applies to header files.
9807 The GNU linker (beginning with version 2.16) has the necessary support
9808 for this option. If your linker does not support this option, the
9809 header file aspect of @option{--sysroot} still works, but the
9810 library aspect does not.
9814 This option has been deprecated. Please use @option{-iquote} instead for
9815 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9816 Any directories you specify with @option{-I} options before the @option{-I-}
9817 option are searched only for the case of @samp{#include "@var{file}"};
9818 they are not searched for @samp{#include <@var{file}>}.
9820 If additional directories are specified with @option{-I} options after
9821 the @option{-I-}, these directories are searched for all @samp{#include}
9822 directives. (Ordinarily @emph{all} @option{-I} directories are used
9825 In addition, the @option{-I-} option inhibits the use of the current
9826 directory (where the current input file came from) as the first search
9827 directory for @samp{#include "@var{file}"}. There is no way to
9828 override this effect of @option{-I-}. With @option{-I.} you can specify
9829 searching the directory that is current when the compiler is
9830 invoked. That is not exactly the same as what the preprocessor does
9831 by default, but it is often satisfactory.
9833 @option{-I-} does not inhibit the use of the standard system directories
9834 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9841 @section Specifying subprocesses and the switches to pass to them
9844 @command{gcc} is a driver program. It performs its job by invoking a
9845 sequence of other programs to do the work of compiling, assembling and
9846 linking. GCC interprets its command-line parameters and uses these to
9847 deduce which programs it should invoke, and which command-line options
9848 it ought to place on their command lines. This behavior is controlled
9849 by @dfn{spec strings}. In most cases there is one spec string for each
9850 program that GCC can invoke, but a few programs have multiple spec
9851 strings to control their behavior. The spec strings built into GCC can
9852 be overridden by using the @option{-specs=} command-line switch to specify
9855 @dfn{Spec files} are plaintext files that are used to construct spec
9856 strings. They consist of a sequence of directives separated by blank
9857 lines. The type of directive is determined by the first non-whitespace
9858 character on the line, which can be one of the following:
9861 @item %@var{command}
9862 Issues a @var{command} to the spec file processor. The commands that can
9866 @item %include <@var{file}>
9867 @cindex @code{%include}
9868 Search for @var{file} and insert its text at the current point in the
9871 @item %include_noerr <@var{file}>
9872 @cindex @code{%include_noerr}
9873 Just like @samp{%include}, but do not generate an error message if the include
9874 file cannot be found.
9876 @item %rename @var{old_name} @var{new_name}
9877 @cindex @code{%rename}
9878 Rename the spec string @var{old_name} to @var{new_name}.
9882 @item *[@var{spec_name}]:
9883 This tells the compiler to create, override or delete the named spec
9884 string. All lines after this directive up to the next directive or
9885 blank line are considered to be the text for the spec string. If this
9886 results in an empty string then the spec is deleted. (Or, if the
9887 spec did not exist, then nothing happens.) Otherwise, if the spec
9888 does not currently exist a new spec is created. If the spec does
9889 exist then its contents are overridden by the text of this
9890 directive, unless the first character of that text is the @samp{+}
9891 character, in which case the text is appended to the spec.
9893 @item [@var{suffix}]:
9894 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9895 and up to the next directive or blank line are considered to make up the
9896 spec string for the indicated suffix. When the compiler encounters an
9897 input file with the named suffix, it processes the spec string in
9898 order to work out how to compile that file. For example:
9905 This says that any input file whose name ends in @samp{.ZZ} should be
9906 passed to the program @samp{z-compile}, which should be invoked with the
9907 command-line switch @option{-input} and with the result of performing the
9908 @samp{%i} substitution. (See below.)
9910 As an alternative to providing a spec string, the text that follows a
9911 suffix directive can be one of the following:
9914 @item @@@var{language}
9915 This says that the suffix is an alias for a known @var{language}. This is
9916 similar to using the @option{-x} command-line switch to GCC to specify a
9917 language explicitly. For example:
9924 Says that .ZZ files are, in fact, C++ source files.
9927 This causes an error messages saying:
9930 @var{name} compiler not installed on this system.
9934 GCC already has an extensive list of suffixes built into it.
9935 This directive adds an entry to the end of the list of suffixes, but
9936 since the list is searched from the end backwards, it is effectively
9937 possible to override earlier entries using this technique.
9941 GCC has the following spec strings built into it. Spec files can
9942 override these strings or create their own. Note that individual
9943 targets can also add their own spec strings to this list.
9946 asm Options to pass to the assembler
9947 asm_final Options to pass to the assembler post-processor
9948 cpp Options to pass to the C preprocessor
9949 cc1 Options to pass to the C compiler
9950 cc1plus Options to pass to the C++ compiler
9951 endfile Object files to include at the end of the link
9952 link Options to pass to the linker
9953 lib Libraries to include on the command line to the linker
9954 libgcc Decides which GCC support library to pass to the linker
9955 linker Sets the name of the linker
9956 predefines Defines to be passed to the C preprocessor
9957 signed_char Defines to pass to CPP to say whether @code{char} is signed
9959 startfile Object files to include at the start of the link
9962 Here is a small example of a spec file:
9968 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9971 This example renames the spec called @samp{lib} to @samp{old_lib} and
9972 then overrides the previous definition of @samp{lib} with a new one.
9973 The new definition adds in some extra command-line options before
9974 including the text of the old definition.
9976 @dfn{Spec strings} are a list of command-line options to be passed to their
9977 corresponding program. In addition, the spec strings can contain
9978 @samp{%}-prefixed sequences to substitute variable text or to
9979 conditionally insert text into the command line. Using these constructs
9980 it is possible to generate quite complex command lines.
9982 Here is a table of all defined @samp{%}-sequences for spec
9983 strings. Note that spaces are not generated automatically around the
9984 results of expanding these sequences. Therefore you can concatenate them
9985 together or combine them with constant text in a single argument.
9989 Substitute one @samp{%} into the program name or argument.
9992 Substitute the name of the input file being processed.
9995 Substitute the basename of the input file being processed.
9996 This is the substring up to (and not including) the last period
9997 and not including the directory.
10000 This is the same as @samp{%b}, but include the file suffix (text after
10004 Marks the argument containing or following the @samp{%d} as a
10005 temporary file name, so that that file is deleted if GCC exits
10006 successfully. Unlike @samp{%g}, this contributes no text to the
10009 @item %g@var{suffix}
10010 Substitute a file name that has suffix @var{suffix} and is chosen
10011 once per compilation, and mark the argument in the same way as
10012 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
10013 name is now chosen in a way that is hard to predict even when previously
10014 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
10015 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
10016 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
10017 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
10018 was simply substituted with a file name chosen once per compilation,
10019 without regard to any appended suffix (which was therefore treated
10020 just like ordinary text), making such attacks more likely to succeed.
10022 @item %u@var{suffix}
10023 Like @samp{%g}, but generates a new temporary file name
10024 each time it appears instead of once per compilation.
10026 @item %U@var{suffix}
10027 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
10028 new one if there is no such last file name. In the absence of any
10029 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
10030 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
10031 involves the generation of two distinct file names, one
10032 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
10033 simply substituted with a file name chosen for the previous @samp{%u},
10034 without regard to any appended suffix.
10036 @item %j@var{suffix}
10037 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
10038 writable, and if @option{-save-temps} is not used;
10039 otherwise, substitute the name
10040 of a temporary file, just like @samp{%u}. This temporary file is not
10041 meant for communication between processes, but rather as a junk
10042 disposal mechanism.
10044 @item %|@var{suffix}
10045 @itemx %m@var{suffix}
10046 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
10047 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
10048 all. These are the two most common ways to instruct a program that it
10049 should read from standard input or write to standard output. If you
10050 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
10051 construct: see for example @file{f/lang-specs.h}.
10053 @item %.@var{SUFFIX}
10054 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
10055 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
10056 terminated by the next space or %.
10059 Marks the argument containing or following the @samp{%w} as the
10060 designated output file of this compilation. This puts the argument
10061 into the sequence of arguments that @samp{%o} substitutes.
10064 Substitutes the names of all the output files, with spaces
10065 automatically placed around them. You should write spaces
10066 around the @samp{%o} as well or the results are undefined.
10067 @samp{%o} is for use in the specs for running the linker.
10068 Input files whose names have no recognized suffix are not compiled
10069 at all, but they are included among the output files, so they are
10073 Substitutes the suffix for object files. Note that this is
10074 handled specially when it immediately follows @samp{%g, %u, or %U},
10075 because of the need for those to form complete file names. The
10076 handling is such that @samp{%O} is treated exactly as if it had already
10077 been substituted, except that @samp{%g, %u, and %U} do not currently
10078 support additional @var{suffix} characters following @samp{%O} as they do
10079 following, for example, @samp{.o}.
10082 Substitutes the standard macro predefinitions for the
10083 current target machine. Use this when running @code{cpp}.
10086 Like @samp{%p}, but puts @samp{__} before and after the name of each
10087 predefined macro, except for macros that start with @samp{__} or with
10088 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
10092 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
10093 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
10094 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
10095 and @option{-imultilib} as necessary.
10098 Current argument is the name of a library or startup file of some sort.
10099 Search for that file in a standard list of directories and substitute
10100 the full name found. The current working directory is included in the
10101 list of directories scanned.
10104 Current argument is the name of a linker script. Search for that file
10105 in the current list of directories to scan for libraries. If the file
10106 is located insert a @option{--script} option into the command line
10107 followed by the full path name found. If the file is not found then
10108 generate an error message. Note: the current working directory is not
10112 Print @var{str} as an error message. @var{str} is terminated by a newline.
10113 Use this when inconsistent options are detected.
10115 @item %(@var{name})
10116 Substitute the contents of spec string @var{name} at this point.
10118 @item %x@{@var{option}@}
10119 Accumulate an option for @samp{%X}.
10122 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
10126 Output the accumulated assembler options specified by @option{-Wa}.
10129 Output the accumulated preprocessor options specified by @option{-Wp}.
10132 Process the @code{asm} spec. This is used to compute the
10133 switches to be passed to the assembler.
10136 Process the @code{asm_final} spec. This is a spec string for
10137 passing switches to an assembler post-processor, if such a program is
10141 Process the @code{link} spec. This is the spec for computing the
10142 command line passed to the linker. Typically it makes use of the
10143 @samp{%L %G %S %D and %E} sequences.
10146 Dump out a @option{-L} option for each directory that GCC believes might
10147 contain startup files. If the target supports multilibs then the
10148 current multilib directory is prepended to each of these paths.
10151 Process the @code{lib} spec. This is a spec string for deciding which
10152 libraries are included on the command line to the linker.
10155 Process the @code{libgcc} spec. This is a spec string for deciding
10156 which GCC support library is included on the command line to the linker.
10159 Process the @code{startfile} spec. This is a spec for deciding which
10160 object files are the first ones passed to the linker. Typically
10161 this might be a file named @file{crt0.o}.
10164 Process the @code{endfile} spec. This is a spec string that specifies
10165 the last object files that are passed to the linker.
10168 Process the @code{cpp} spec. This is used to construct the arguments
10169 to be passed to the C preprocessor.
10172 Process the @code{cc1} spec. This is used to construct the options to be
10173 passed to the actual C compiler (@samp{cc1}).
10176 Process the @code{cc1plus} spec. This is used to construct the options to be
10177 passed to the actual C++ compiler (@samp{cc1plus}).
10180 Substitute the variable part of a matched option. See below.
10181 Note that each comma in the substituted string is replaced by
10185 Remove all occurrences of @code{-S} from the command line. Note---this
10186 command is position dependent. @samp{%} commands in the spec string
10187 before this one see @code{-S}, @samp{%} commands in the spec string
10188 after this one do not.
10190 @item %:@var{function}(@var{args})
10191 Call the named function @var{function}, passing it @var{args}.
10192 @var{args} is first processed as a nested spec string, then split
10193 into an argument vector in the usual fashion. The function returns
10194 a string which is processed as if it had appeared literally as part
10195 of the current spec.
10197 The following built-in spec functions are provided:
10200 @item @code{getenv}
10201 The @code{getenv} spec function takes two arguments: an environment
10202 variable name and a string. If the environment variable is not
10203 defined, a fatal error is issued. Otherwise, the return value is the
10204 value of the environment variable concatenated with the string. For
10205 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
10208 %:getenv(TOPDIR /include)
10211 expands to @file{/path/to/top/include}.
10213 @item @code{if-exists}
10214 The @code{if-exists} spec function takes one argument, an absolute
10215 pathname to a file. If the file exists, @code{if-exists} returns the
10216 pathname. Here is a small example of its usage:
10220 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
10223 @item @code{if-exists-else}
10224 The @code{if-exists-else} spec function is similar to the @code{if-exists}
10225 spec function, except that it takes two arguments. The first argument is
10226 an absolute pathname to a file. If the file exists, @code{if-exists-else}
10227 returns the pathname. If it does not exist, it returns the second argument.
10228 This way, @code{if-exists-else} can be used to select one file or another,
10229 based on the existence of the first. Here is a small example of its usage:
10233 crt0%O%s %:if-exists(crti%O%s) \
10234 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
10237 @item @code{replace-outfile}
10238 The @code{replace-outfile} spec function takes two arguments. It looks for the
10239 first argument in the outfiles array and replaces it with the second argument. Here
10240 is a small example of its usage:
10243 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10246 @item @code{remove-outfile}
10247 The @code{remove-outfile} spec function takes one argument. It looks for the
10248 first argument in the outfiles array and removes it. Here is a small example
10252 %:remove-outfile(-lm)
10255 @item @code{pass-through-libs}
10256 The @code{pass-through-libs} spec function takes any number of arguments. It
10257 finds any @option{-l} options and any non-options ending in @file{.a} (which it
10258 assumes are the names of linker input library archive files) and returns a
10259 result containing all the found arguments each prepended by
10260 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10261 intended to be passed to the LTO linker plugin.
10264 %:pass-through-libs(%G %L %G)
10267 @item @code{print-asm-header}
10268 The @code{print-asm-header} function takes no arguments and simply
10269 prints a banner like:
10275 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10278 It is used to separate compiler options from assembler options
10279 in the @option{--target-help} output.
10282 @item %@{@code{S}@}
10283 Substitutes the @code{-S} switch, if that switch is given to GCC@.
10284 If that switch is not specified, this substitutes nothing. Note that
10285 the leading dash is omitted when specifying this option, and it is
10286 automatically inserted if the substitution is performed. Thus the spec
10287 string @samp{%@{foo@}} matches the command-line option @option{-foo}
10288 and outputs the command-line option @option{-foo}.
10290 @item %W@{@code{S}@}
10291 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10292 deleted on failure.
10294 @item %@{@code{S}*@}
10295 Substitutes all the switches specified to GCC whose names start
10296 with @code{-S}, but which also take an argument. This is used for
10297 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10298 GCC considers @option{-o foo} as being
10299 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
10300 text, including the space. Thus two arguments are generated.
10302 @item %@{@code{S}*&@code{T}*@}
10303 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10304 (the order of @code{S} and @code{T} in the spec is not significant).
10305 There can be any number of ampersand-separated variables; for each the
10306 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10308 @item %@{@code{S}:@code{X}@}
10309 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
10311 @item %@{!@code{S}:@code{X}@}
10312 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
10314 @item %@{@code{S}*:@code{X}@}
10315 Substitutes @code{X} if one or more switches whose names start with
10316 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10317 once, no matter how many such switches appeared. However, if @code{%*}
10318 appears somewhere in @code{X}, then @code{X} is substituted once
10319 for each matching switch, with the @code{%*} replaced by the part of
10320 that switch matching the @code{*}.
10322 @item %@{.@code{S}:@code{X}@}
10323 Substitutes @code{X}, if processing a file with suffix @code{S}.
10325 @item %@{!.@code{S}:@code{X}@}
10326 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10328 @item %@{,@code{S}:@code{X}@}
10329 Substitutes @code{X}, if processing a file for language @code{S}.
10331 @item %@{!,@code{S}:@code{X}@}
10332 Substitutes @code{X}, if not processing a file for language @code{S}.
10334 @item %@{@code{S}|@code{P}:@code{X}@}
10335 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
10336 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10337 @code{*} sequences as well, although they have a stronger binding than
10338 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10339 alternatives must be starred, and only the first matching alternative
10342 For example, a spec string like this:
10345 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10349 outputs the following command-line options from the following input
10350 command-line options:
10355 -d fred.c -foo -baz -boggle
10356 -d jim.d -bar -baz -boggle
10359 @item %@{S:X; T:Y; :D@}
10361 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
10362 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10363 be as many clauses as you need. This may be combined with @code{.},
10364 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10369 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10370 construct may contain other nested @samp{%} constructs or spaces, or
10371 even newlines. They are processed as usual, as described above.
10372 Trailing white space in @code{X} is ignored. White space may also
10373 appear anywhere on the left side of the colon in these constructs,
10374 except between @code{.} or @code{*} and the corresponding word.
10376 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10377 handled specifically in these constructs. If another value of
10378 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10379 @option{-W} switch is found later in the command line, the earlier
10380 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10381 just one letter, which passes all matching options.
10383 The character @samp{|} at the beginning of the predicate text is used to
10384 indicate that a command should be piped to the following command, but
10385 only if @option{-pipe} is specified.
10387 It is built into GCC which switches take arguments and which do not.
10388 (You might think it would be useful to generalize this to allow each
10389 compiler's spec to say which switches take arguments. But this cannot
10390 be done in a consistent fashion. GCC cannot even decide which input
10391 files have been specified without knowing which switches take arguments,
10392 and it must know which input files to compile in order to tell which
10395 GCC also knows implicitly that arguments starting in @option{-l} are to be
10396 treated as compiler output files, and passed to the linker in their
10397 proper position among the other output files.
10399 @c man begin OPTIONS
10401 @node Target Options
10402 @section Specifying Target Machine and Compiler Version
10403 @cindex target options
10404 @cindex cross compiling
10405 @cindex specifying machine version
10406 @cindex specifying compiler version and target machine
10407 @cindex compiler version, specifying
10408 @cindex target machine, specifying
10410 The usual way to run GCC is to run the executable called @command{gcc}, or
10411 @command{@var{machine}-gcc} when cross-compiling, or
10412 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10413 one that was installed last.
10415 @node Submodel Options
10416 @section Hardware Models and Configurations
10417 @cindex submodel options
10418 @cindex specifying hardware config
10419 @cindex hardware models and configurations, specifying
10420 @cindex machine dependent options
10422 Each target machine types can have its own
10423 special options, starting with @samp{-m}, to choose among various
10424 hardware models or configurations---for example, 68010 vs 68020,
10425 floating coprocessor or none. A single installed version of the
10426 compiler can compile for any model or configuration, according to the
10429 Some configurations of the compiler also support additional special
10430 options, usually for compatibility with other compilers on the same
10433 @c This list is ordered alphanumerically by subsection name.
10434 @c It should be the same order and spelling as these options are listed
10435 @c in Machine Dependent Options
10438 * Adapteva Epiphany Options::
10441 * Blackfin Options::
10446 * DEC Alpha Options::
10449 * GNU/Linux Options::
10452 * i386 and x86-64 Options::
10453 * i386 and x86-64 Windows Options::
10461 * MicroBlaze Options::
10464 * MN10300 Options::
10466 * picoChip Options::
10467 * PowerPC Options::
10469 * RS/6000 and PowerPC Options::
10471 * S/390 and zSeries Options::
10474 * Solaris 2 Options::
10477 * System V Options::
10478 * TILE-Gx Options::
10479 * TILEPro Options::
10483 * VxWorks Options::
10485 * Xstormy16 Options::
10487 * zSeries Options::
10490 @node Adapteva Epiphany Options
10491 @subsection Adapteva Epiphany Options
10493 These @samp{-m} options are defined for Adapteva Epiphany:
10496 @item -mhalf-reg-file
10497 @opindex mhalf-reg-file
10498 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
10499 That allows code to run on hardware variants that lack these registers.
10501 @item -mprefer-short-insn-regs
10502 @opindex mprefer-short-insn-regs
10503 Preferrentially allocate registers that allow short instruction generation.
10504 This can result in increasesd instruction count, so if this reduces or
10505 increases code size might vary from case to case.
10507 @item -mbranch-cost=@var{num}
10508 @opindex mbranch-cost
10509 Set the cost of branches to roughly @var{num} ``simple'' instructions.
10510 This cost is only a heuristic and is not guaranteed to produce
10511 consistent results across releases.
10515 Enable the generation of conditional moves.
10517 @item -mnops=@var{num}
10519 Emit @var{num} nops before every other generated instruction.
10521 @item -mno-soft-cmpsf
10522 @opindex mno-soft-cmpsf
10523 For single-precision floating-point comparisons, emit an @code{fsub} instruction
10524 and test the flags. This is faster than a software comparison, but can
10525 get incorrect results in the presence of NaNs, or when two different small
10526 numbers are compared such that their difference is calculated as zero.
10527 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
10528 software comparisons.
10530 @item -mstack-offset=@var{num}
10531 @opindex mstack-offset
10532 Set the offset between the top of the stack and the stack pointer.
10533 E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7
10534 can be used by leaf functions without stack allocation.
10535 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
10536 Note also that this option changes the ABI, compiling a program with a
10537 different stack offset than the libraries have been compiled with
10538 generally does not work.
10539 This option can be useful if you want to evaluate if a different stack
10540 offset would give you better code, but to actually use a different stack
10541 offset to build working programs, it is recommended to configure the
10542 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
10544 @item -mno-round-nearest
10545 @opindex mno-round-nearest
10546 Make the scheduler assume that the rounding mode has been set to
10547 truncating. The default is @option{-mround-nearest}.
10550 @opindex mlong-calls
10551 If not otherwise specified by an attribute, assume all calls might be beyond
10552 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
10553 function address into a register before performing a (otherwise direct) call.
10554 This is the default.
10556 @item -mshort-calls
10557 @opindex short-calls
10558 If not otherwise specified by an attribute, assume all direct calls are
10559 in the range of the @code{b} / @code{bl} instructions, so use these instructions
10560 for direct calls. The default is @option{-mlong-calls}.
10564 Assume addresses can be loaded as 16-bit unsigned values. This does not
10565 apply to function addresses for which @option{-mlong-calls} semantics
10568 @item -mfp-mode=@var{mode}
10570 Set the prevailing mode of the floating-point unit.
10571 This determines the floating-point mode that is provided and expected
10572 at function call and return time. Making this mode match the mode you
10573 predominantly need at function start can make your programs smaller and
10574 faster by avoiding unnecessary mode switches.
10576 @var{mode} can be set to one the following values:
10580 Any mode at function entry is valid, and retained or restored when
10581 the function returns, and when it calls other functions.
10582 This mode is useful for compiling libraries or other compilation units
10583 you might want to incorporate into different programs with different
10584 prevailing FPU modes, and the convenience of being able to use a single
10585 object file outweighs the size and speed overhead for any extra
10586 mode switching that might be needed, compared with what would be needed
10587 with a more specific choice of prevailing FPU mode.
10590 This is the mode used for floating-point calculations with
10591 truncating (i.e.@: round towards zero) rounding mode. That includes
10592 conversion from floating point to integer.
10594 @item round-nearest
10595 This is the mode used for floating-point calculations with
10596 round-to-nearest-or-even rounding mode.
10599 This is the mode used to perform integer calculations in the FPU, e.g.@:
10600 integer multiply, or integer multiply-and-accumulate.
10603 The default is @option{-mfp-mode=caller}
10605 @item -mnosplit-lohi
10606 @opindex mnosplit-lohi
10608 @opindex mno-postinc
10609 @item -mno-postmodify
10610 @opindex mno-postmodify
10611 Code generation tweaks that disable, respectively, splitting of 32-bit
10612 loads, generation of post-increment addresses, and generation of
10613 post-modify addresses. The defaults are @option{msplit-lohi},
10614 @option{-mpost-inc}, and @option{-mpost-modify}.
10616 @item -mnovect-double
10617 @opindex mno-vect-double
10618 Change the preferred SIMD mode to SImode. The default is
10619 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
10621 @item -max-vect-align=@var{num}
10622 @opindex max-vect-align
10623 The maximum alignment for SIMD vector mode types.
10624 @var{num} may be 4 or 8. The default is 8.
10625 Note that this is an ABI change, even though many library function
10626 interfaces are unaffected if they don't use SIMD vector modes
10627 in places that affect size and/or alignment of relevant types.
10629 @item -msplit-vecmove-early
10630 @opindex msplit-vecmove-early
10631 Split vector moves into single word moves before reload. In theory this
10632 can give better register allocation, but so far the reverse seems to be
10633 generally the case.
10635 @item -m1reg-@var{reg}
10637 Specify a register to hold the constant @minus{}1, which makes loading small negative
10638 constants and certain bitmasks faster.
10639 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
10640 which specify use of that register as a fixed register,
10641 and @samp{none}, which means that no register is used for this
10642 purpose. The default is @option{-m1reg-none}.
10647 @subsection ARM Options
10648 @cindex ARM options
10650 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10654 @item -mabi=@var{name}
10656 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10657 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10660 @opindex mapcs-frame
10661 Generate a stack frame that is compliant with the ARM Procedure Call
10662 Standard for all functions, even if this is not strictly necessary for
10663 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10664 with this option causes the stack frames not to be generated for
10665 leaf functions. The default is @option{-mno-apcs-frame}.
10669 This is a synonym for @option{-mapcs-frame}.
10672 @c not currently implemented
10673 @item -mapcs-stack-check
10674 @opindex mapcs-stack-check
10675 Generate code to check the amount of stack space available upon entry to
10676 every function (that actually uses some stack space). If there is
10677 insufficient space available then either the function
10678 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} is
10679 called, depending upon the amount of stack space required. The runtime
10680 system is required to provide these functions. The default is
10681 @option{-mno-apcs-stack-check}, since this produces smaller code.
10683 @c not currently implemented
10685 @opindex mapcs-float
10686 Pass floating-point arguments using the floating-point registers. This is
10687 one of the variants of the APCS@. This option is recommended if the
10688 target hardware has a floating-point unit or if a lot of floating-point
10689 arithmetic is going to be performed by the code. The default is
10690 @option{-mno-apcs-float}, since integer only code is slightly increased in
10691 size if @option{-mapcs-float} is used.
10693 @c not currently implemented
10694 @item -mapcs-reentrant
10695 @opindex mapcs-reentrant
10696 Generate reentrant, position independent code. The default is
10697 @option{-mno-apcs-reentrant}.
10700 @item -mthumb-interwork
10701 @opindex mthumb-interwork
10702 Generate code that supports calling between the ARM and Thumb
10703 instruction sets. Without this option, on pre-v5 architectures, the
10704 two instruction sets cannot be reliably used inside one program. The
10705 default is @option{-mno-thumb-interwork}, since slightly larger code
10706 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10707 configurations this option is meaningless.
10709 @item -mno-sched-prolog
10710 @opindex mno-sched-prolog
10711 Prevent the reordering of instructions in the function prologue, or the
10712 merging of those instruction with the instructions in the function's
10713 body. This means that all functions start with a recognizable set
10714 of instructions (or in fact one of a choice from a small set of
10715 different function prologues), and this information can be used to
10716 locate the start of functions inside an executable piece of code. The
10717 default is @option{-msched-prolog}.
10719 @item -mfloat-abi=@var{name}
10720 @opindex mfloat-abi
10721 Specifies which floating-point ABI to use. Permissible values
10722 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10724 Specifying @samp{soft} causes GCC to generate output containing
10725 library calls for floating-point operations.
10726 @samp{softfp} allows the generation of code using hardware floating-point
10727 instructions, but still uses the soft-float calling conventions.
10728 @samp{hard} allows generation of floating-point instructions
10729 and uses FPU-specific calling conventions.
10731 The default depends on the specific target configuration. Note that
10732 the hard-float and soft-float ABIs are not link-compatible; you must
10733 compile your entire program with the same ABI, and link with a
10734 compatible set of libraries.
10736 @item -mlittle-endian
10737 @opindex mlittle-endian
10738 Generate code for a processor running in little-endian mode. This is
10739 the default for all standard configurations.
10742 @opindex mbig-endian
10743 Generate code for a processor running in big-endian mode; the default is
10744 to compile code for a little-endian processor.
10746 @item -mwords-little-endian
10747 @opindex mwords-little-endian
10748 This option only applies when generating code for big-endian processors.
10749 Generate code for a little-endian word order but a big-endian byte
10750 order. That is, a byte order of the form @samp{32107654}. Note: this
10751 option should only be used if you require compatibility with code for
10752 big-endian ARM processors generated by versions of the compiler prior to
10753 2.8. This option is now deprecated.
10755 @item -mcpu=@var{name}
10757 This specifies the name of the target ARM processor. GCC uses this name
10758 to determine what kind of instructions it can emit when generating
10759 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10760 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10761 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10762 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10763 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10765 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10766 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10767 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10768 @samp{strongarm1110},
10769 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10770 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10771 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10772 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10773 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10774 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10775 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10776 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
10777 @samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10778 @samp{cortex-m4}, @samp{cortex-m3},
10781 @samp{cortex-m0plus},
10782 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10783 @samp{fa526}, @samp{fa626},
10784 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10787 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10788 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10789 See @option{-mtune} for more information.
10791 @option{-mcpu=native} causes the compiler to auto-detect the CPU
10792 of the build computer. At present, this feature is only supported on
10793 Linux, and not all architectures are recognized. If the auto-detect is
10794 unsuccessful the option has no effect.
10796 @item -mtune=@var{name}
10798 This option is very similar to the @option{-mcpu=} option, except that
10799 instead of specifying the actual target processor type, and hence
10800 restricting which instructions can be used, it specifies that GCC should
10801 tune the performance of the code as if the target were of the type
10802 specified in this option, but still choosing the instructions it
10803 generates based on the CPU specified by a @option{-mcpu=} option.
10804 For some ARM implementations better performance can be obtained by using
10807 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10808 performance for a blend of processors within architecture @var{arch}.
10809 The aim is to generate code that run well on the current most popular
10810 processors, balancing between optimizations that benefit some CPUs in the
10811 range, and avoiding performance pitfalls of other CPUs. The effects of
10812 this option may change in future GCC versions as CPU models come and go.
10814 @option{-mtune=native} causes the compiler to auto-detect the CPU
10815 of the build computer. At present, this feature is only supported on
10816 Linux, and not all architectures are recognized. If the auto-detect is
10817 unsuccessful the option has no effect.
10819 @item -march=@var{name}
10821 This specifies the name of the target ARM architecture. GCC uses this
10822 name to determine what kind of instructions it can emit when generating
10823 assembly code. This option can be used in conjunction with or instead
10824 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10825 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10826 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10827 @samp{armv6}, @samp{armv6j},
10828 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10829 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10830 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10832 @option{-march=native} causes the compiler to auto-detect the architecture
10833 of the build computer. At present, this feature is only supported on
10834 Linux, and not all architectures are recognized. If the auto-detect is
10835 unsuccessful the option has no effect.
10837 @item -mfpu=@var{name}
10838 @itemx -mfpe=@var{number}
10839 @itemx -mfp=@var{number}
10843 This specifies what floating-point hardware (or hardware emulation) is
10844 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10845 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10846 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10847 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10848 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10849 @option{-mfp} and @option{-mfpe} are synonyms for
10850 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10853 If @option{-msoft-float} is specified this specifies the format of
10854 floating-point values.
10856 If the selected floating-point hardware includes the NEON extension
10857 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10858 operations are not generated by GCC's auto-vectorization pass unless
10859 @option{-funsafe-math-optimizations} is also specified. This is
10860 because NEON hardware does not fully implement the IEEE 754 standard for
10861 floating-point arithmetic (in particular denormal values are treated as
10862 zero), so the use of NEON instructions may lead to a loss of precision.
10864 @item -mfp16-format=@var{name}
10865 @opindex mfp16-format
10866 Specify the format of the @code{__fp16} half-precision floating-point type.
10867 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10868 the default is @samp{none}, in which case the @code{__fp16} type is not
10869 defined. @xref{Half-Precision}, for more information.
10871 @item -mstructure-size-boundary=@var{n}
10872 @opindex mstructure-size-boundary
10873 The sizes of all structures and unions are rounded up to a multiple
10874 of the number of bits set by this option. Permissible values are 8, 32
10875 and 64. The default value varies for different toolchains. For the COFF
10876 targeted toolchain the default value is 8. A value of 64 is only allowed
10877 if the underlying ABI supports it.
10879 Specifying the larger number can produce faster, more efficient code, but
10880 can also increase the size of the program. Different values are potentially
10881 incompatible. Code compiled with one value cannot necessarily expect to
10882 work with code or libraries compiled with another value, if they exchange
10883 information using structures or unions.
10885 @item -mabort-on-noreturn
10886 @opindex mabort-on-noreturn
10887 Generate a call to the function @code{abort} at the end of a
10888 @code{noreturn} function. It is executed if the function tries to
10892 @itemx -mno-long-calls
10893 @opindex mlong-calls
10894 @opindex mno-long-calls
10895 Tells the compiler to perform function calls by first loading the
10896 address of the function into a register and then performing a subroutine
10897 call on this register. This switch is needed if the target function
10898 lies outside of the 64-megabyte addressing range of the offset-based
10899 version of subroutine call instruction.
10901 Even if this switch is enabled, not all function calls are turned
10902 into long calls. The heuristic is that static functions, functions
10903 that have the @samp{short-call} attribute, functions that are inside
10904 the scope of a @samp{#pragma no_long_calls} directive, and functions whose
10905 definitions have already been compiled within the current compilation
10906 unit are not turned into long calls. The exceptions to this rule are
10907 that weak function definitions, functions with the @samp{long-call}
10908 attribute or the @samp{section} attribute, and functions that are within
10909 the scope of a @samp{#pragma long_calls} directive are always
10910 turned into long calls.
10912 This feature is not enabled by default. Specifying
10913 @option{-mno-long-calls} restores the default behavior, as does
10914 placing the function calls within the scope of a @samp{#pragma
10915 long_calls_off} directive. Note these switches have no effect on how
10916 the compiler generates code to handle function calls via function
10919 @item -msingle-pic-base
10920 @opindex msingle-pic-base
10921 Treat the register used for PIC addressing as read-only, rather than
10922 loading it in the prologue for each function. The runtime system is
10923 responsible for initializing this register with an appropriate value
10924 before execution begins.
10926 @item -mpic-register=@var{reg}
10927 @opindex mpic-register
10928 Specify the register to be used for PIC addressing. The default is R10
10929 unless stack-checking is enabled, when R9 is used.
10931 @item -mcirrus-fix-invalid-insns
10932 @opindex mcirrus-fix-invalid-insns
10933 @opindex mno-cirrus-fix-invalid-insns
10934 Insert NOPs into the instruction stream to in order to work around
10935 problems with invalid Maverick instruction combinations. This option
10936 is only valid if the @option{-mcpu=ep9312} option has been used to
10937 enable generation of instructions for the Cirrus Maverick floating-point
10938 co-processor. This option is not enabled by default, since the
10939 problem is only present in older Maverick implementations. The default
10940 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10943 @item -mpoke-function-name
10944 @opindex mpoke-function-name
10945 Write the name of each function into the text section, directly
10946 preceding the function prologue. The generated code is similar to this:
10950 .ascii "arm_poke_function_name", 0
10953 .word 0xff000000 + (t1 - t0)
10954 arm_poke_function_name
10956 stmfd sp!, @{fp, ip, lr, pc@}
10960 When performing a stack backtrace, code can inspect the value of
10961 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10962 location @code{pc - 12} and the top 8 bits are set, then we know that
10963 there is a function name embedded immediately preceding this location
10964 and has length @code{((pc[-3]) & 0xff000000)}.
10971 Select between generating code that executes in ARM and Thumb
10972 states. The default for most configurations is to generate code
10973 that executes in ARM state, but the default can be changed by
10974 configuring GCC with the @option{--with-mode=}@var{state}
10978 @opindex mtpcs-frame
10979 Generate a stack frame that is compliant with the Thumb Procedure Call
10980 Standard for all non-leaf functions. (A leaf function is one that does
10981 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10983 @item -mtpcs-leaf-frame
10984 @opindex mtpcs-leaf-frame
10985 Generate a stack frame that is compliant with the Thumb Procedure Call
10986 Standard for all leaf functions. (A leaf function is one that does
10987 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10989 @item -mcallee-super-interworking
10990 @opindex mcallee-super-interworking
10991 Gives all externally visible functions in the file being compiled an ARM
10992 instruction set header which switches to Thumb mode before executing the
10993 rest of the function. This allows these functions to be called from
10994 non-interworking code. This option is not valid in AAPCS configurations
10995 because interworking is enabled by default.
10997 @item -mcaller-super-interworking
10998 @opindex mcaller-super-interworking
10999 Allows calls via function pointers (including virtual functions) to
11000 execute correctly regardless of whether the target code has been
11001 compiled for interworking or not. There is a small overhead in the cost
11002 of executing a function pointer if this option is enabled. This option
11003 is not valid in AAPCS configurations because interworking is enabled
11006 @item -mtp=@var{name}
11008 Specify the access model for the thread local storage pointer. The valid
11009 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
11010 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
11011 (supported in the arm6k architecture), and @option{auto}, which uses the
11012 best available method for the selected processor. The default setting is
11015 @item -mtls-dialect=@var{dialect}
11016 @opindex mtls-dialect
11017 Specify the dialect to use for accessing thread local storage. Two
11018 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
11019 @samp{gnu} dialect selects the original GNU scheme for supporting
11020 local and global dynamic TLS models. The @samp{gnu2} dialect
11021 selects the GNU descriptor scheme, which provides better performance
11022 for shared libraries. The GNU descriptor scheme is compatible with
11023 the original scheme, but does require new assembler, linker and
11024 library support. Initial and local exec TLS models are unaffected by
11025 this option and always use the original scheme.
11027 @item -mword-relocations
11028 @opindex mword-relocations
11029 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
11030 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
11031 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
11034 @item -mfix-cortex-m3-ldrd
11035 @opindex mfix-cortex-m3-ldrd
11036 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
11037 with overlapping destination and base registers are used. This option avoids
11038 generating these instructions. This option is enabled by default when
11039 @option{-mcpu=cortex-m3} is specified.
11044 @subsection AVR Options
11045 @cindex AVR Options
11047 These options are defined for AVR implementations:
11050 @item -mmcu=@var{mcu}
11052 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
11054 For a complete list of @var{mcu} values that are supported by @command{avr-gcc},
11055 see the compiler output when called with the @option{--help=target}
11056 command line option.
11057 The default for this option is@tie{}@code{avr2}.
11059 GCC supports the following AVR devices and ISAs:
11061 @include avr-mmcu.texi
11063 @item -maccumulate-args
11064 @opindex maccumulate-args
11065 Accumulate outgoing function arguments and acquire/release the needed
11066 stack space for outgoing function arguments once in function
11067 prologue/epilogue. Without this option, outgoing arguments are pushed
11068 before calling a function and popped afterwards.
11070 Popping the arguments after the function call can be expensive on
11071 AVR so that accumulating the stack space might lead to smaller
11072 executables because arguments need not to be removed from the
11073 stack after such a function call.
11075 This option can lead to reduced code size for functions that perform
11076 several calls to functions that get their arguments on the stack like
11077 calls to printf-like functions.
11079 @item -mbranch-cost=@var{cost}
11080 @opindex mbranch-cost
11081 Set the branch costs for conditional branch instructions to
11082 @var{cost}. Reasonable values for @var{cost} are small, non-negative
11083 integers. The default branch cost is 0.
11085 @item -mcall-prologues
11086 @opindex mcall-prologues
11087 Functions prologues/epilogues are expanded as calls to appropriate
11088 subroutines. Code size is smaller.
11092 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
11093 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
11094 and @code{long long} is 4 bytes. Please note that this option does not
11095 conform to the C standards, but it results in smaller code
11098 @item -mno-interrupts
11099 @opindex mno-interrupts
11100 Generated code is not compatible with hardware interrupts.
11101 Code size is smaller.
11105 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
11106 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
11107 Setting @code{-mrelax} just adds the @code{--relax} option to the
11108 linker command line when the linker is called.
11110 Jump relaxing is performed by the linker because jump offsets are not
11111 known before code is located. Therefore, the assembler code generated by the
11112 compiler is the same, but the instructions in the executable may
11113 differ from instructions in the assembler code.
11115 Relaxing must be turned on if linker stubs are needed, see the
11116 section on @code{EIND} and linker stubs below.
11118 @item -mshort-calls
11119 @opindex mshort-calls
11120 Use @code{RCALL}/@code{RJMP} instructions even on devices with
11121 16@tie{}KiB or more of program memory, i.e.@: on devices that
11122 have the @code{CALL} and @code{JMP} instructions.
11123 See also the @code{-mrelax} command line option.
11127 Treat the stack pointer register as an 8-bit register,
11128 i.e.@: assume the high byte of the stack pointer is zero.
11129 In general, you don't need to set this option by hand.
11131 This option is used internally by the compiler to select and
11132 build multilibs for architectures @code{avr2} and @code{avr25}.
11133 These architectures mix devices with and without @code{SPH}.
11134 For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25}
11135 the compiler driver will add or remove this option from the compiler
11136 proper's command line, because the compiler then knows if the device
11137 or architecture has an 8-bit stack pointer and thus no @code{SPH}
11142 Use address register @code{X} in a way proposed by the hardware. This means
11143 that @code{X} is only used in indirect, post-increment or
11144 pre-decrement addressing.
11146 Without this option, the @code{X} register may be used in the same way
11147 as @code{Y} or @code{Z} which then is emulated by additional
11149 For example, loading a value with @code{X+const} addressing with a
11150 small non-negative @code{const < 64} to a register @var{Rn} is
11154 adiw r26, const ; X += const
11155 ld @var{Rn}, X ; @var{Rn} = *X
11156 sbiw r26, const ; X -= const
11160 @opindex mtiny-stack
11161 Only change the lower 8@tie{}bits of the stack pointer.
11164 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
11165 @cindex @code{EIND}
11166 Pointers in the implementation are 16@tie{}bits wide.
11167 The address of a function or label is represented as word address so
11168 that indirect jumps and calls can target any code address in the
11169 range of 64@tie{}Ki words.
11171 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
11172 bytes of program memory space, there is a special function register called
11173 @code{EIND} that serves as most significant part of the target address
11174 when @code{EICALL} or @code{EIJMP} instructions are used.
11176 Indirect jumps and calls on these devices are handled as follows by
11177 the compiler and are subject to some limitations:
11182 The compiler never sets @code{EIND}.
11185 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
11186 instructions or might read @code{EIND} directly in order to emulate an
11187 indirect call/jump by means of a @code{RET} instruction.
11190 The compiler assumes that @code{EIND} never changes during the startup
11191 code or during the application. In particular, @code{EIND} is not
11192 saved/restored in function or interrupt service routine
11196 For indirect calls to functions and computed goto, the linker
11197 generates @emph{stubs}. Stubs are jump pads sometimes also called
11198 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
11199 The stub contains a direct jump to the desired address.
11202 Linker relaxation must be turned on so that the linker will generate
11203 the stubs correctly an all situaltion. See the compiler option
11204 @code{-mrelax} and the linler option @code{--relax}.
11205 There are corner cases where the linker is supposed to generate stubs
11206 but aborts without relaxation and without a helpful error message.
11209 The default linker script is arranged for code with @code{EIND = 0}.
11210 If code is supposed to work for a setup with @code{EIND != 0}, a custom
11211 linker script has to be used in order to place the sections whose
11212 name start with @code{.trampolines} into the segment where @code{EIND}
11216 The startup code from libgcc never sets @code{EIND}.
11217 Notice that startup code is a blend of code from libgcc and AVR-LibC.
11218 For the impact of AVR-LibC on @code{EIND}, see the
11219 @w{@uref{http://nongnu.org/avr-libc/user-manual,AVR-LibC user manual}}.
11222 It is legitimate for user-specific startup code to set up @code{EIND}
11223 early, for example by means of initialization code located in
11224 section @code{.init3}. Such code runs prior to general startup code
11225 that initializes RAM and calls constructors, but after the bit
11226 of startup code from AVR-LibC that sets @code{EIND} to the segment
11227 where the vector table is located.
11229 #include <avr/io.h>
11232 __attribute__((section(".init3"),naked,used,no_instrument_function))
11233 init3_set_eind (void)
11235 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
11236 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
11241 The @code{__trampolines_start} symbol is defined in the linker script.
11244 Stubs are generated automatically by the linker if
11245 the following two conditions are met:
11248 @item The address of a label is taken by means of the @code{gs} modifier
11249 (short for @emph{generate stubs}) like so:
11251 LDI r24, lo8(gs(@var{func}))
11252 LDI r25, hi8(gs(@var{func}))
11254 @item The final location of that label is in a code segment
11255 @emph{outside} the segment where the stubs are located.
11259 The compiler emits such @code{gs} modifiers for code labels in the
11260 following situations:
11262 @item Taking address of a function or code label.
11263 @item Computed goto.
11264 @item If prologue-save function is used, see @option{-mcall-prologues}
11265 command-line option.
11266 @item Switch/case dispatch tables. If you do not want such dispatch
11267 tables you can specify the @option{-fno-jump-tables} command-line option.
11268 @item C and C++ constructors/destructors called during startup/shutdown.
11269 @item If the tools hit a @code{gs()} modifier explained above.
11273 Jumping to non-symbolic addresses like so is @emph{not} supported:
11278 /* Call function at word address 0x2 */
11279 return ((int(*)(void)) 0x2)();
11283 Instead, a stub has to be set up, i.e.@: the function has to be called
11284 through a symbol (@code{func_4} in the example):
11289 extern int func_4 (void);
11291 /* Call function at byte address 0x4 */
11296 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
11297 Alternatively, @code{func_4} can be defined in the linker script.
11300 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
11301 @cindex @code{RAMPD}
11302 @cindex @code{RAMPX}
11303 @cindex @code{RAMPY}
11304 @cindex @code{RAMPZ}
11305 Some AVR devices support memories larger than the 64@tie{}KiB range
11306 that can be accessed with 16-bit pointers. To access memory locations
11307 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
11308 register is used as high part of the address:
11309 The @code{X}, @code{Y}, @code{Z} address register is concatenated
11310 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
11311 register, respectively, to get a wide address. Similarly,
11312 @code{RAMPD} is used together with direct addressing.
11316 The startup code initializes the @code{RAMP} special function
11317 registers with zero.
11320 If a @ref{AVR Named Address Spaces,named address space} other than
11321 generic or @code{__flash} is used, then @code{RAMPZ} is set
11322 as needed before the operation.
11325 If the device supports RAM larger than 64@tie{KiB} and the compiler
11326 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
11327 is reset to zero after the operation.
11330 If the device comes with a specific @code{RAMP} register, the ISR
11331 prologue/epilogue saves/restores that SFR and initializes it with
11332 zero in case the ISR code might (implicitly) use it.
11335 RAM larger than 64@tie{KiB} is not supported by GCC for AVR targets.
11336 If you use inline assembler to read from locations outside the
11337 16-bit address range and change one of the @code{RAMP} registers,
11338 you must reset it to zero after the access.
11342 @subsubsection AVR Built-in Macros
11344 GCC defines several built-in macros so that the user code can test
11345 for the presence or absence of features. Almost any of the following
11346 built-in macros are deduced from device capabilities and thus
11347 triggered by the @code{-mmcu=} command-line option.
11349 For even more AVR-specific built-in macros see
11350 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
11354 @item __AVR_@var{Device}__
11355 Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects
11356 the device's name. For example, @code{-mmcu=atmega8} defines the
11357 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
11358 @code{__AVR_ATtiny261A__}, etc.
11360 The built-in macros' names follow
11361 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
11362 the device name as from the AVR user manual. The difference between
11363 @var{Device} in the built-in macro and @var{device} in
11364 @code{-mmcu=@var{device}} is that the latter is always lowercase.
11366 @item __AVR_HAVE_ELPM__
11367 The device has the the @code{ELPM} instruction.
11369 @item __AVR_HAVE_ELPMX__
11370 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
11371 R@var{n},Z+} instructions.
11373 @item __AVR_HAVE_MOVW__
11374 The device has the @code{MOVW} instruction to perform 16-bit
11375 register-register moves.
11377 @item __AVR_HAVE_LPMX__
11378 The device has the @code{LPM R@var{n},Z} and
11379 @code{LPM R@var{n},Z+} instructions.
11381 @item __AVR_HAVE_MUL__
11382 The device has a hardware multiplier.
11384 @item __AVR_HAVE_JMP_CALL__
11385 The device has the @code{JMP} and @code{CALL} instructions.
11386 This is the case for devices with at least 16@tie{}KiB of program
11387 memory and if @code{-mshort-calls} is not set.
11389 @item __AVR_HAVE_EIJMP_EICALL__
11390 @item __AVR_3_BYTE_PC__
11391 The device has the @code{EIJMP} and @code{EICALL} instructions.
11392 This is the case for devices with more than 128@tie{}KiB of program memory.
11393 This also means that the program counter
11394 (PC) is 3@tie{}bytes wide.
11396 @item __AVR_2_BYTE_PC__
11397 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
11398 with up to 128@tie{}KiB of program memory.
11400 @item __AVR_HAVE_8BIT_SP__
11401 @item __AVR_HAVE_16BIT_SP__
11402 The stack pointer (SP) register is treated as 8-bit respectively
11403 16-bit register by the compiler.
11404 The definition of these macros is affected by @code{-mtiny-stack}.
11406 @item __AVR_HAVE_SPH__
11408 The device has the SPH (high part of stack pointer) special function
11409 register or has an 8-bit stack pointer, respectively.
11410 The definition of these macros is affected by @code{-mmcu=} and
11411 in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also
11414 @item __AVR_HAVE_RAMPD__
11415 @item __AVR_HAVE_RAMPX__
11416 @item __AVR_HAVE_RAMPY__
11417 @item __AVR_HAVE_RAMPZ__
11418 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
11419 @code{RAMPZ} special function register, respectively.
11421 @item __NO_INTERRUPTS__
11422 This macro reflects the @code{-mno-interrupts} command line option.
11424 @item __AVR_ERRATA_SKIP__
11425 @item __AVR_ERRATA_SKIP_JMP_CALL__
11426 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
11427 instructions because of a hardware erratum. Skip instructions are
11428 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
11429 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
11432 @item __AVR_SFR_OFFSET__=@var{offset}
11433 Instructions that can address I/O special function registers directly
11434 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
11435 address as if addressed by an instruction to access RAM like @code{LD}
11436 or @code{STS}. This offset depends on the device architecture and has
11437 to be subtracted from the RAM address in order to get the
11438 respective I/O@tie{}address.
11442 @node Blackfin Options
11443 @subsection Blackfin Options
11444 @cindex Blackfin Options
11447 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
11449 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
11450 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
11451 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
11452 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
11453 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
11454 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
11455 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
11456 @samp{bf561}, @samp{bf592}.
11457 The optional @var{sirevision} specifies the silicon revision of the target
11458 Blackfin processor. Any workarounds available for the targeted silicon revision
11459 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
11460 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
11461 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
11462 hexadecimal digits representing the major and minor numbers in the silicon
11463 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
11464 is not defined. If @var{sirevision} is @samp{any}, the
11465 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
11466 If this optional @var{sirevision} is not used, GCC assumes the latest known
11467 silicon revision of the targeted Blackfin processor.
11469 Support for @samp{bf561} is incomplete. For @samp{bf561},
11470 Only the processor macro is defined.
11471 Without this option, @samp{bf532} is used as the processor by default.
11472 The corresponding predefined processor macros for @var{cpu} is to
11473 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
11474 provided by libgloss to be linked in if @option{-msim} is not given.
11478 Specifies that the program will be run on the simulator. This causes
11479 the simulator BSP provided by libgloss to be linked in. This option
11480 has effect only for @samp{bfin-elf} toolchain.
11481 Certain other options, such as @option{-mid-shared-library} and
11482 @option{-mfdpic}, imply @option{-msim}.
11484 @item -momit-leaf-frame-pointer
11485 @opindex momit-leaf-frame-pointer
11486 Don't keep the frame pointer in a register for leaf functions. This
11487 avoids the instructions to save, set up and restore frame pointers and
11488 makes an extra register available in leaf functions. The option
11489 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
11490 which might make debugging harder.
11492 @item -mspecld-anomaly
11493 @opindex mspecld-anomaly
11494 When enabled, the compiler ensures that the generated code does not
11495 contain speculative loads after jump instructions. If this option is used,
11496 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
11498 @item -mno-specld-anomaly
11499 @opindex mno-specld-anomaly
11500 Don't generate extra code to prevent speculative loads from occurring.
11502 @item -mcsync-anomaly
11503 @opindex mcsync-anomaly
11504 When enabled, the compiler ensures that the generated code does not
11505 contain CSYNC or SSYNC instructions too soon after conditional branches.
11506 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
11508 @item -mno-csync-anomaly
11509 @opindex mno-csync-anomaly
11510 Don't generate extra code to prevent CSYNC or SSYNC instructions from
11511 occurring too soon after a conditional branch.
11515 When enabled, the compiler is free to take advantage of the knowledge that
11516 the entire program fits into the low 64k of memory.
11519 @opindex mno-low-64k
11520 Assume that the program is arbitrarily large. This is the default.
11522 @item -mstack-check-l1
11523 @opindex mstack-check-l1
11524 Do stack checking using information placed into L1 scratchpad memory by the
11527 @item -mid-shared-library
11528 @opindex mid-shared-library
11529 Generate code that supports shared libraries via the library ID method.
11530 This allows for execute in place and shared libraries in an environment
11531 without virtual memory management. This option implies @option{-fPIC}.
11532 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11534 @item -mno-id-shared-library
11535 @opindex mno-id-shared-library
11536 Generate code that doesn't assume ID based shared libraries are being used.
11537 This is the default.
11539 @item -mleaf-id-shared-library
11540 @opindex mleaf-id-shared-library
11541 Generate code that supports shared libraries via the library ID method,
11542 but assumes that this library or executable won't link against any other
11543 ID shared libraries. That allows the compiler to use faster code for jumps
11546 @item -mno-leaf-id-shared-library
11547 @opindex mno-leaf-id-shared-library
11548 Do not assume that the code being compiled won't link against any ID shared
11549 libraries. Slower code is generated for jump and call insns.
11551 @item -mshared-library-id=n
11552 @opindex mshared-library-id
11553 Specifies the identification number of the ID-based shared library being
11554 compiled. Specifying a value of 0 generates more compact code; specifying
11555 other values forces the allocation of that number to the current
11556 library but is no more space- or time-efficient than omitting this option.
11560 Generate code that allows the data segment to be located in a different
11561 area of memory from the text segment. This allows for execute in place in
11562 an environment without virtual memory management by eliminating relocations
11563 against the text section.
11565 @item -mno-sep-data
11566 @opindex mno-sep-data
11567 Generate code that assumes that the data segment follows the text segment.
11568 This is the default.
11571 @itemx -mno-long-calls
11572 @opindex mlong-calls
11573 @opindex mno-long-calls
11574 Tells the compiler to perform function calls by first loading the
11575 address of the function into a register and then performing a subroutine
11576 call on this register. This switch is needed if the target function
11577 lies outside of the 24-bit addressing range of the offset-based
11578 version of subroutine call instruction.
11580 This feature is not enabled by default. Specifying
11581 @option{-mno-long-calls} restores the default behavior. Note these
11582 switches have no effect on how the compiler generates code to handle
11583 function calls via function pointers.
11587 Link with the fast floating-point library. This library relaxes some of
11588 the IEEE floating-point standard's rules for checking inputs against
11589 Not-a-Number (NAN), in the interest of performance.
11592 @opindex minline-plt
11593 Enable inlining of PLT entries in function calls to functions that are
11594 not known to bind locally. It has no effect without @option{-mfdpic}.
11597 @opindex mmulticore
11598 Build standalone application for multicore Blackfin processor. Proper
11599 start files and link scripts are used to support multicore.
11600 This option defines @code{__BFIN_MULTICORE}. It can only be used with
11601 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
11602 @option{-mcorea} or @option{-mcoreb}. If it's used without
11603 @option{-mcorea} or @option{-mcoreb}, single application/dual core
11604 programming model is used. In this model, the main function of Core B
11605 should be named as coreb_main. If it's used with @option{-mcorea} or
11606 @option{-mcoreb}, one application per core programming model is used.
11607 If this option is not used, single core application programming
11612 Build standalone application for Core A of BF561 when using
11613 one application per core programming model. Proper start files
11614 and link scripts are used to support Core A. This option
11615 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
11619 Build standalone application for Core B of BF561 when using
11620 one application per core programming model. Proper start files
11621 and link scripts are used to support Core B. This option
11622 defines @code{__BFIN_COREB}. When this option is used, coreb_main
11623 should be used instead of main. It must be used with
11624 @option{-mmulticore}.
11628 Build standalone application for SDRAM. Proper start files and
11629 link scripts are used to put the application into SDRAM.
11630 Loader should initialize SDRAM before loading the application
11631 into SDRAM. This option defines @code{__BFIN_SDRAM}.
11635 Assume that ICPLBs are enabled at run time. This has an effect on certain
11636 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
11637 are enabled; for standalone applications the default is off.
11641 @subsection C6X Options
11642 @cindex C6X Options
11645 @item -march=@var{name}
11647 This specifies the name of the target architecture. GCC uses this
11648 name to determine what kind of instructions it can emit when generating
11649 assembly code. Permissible names are: @samp{c62x},
11650 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
11653 @opindex mbig-endian
11654 Generate code for a big-endian target.
11656 @item -mlittle-endian
11657 @opindex mlittle-endian
11658 Generate code for a little-endian target. This is the default.
11662 Choose startup files and linker script suitable for the simulator.
11664 @item -msdata=default
11665 @opindex msdata=default
11666 Put small global and static data in the @samp{.neardata} section,
11667 which is pointed to by register @code{B14}. Put small uninitialized
11668 global and static data in the @samp{.bss} section, which is adjacent
11669 to the @samp{.neardata} section. Put small read-only data into the
11670 @samp{.rodata} section. The corresponding sections used for large
11671 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
11674 @opindex msdata=all
11675 Put all data, not just small objets, into the sections reserved for
11676 small data, and use addressing relative to the @code{B14} register to
11680 @opindex msdata=none
11681 Make no use of the sections reserved for small data, and use absolute
11682 addresses to access all data. Put all initialized global and static
11683 data in the @samp{.fardata} section, and all uninitialized data in the
11684 @samp{.far} section. Put all constant data into the @samp{.const}
11689 @subsection CRIS Options
11690 @cindex CRIS Options
11692 These options are defined specifically for the CRIS ports.
11695 @item -march=@var{architecture-type}
11696 @itemx -mcpu=@var{architecture-type}
11699 Generate code for the specified architecture. The choices for
11700 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
11701 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
11702 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
11705 @item -mtune=@var{architecture-type}
11707 Tune to @var{architecture-type} everything applicable about the generated
11708 code, except for the ABI and the set of available instructions. The
11709 choices for @var{architecture-type} are the same as for
11710 @option{-march=@var{architecture-type}}.
11712 @item -mmax-stack-frame=@var{n}
11713 @opindex mmax-stack-frame
11714 Warn when the stack frame of a function exceeds @var{n} bytes.
11720 The options @option{-metrax4} and @option{-metrax100} are synonyms for
11721 @option{-march=v3} and @option{-march=v8} respectively.
11723 @item -mmul-bug-workaround
11724 @itemx -mno-mul-bug-workaround
11725 @opindex mmul-bug-workaround
11726 @opindex mno-mul-bug-workaround
11727 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
11728 models where it applies. This option is active by default.
11732 Enable CRIS-specific verbose debug-related information in the assembly
11733 code. This option also has the effect to turn off the @samp{#NO_APP}
11734 formatted-code indicator to the assembler at the beginning of the
11739 Do not use condition-code results from previous instruction; always emit
11740 compare and test instructions before use of condition codes.
11742 @item -mno-side-effects
11743 @opindex mno-side-effects
11744 Do not emit instructions with side-effects in addressing modes other than
11747 @item -mstack-align
11748 @itemx -mno-stack-align
11749 @itemx -mdata-align
11750 @itemx -mno-data-align
11751 @itemx -mconst-align
11752 @itemx -mno-const-align
11753 @opindex mstack-align
11754 @opindex mno-stack-align
11755 @opindex mdata-align
11756 @opindex mno-data-align
11757 @opindex mconst-align
11758 @opindex mno-const-align
11759 These options (no-options) arranges (eliminate arrangements) for the
11760 stack-frame, individual data and constants to be aligned for the maximum
11761 single data access size for the chosen CPU model. The default is to
11762 arrange for 32-bit alignment. ABI details such as structure layout are
11763 not affected by these options.
11771 Similar to the stack- data- and const-align options above, these options
11772 arrange for stack-frame, writable data and constants to all be 32-bit,
11773 16-bit or 8-bit aligned. The default is 32-bit alignment.
11775 @item -mno-prologue-epilogue
11776 @itemx -mprologue-epilogue
11777 @opindex mno-prologue-epilogue
11778 @opindex mprologue-epilogue
11779 With @option{-mno-prologue-epilogue}, the normal function prologue and
11780 epilogue which set up the stack frame are omitted and no return
11781 instructions or return sequences are generated in the code. Use this
11782 option only together with visual inspection of the compiled code: no
11783 warnings or errors are generated when call-saved registers must be saved,
11784 or storage for local variable needs to be allocated.
11788 @opindex mno-gotplt
11790 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11791 instruction sequences that load addresses for functions from the PLT part
11792 of the GOT rather than (traditional on other architectures) calls to the
11793 PLT@. The default is @option{-mgotplt}.
11797 Legacy no-op option only recognized with the cris-axis-elf and
11798 cris-axis-linux-gnu targets.
11802 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11806 This option, recognized for the cris-axis-elf arranges
11807 to link with input-output functions from a simulator library. Code,
11808 initialized data and zero-initialized data are allocated consecutively.
11812 Like @option{-sim}, but pass linker options to locate initialized data at
11813 0x40000000 and zero-initialized data at 0x80000000.
11817 @subsection CR16 Options
11818 @cindex CR16 Options
11820 These options are defined specifically for the CR16 ports.
11826 Enable the use of multiply-accumulate instructions. Disabled by default.
11830 @opindex mcr16cplus
11832 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
11837 Links the library libsim.a which is in compatible with simulator. Applicable
11838 to ELF compiler only.
11842 Choose integer type as 32-bit wide.
11846 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
11848 @item -mdata-model=@var{model}
11849 @opindex mdata-model
11850 Choose a data model. The choices for @var{model} are @samp{near},
11851 @samp{far} or @samp{medium}. @samp{medium} is default.
11852 However, @samp{far} is not valid with @option{-mcr16c}, as the
11853 CR16C architecture does not support the far data model.
11856 @node Darwin Options
11857 @subsection Darwin Options
11858 @cindex Darwin options
11860 These options are defined for all architectures running the Darwin operating
11863 FSF GCC on Darwin does not create ``fat'' object files; it creates
11864 an object file for the single architecture that GCC was built to
11865 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11866 @option{-arch} options are used; it does so by running the compiler or
11867 linker multiple times and joining the results together with
11870 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11871 @samp{i686}) is determined by the flags that specify the ISA
11872 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11873 @option{-force_cpusubtype_ALL} option can be used to override this.
11875 The Darwin tools vary in their behavior when presented with an ISA
11876 mismatch. The assembler, @file{as}, only permits instructions to
11877 be used that are valid for the subtype of the file it is generating,
11878 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11879 The linker for shared libraries, @file{/usr/bin/libtool}, fails
11880 and prints an error if asked to create a shared library with a less
11881 restrictive subtype than its input files (for instance, trying to put
11882 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11883 for executables, @command{ld}, quietly gives the executable the most
11884 restrictive subtype of any of its input files.
11889 Add the framework directory @var{dir} to the head of the list of
11890 directories to be searched for header files. These directories are
11891 interleaved with those specified by @option{-I} options and are
11892 scanned in a left-to-right order.
11894 A framework directory is a directory with frameworks in it. A
11895 framework is a directory with a @file{Headers} and/or
11896 @file{PrivateHeaders} directory contained directly in it that ends
11897 in @file{.framework}. The name of a framework is the name of this
11898 directory excluding the @file{.framework}. Headers associated with
11899 the framework are found in one of those two directories, with
11900 @file{Headers} being searched first. A subframework is a framework
11901 directory that is in a framework's @file{Frameworks} directory.
11902 Includes of subframework headers can only appear in a header of a
11903 framework that contains the subframework, or in a sibling subframework
11904 header. Two subframeworks are siblings if they occur in the same
11905 framework. A subframework should not have the same name as a
11906 framework; a warning is issued if this is violated. Currently a
11907 subframework cannot have subframeworks; in the future, the mechanism
11908 may be extended to support this. The standard frameworks can be found
11909 in @file{/System/Library/Frameworks} and
11910 @file{/Library/Frameworks}. An example include looks like
11911 @code{#include <Framework/header.h>}, where @file{Framework} denotes
11912 the name of the framework and @file{header.h} is found in the
11913 @file{PrivateHeaders} or @file{Headers} directory.
11915 @item -iframework@var{dir}
11916 @opindex iframework
11917 Like @option{-F} except the directory is a treated as a system
11918 directory. The main difference between this @option{-iframework} and
11919 @option{-F} is that with @option{-iframework} the compiler does not
11920 warn about constructs contained within header files found via
11921 @var{dir}. This option is valid only for the C family of languages.
11925 Emit debugging information for symbols that are used. For stabs
11926 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11927 This is by default ON@.
11931 Emit debugging information for all symbols and types.
11933 @item -mmacosx-version-min=@var{version}
11934 The earliest version of MacOS X that this executable will run on
11935 is @var{version}. Typical values of @var{version} include @code{10.1},
11936 @code{10.2}, and @code{10.3.9}.
11938 If the compiler was built to use the system's headers by default,
11939 then the default for this option is the system version on which the
11940 compiler is running, otherwise the default is to make choices that
11941 are compatible with as many systems and code bases as possible.
11945 Enable kernel development mode. The @option{-mkernel} option sets
11946 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11947 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11948 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11949 applicable. This mode also sets @option{-mno-altivec},
11950 @option{-msoft-float}, @option{-fno-builtin} and
11951 @option{-mlong-branch} for PowerPC targets.
11953 @item -mone-byte-bool
11954 @opindex mone-byte-bool
11955 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11956 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11957 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11958 option has no effect on x86.
11960 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11961 to generate code that is not binary compatible with code generated
11962 without that switch. Using this switch may require recompiling all
11963 other modules in a program, including system libraries. Use this
11964 switch to conform to a non-default data model.
11966 @item -mfix-and-continue
11967 @itemx -ffix-and-continue
11968 @itemx -findirect-data
11969 @opindex mfix-and-continue
11970 @opindex ffix-and-continue
11971 @opindex findirect-data
11972 Generate code suitable for fast turn around development. Needed to
11973 enable GDB to dynamically load @code{.o} files into already running
11974 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11975 are provided for backwards compatibility.
11979 Loads all members of static archive libraries.
11980 See man ld(1) for more information.
11982 @item -arch_errors_fatal
11983 @opindex arch_errors_fatal
11984 Cause the errors having to do with files that have the wrong architecture
11987 @item -bind_at_load
11988 @opindex bind_at_load
11989 Causes the output file to be marked such that the dynamic linker will
11990 bind all undefined references when the file is loaded or launched.
11994 Produce a Mach-o bundle format file.
11995 See man ld(1) for more information.
11997 @item -bundle_loader @var{executable}
11998 @opindex bundle_loader
11999 This option specifies the @var{executable} that will load the build
12000 output file being linked. See man ld(1) for more information.
12003 @opindex dynamiclib
12004 When passed this option, GCC produces a dynamic library instead of
12005 an executable when linking, using the Darwin @file{libtool} command.
12007 @item -force_cpusubtype_ALL
12008 @opindex force_cpusubtype_ALL
12009 This causes GCC's output file to have the @var{ALL} subtype, instead of
12010 one controlled by the @option{-mcpu} or @option{-march} option.
12012 @item -allowable_client @var{client_name}
12013 @itemx -client_name
12014 @itemx -compatibility_version
12015 @itemx -current_version
12017 @itemx -dependency-file
12019 @itemx -dylinker_install_name
12021 @itemx -exported_symbols_list
12024 @itemx -flat_namespace
12025 @itemx -force_flat_namespace
12026 @itemx -headerpad_max_install_names
12029 @itemx -install_name
12030 @itemx -keep_private_externs
12031 @itemx -multi_module
12032 @itemx -multiply_defined
12033 @itemx -multiply_defined_unused
12036 @itemx -no_dead_strip_inits_and_terms
12037 @itemx -nofixprebinding
12038 @itemx -nomultidefs
12040 @itemx -noseglinkedit
12041 @itemx -pagezero_size
12043 @itemx -prebind_all_twolevel_modules
12044 @itemx -private_bundle
12046 @itemx -read_only_relocs
12048 @itemx -sectobjectsymbols
12052 @itemx -sectobjectsymbols
12055 @itemx -segs_read_only_addr
12057 @itemx -segs_read_write_addr
12058 @itemx -seg_addr_table
12059 @itemx -seg_addr_table_filename
12060 @itemx -seglinkedit
12062 @itemx -segs_read_only_addr
12063 @itemx -segs_read_write_addr
12064 @itemx -single_module
12066 @itemx -sub_library
12068 @itemx -sub_umbrella
12069 @itemx -twolevel_namespace
12072 @itemx -unexported_symbols_list
12073 @itemx -weak_reference_mismatches
12074 @itemx -whatsloaded
12075 @opindex allowable_client
12076 @opindex client_name
12077 @opindex compatibility_version
12078 @opindex current_version
12079 @opindex dead_strip
12080 @opindex dependency-file
12081 @opindex dylib_file
12082 @opindex dylinker_install_name
12084 @opindex exported_symbols_list
12086 @opindex flat_namespace
12087 @opindex force_flat_namespace
12088 @opindex headerpad_max_install_names
12089 @opindex image_base
12091 @opindex install_name
12092 @opindex keep_private_externs
12093 @opindex multi_module
12094 @opindex multiply_defined
12095 @opindex multiply_defined_unused
12096 @opindex noall_load
12097 @opindex no_dead_strip_inits_and_terms
12098 @opindex nofixprebinding
12099 @opindex nomultidefs
12101 @opindex noseglinkedit
12102 @opindex pagezero_size
12104 @opindex prebind_all_twolevel_modules
12105 @opindex private_bundle
12106 @opindex read_only_relocs
12108 @opindex sectobjectsymbols
12111 @opindex sectcreate
12112 @opindex sectobjectsymbols
12115 @opindex segs_read_only_addr
12116 @opindex segs_read_write_addr
12117 @opindex seg_addr_table
12118 @opindex seg_addr_table_filename
12119 @opindex seglinkedit
12121 @opindex segs_read_only_addr
12122 @opindex segs_read_write_addr
12123 @opindex single_module
12125 @opindex sub_library
12126 @opindex sub_umbrella
12127 @opindex twolevel_namespace
12130 @opindex unexported_symbols_list
12131 @opindex weak_reference_mismatches
12132 @opindex whatsloaded
12133 These options are passed to the Darwin linker. The Darwin linker man page
12134 describes them in detail.
12137 @node DEC Alpha Options
12138 @subsection DEC Alpha Options
12140 These @samp{-m} options are defined for the DEC Alpha implementations:
12143 @item -mno-soft-float
12144 @itemx -msoft-float
12145 @opindex mno-soft-float
12146 @opindex msoft-float
12147 Use (do not use) the hardware floating-point instructions for
12148 floating-point operations. When @option{-msoft-float} is specified,
12149 functions in @file{libgcc.a} are used to perform floating-point
12150 operations. Unless they are replaced by routines that emulate the
12151 floating-point operations, or compiled in such a way as to call such
12152 emulations routines, these routines issue floating-point
12153 operations. If you are compiling for an Alpha without floating-point
12154 operations, you must ensure that the library is built so as not to call
12157 Note that Alpha implementations without floating-point operations are
12158 required to have floating-point registers.
12161 @itemx -mno-fp-regs
12163 @opindex mno-fp-regs
12164 Generate code that uses (does not use) the floating-point register set.
12165 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
12166 register set is not used, floating-point operands are passed in integer
12167 registers as if they were integers and floating-point results are passed
12168 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
12169 so any function with a floating-point argument or return value called by code
12170 compiled with @option{-mno-fp-regs} must also be compiled with that
12173 A typical use of this option is building a kernel that does not use,
12174 and hence need not save and restore, any floating-point registers.
12178 The Alpha architecture implements floating-point hardware optimized for
12179 maximum performance. It is mostly compliant with the IEEE floating-point
12180 standard. However, for full compliance, software assistance is
12181 required. This option generates code fully IEEE-compliant code
12182 @emph{except} that the @var{inexact-flag} is not maintained (see below).
12183 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
12184 defined during compilation. The resulting code is less efficient but is
12185 able to correctly support denormalized numbers and exceptional IEEE
12186 values such as not-a-number and plus/minus infinity. Other Alpha
12187 compilers call this option @option{-ieee_with_no_inexact}.
12189 @item -mieee-with-inexact
12190 @opindex mieee-with-inexact
12191 This is like @option{-mieee} except the generated code also maintains
12192 the IEEE @var{inexact-flag}. Turning on this option causes the
12193 generated code to implement fully-compliant IEEE math. In addition to
12194 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
12195 macro. On some Alpha implementations the resulting code may execute
12196 significantly slower than the code generated by default. Since there is
12197 very little code that depends on the @var{inexact-flag}, you should
12198 normally not specify this option. Other Alpha compilers call this
12199 option @option{-ieee_with_inexact}.
12201 @item -mfp-trap-mode=@var{trap-mode}
12202 @opindex mfp-trap-mode
12203 This option controls what floating-point related traps are enabled.
12204 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
12205 The trap mode can be set to one of four values:
12209 This is the default (normal) setting. The only traps that are enabled
12210 are the ones that cannot be disabled in software (e.g., division by zero
12214 In addition to the traps enabled by @samp{n}, underflow traps are enabled
12218 Like @samp{u}, but the instructions are marked to be safe for software
12219 completion (see Alpha architecture manual for details).
12222 Like @samp{su}, but inexact traps are enabled as well.
12225 @item -mfp-rounding-mode=@var{rounding-mode}
12226 @opindex mfp-rounding-mode
12227 Selects the IEEE rounding mode. Other Alpha compilers call this option
12228 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
12233 Normal IEEE rounding mode. Floating-point numbers are rounded towards
12234 the nearest machine number or towards the even machine number in case
12238 Round towards minus infinity.
12241 Chopped rounding mode. Floating-point numbers are rounded towards zero.
12244 Dynamic rounding mode. A field in the floating-point control register
12245 (@var{fpcr}, see Alpha architecture reference manual) controls the
12246 rounding mode in effect. The C library initializes this register for
12247 rounding towards plus infinity. Thus, unless your program modifies the
12248 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
12251 @item -mtrap-precision=@var{trap-precision}
12252 @opindex mtrap-precision
12253 In the Alpha architecture, floating-point traps are imprecise. This
12254 means without software assistance it is impossible to recover from a
12255 floating trap and program execution normally needs to be terminated.
12256 GCC can generate code that can assist operating system trap handlers
12257 in determining the exact location that caused a floating-point trap.
12258 Depending on the requirements of an application, different levels of
12259 precisions can be selected:
12263 Program precision. This option is the default and means a trap handler
12264 can only identify which program caused a floating-point exception.
12267 Function precision. The trap handler can determine the function that
12268 caused a floating-point exception.
12271 Instruction precision. The trap handler can determine the exact
12272 instruction that caused a floating-point exception.
12275 Other Alpha compilers provide the equivalent options called
12276 @option{-scope_safe} and @option{-resumption_safe}.
12278 @item -mieee-conformant
12279 @opindex mieee-conformant
12280 This option marks the generated code as IEEE conformant. You must not
12281 use this option unless you also specify @option{-mtrap-precision=i} and either
12282 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
12283 is to emit the line @samp{.eflag 48} in the function prologue of the
12284 generated assembly file.
12286 @item -mbuild-constants
12287 @opindex mbuild-constants
12288 Normally GCC examines a 32- or 64-bit integer constant to
12289 see if it can construct it from smaller constants in two or three
12290 instructions. If it cannot, it outputs the constant as a literal and
12291 generates code to load it from the data segment at run time.
12293 Use this option to require GCC to construct @emph{all} integer constants
12294 using code, even if it takes more instructions (the maximum is six).
12296 You typically use this option to build a shared library dynamic
12297 loader. Itself a shared library, it must relocate itself in memory
12298 before it can find the variables and constants in its own data segment.
12316 Indicate whether GCC should generate code to use the optional BWX,
12317 CIX, FIX and MAX instruction sets. The default is to use the instruction
12318 sets supported by the CPU type specified via @option{-mcpu=} option or that
12319 of the CPU on which GCC was built if none is specified.
12322 @itemx -mfloat-ieee
12323 @opindex mfloat-vax
12324 @opindex mfloat-ieee
12325 Generate code that uses (does not use) VAX F and G floating-point
12326 arithmetic instead of IEEE single and double precision.
12328 @item -mexplicit-relocs
12329 @itemx -mno-explicit-relocs
12330 @opindex mexplicit-relocs
12331 @opindex mno-explicit-relocs
12332 Older Alpha assemblers provided no way to generate symbol relocations
12333 except via assembler macros. Use of these macros does not allow
12334 optimal instruction scheduling. GNU binutils as of version 2.12
12335 supports a new syntax that allows the compiler to explicitly mark
12336 which relocations should apply to which instructions. This option
12337 is mostly useful for debugging, as GCC detects the capabilities of
12338 the assembler when it is built and sets the default accordingly.
12341 @itemx -mlarge-data
12342 @opindex msmall-data
12343 @opindex mlarge-data
12344 When @option{-mexplicit-relocs} is in effect, static data is
12345 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
12346 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
12347 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
12348 16-bit relocations off of the @code{$gp} register. This limits the
12349 size of the small data area to 64KB, but allows the variables to be
12350 directly accessed via a single instruction.
12352 The default is @option{-mlarge-data}. With this option the data area
12353 is limited to just below 2GB@. Programs that require more than 2GB of
12354 data must use @code{malloc} or @code{mmap} to allocate the data in the
12355 heap instead of in the program's data segment.
12357 When generating code for shared libraries, @option{-fpic} implies
12358 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
12361 @itemx -mlarge-text
12362 @opindex msmall-text
12363 @opindex mlarge-text
12364 When @option{-msmall-text} is used, the compiler assumes that the
12365 code of the entire program (or shared library) fits in 4MB, and is
12366 thus reachable with a branch instruction. When @option{-msmall-data}
12367 is used, the compiler can assume that all local symbols share the
12368 same @code{$gp} value, and thus reduce the number of instructions
12369 required for a function call from 4 to 1.
12371 The default is @option{-mlarge-text}.
12373 @item -mcpu=@var{cpu_type}
12375 Set the instruction set and instruction scheduling parameters for
12376 machine type @var{cpu_type}. You can specify either the @samp{EV}
12377 style name or the corresponding chip number. GCC supports scheduling
12378 parameters for the EV4, EV5 and EV6 family of processors and
12379 chooses the default values for the instruction set from the processor
12380 you specify. If you do not specify a processor type, GCC defaults
12381 to the processor on which the compiler was built.
12383 Supported values for @var{cpu_type} are
12389 Schedules as an EV4 and has no instruction set extensions.
12393 Schedules as an EV5 and has no instruction set extensions.
12397 Schedules as an EV5 and supports the BWX extension.
12402 Schedules as an EV5 and supports the BWX and MAX extensions.
12406 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
12410 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
12413 Native toolchains also support the value @samp{native},
12414 which selects the best architecture option for the host processor.
12415 @option{-mcpu=native} has no effect if GCC does not recognize
12418 @item -mtune=@var{cpu_type}
12420 Set only the instruction scheduling parameters for machine type
12421 @var{cpu_type}. The instruction set is not changed.
12423 Native toolchains also support the value @samp{native},
12424 which selects the best architecture option for the host processor.
12425 @option{-mtune=native} has no effect if GCC does not recognize
12428 @item -mmemory-latency=@var{time}
12429 @opindex mmemory-latency
12430 Sets the latency the scheduler should assume for typical memory
12431 references as seen by the application. This number is highly
12432 dependent on the memory access patterns used by the application
12433 and the size of the external cache on the machine.
12435 Valid options for @var{time} are
12439 A decimal number representing clock cycles.
12445 The compiler contains estimates of the number of clock cycles for
12446 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
12447 (also called Dcache, Scache, and Bcache), as well as to main memory.
12448 Note that L3 is only valid for EV5.
12454 @subsection FR30 Options
12455 @cindex FR30 Options
12457 These options are defined specifically for the FR30 port.
12461 @item -msmall-model
12462 @opindex msmall-model
12463 Use the small address space model. This can produce smaller code, but
12464 it does assume that all symbolic values and addresses fit into a
12469 Assume that runtime support has been provided and so there is no need
12470 to include the simulator library (@file{libsim.a}) on the linker
12476 @subsection FRV Options
12477 @cindex FRV Options
12483 Only use the first 32 general-purpose registers.
12488 Use all 64 general-purpose registers.
12493 Use only the first 32 floating-point registers.
12498 Use all 64 floating-point registers.
12501 @opindex mhard-float
12503 Use hardware instructions for floating-point operations.
12506 @opindex msoft-float
12508 Use library routines for floating-point operations.
12513 Dynamically allocate condition code registers.
12518 Do not try to dynamically allocate condition code registers, only
12519 use @code{icc0} and @code{fcc0}.
12524 Change ABI to use double word insns.
12529 Do not use double word instructions.
12534 Use floating-point double instructions.
12537 @opindex mno-double
12539 Do not use floating-point double instructions.
12544 Use media instructions.
12549 Do not use media instructions.
12554 Use multiply and add/subtract instructions.
12557 @opindex mno-muladd
12559 Do not use multiply and add/subtract instructions.
12564 Select the FDPIC ABI, which uses function descriptors to represent
12565 pointers to functions. Without any PIC/PIE-related options, it
12566 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
12567 assumes GOT entries and small data are within a 12-bit range from the
12568 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
12569 are computed with 32 bits.
12570 With a @samp{bfin-elf} target, this option implies @option{-msim}.
12573 @opindex minline-plt
12575 Enable inlining of PLT entries in function calls to functions that are
12576 not known to bind locally. It has no effect without @option{-mfdpic}.
12577 It's enabled by default if optimizing for speed and compiling for
12578 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
12579 optimization option such as @option{-O3} or above is present in the
12585 Assume a large TLS segment when generating thread-local code.
12590 Do not assume a large TLS segment when generating thread-local code.
12595 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
12596 that is known to be in read-only sections. It's enabled by default,
12597 except for @option{-fpic} or @option{-fpie}: even though it may help
12598 make the global offset table smaller, it trades 1 instruction for 4.
12599 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
12600 one of which may be shared by multiple symbols, and it avoids the need
12601 for a GOT entry for the referenced symbol, so it's more likely to be a
12602 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
12604 @item -multilib-library-pic
12605 @opindex multilib-library-pic
12607 Link with the (library, not FD) pic libraries. It's implied by
12608 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
12609 @option{-fpic} without @option{-mfdpic}. You should never have to use
12613 @opindex mlinked-fp
12615 Follow the EABI requirement of always creating a frame pointer whenever
12616 a stack frame is allocated. This option is enabled by default and can
12617 be disabled with @option{-mno-linked-fp}.
12620 @opindex mlong-calls
12622 Use indirect addressing to call functions outside the current
12623 compilation unit. This allows the functions to be placed anywhere
12624 within the 32-bit address space.
12626 @item -malign-labels
12627 @opindex malign-labels
12629 Try to align labels to an 8-byte boundary by inserting nops into the
12630 previous packet. This option only has an effect when VLIW packing
12631 is enabled. It doesn't create new packets; it merely adds nops to
12634 @item -mlibrary-pic
12635 @opindex mlibrary-pic
12637 Generate position-independent EABI code.
12642 Use only the first four media accumulator registers.
12647 Use all eight media accumulator registers.
12652 Pack VLIW instructions.
12657 Do not pack VLIW instructions.
12660 @opindex mno-eflags
12662 Do not mark ABI switches in e_flags.
12665 @opindex mcond-move
12667 Enable the use of conditional-move instructions (default).
12669 This switch is mainly for debugging the compiler and will likely be removed
12670 in a future version.
12672 @item -mno-cond-move
12673 @opindex mno-cond-move
12675 Disable the use of conditional-move instructions.
12677 This switch is mainly for debugging the compiler and will likely be removed
12678 in a future version.
12683 Enable the use of conditional set instructions (default).
12685 This switch is mainly for debugging the compiler and will likely be removed
12686 in a future version.
12691 Disable the use of conditional set instructions.
12693 This switch is mainly for debugging the compiler and will likely be removed
12694 in a future version.
12697 @opindex mcond-exec
12699 Enable the use of conditional execution (default).
12701 This switch is mainly for debugging the compiler and will likely be removed
12702 in a future version.
12704 @item -mno-cond-exec
12705 @opindex mno-cond-exec
12707 Disable the use of conditional execution.
12709 This switch is mainly for debugging the compiler and will likely be removed
12710 in a future version.
12712 @item -mvliw-branch
12713 @opindex mvliw-branch
12715 Run a pass to pack branches into VLIW instructions (default).
12717 This switch is mainly for debugging the compiler and will likely be removed
12718 in a future version.
12720 @item -mno-vliw-branch
12721 @opindex mno-vliw-branch
12723 Do not run a pass to pack branches into VLIW instructions.
12725 This switch is mainly for debugging the compiler and will likely be removed
12726 in a future version.
12728 @item -mmulti-cond-exec
12729 @opindex mmulti-cond-exec
12731 Enable optimization of @code{&&} and @code{||} in conditional execution
12734 This switch is mainly for debugging the compiler and will likely be removed
12735 in a future version.
12737 @item -mno-multi-cond-exec
12738 @opindex mno-multi-cond-exec
12740 Disable optimization of @code{&&} and @code{||} in conditional execution.
12742 This switch is mainly for debugging the compiler and will likely be removed
12743 in a future version.
12745 @item -mnested-cond-exec
12746 @opindex mnested-cond-exec
12748 Enable nested conditional execution optimizations (default).
12750 This switch is mainly for debugging the compiler and will likely be removed
12751 in a future version.
12753 @item -mno-nested-cond-exec
12754 @opindex mno-nested-cond-exec
12756 Disable nested conditional execution optimizations.
12758 This switch is mainly for debugging the compiler and will likely be removed
12759 in a future version.
12761 @item -moptimize-membar
12762 @opindex moptimize-membar
12764 This switch removes redundant @code{membar} instructions from the
12765 compiler generated code. It is enabled by default.
12767 @item -mno-optimize-membar
12768 @opindex mno-optimize-membar
12770 This switch disables the automatic removal of redundant @code{membar}
12771 instructions from the generated code.
12773 @item -mtomcat-stats
12774 @opindex mtomcat-stats
12776 Cause gas to print out tomcat statistics.
12778 @item -mcpu=@var{cpu}
12781 Select the processor type for which to generate code. Possible values are
12782 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12783 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12787 @node GNU/Linux Options
12788 @subsection GNU/Linux Options
12790 These @samp{-m} options are defined for GNU/Linux targets:
12795 Use the GNU C library. This is the default except
12796 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12800 Use uClibc C library. This is the default on
12801 @samp{*-*-linux-*uclibc*} targets.
12805 Use Bionic C library. This is the default on
12806 @samp{*-*-linux-*android*} targets.
12810 Compile code compatible with Android platform. This is the default on
12811 @samp{*-*-linux-*android*} targets.
12813 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12814 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12815 this option makes the GCC driver pass Android-specific options to the linker.
12816 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12819 @item -tno-android-cc
12820 @opindex tno-android-cc
12821 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12822 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12823 @option{-fno-rtti} by default.
12825 @item -tno-android-ld
12826 @opindex tno-android-ld
12827 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12828 linking options to the linker.
12832 @node H8/300 Options
12833 @subsection H8/300 Options
12835 These @samp{-m} options are defined for the H8/300 implementations:
12840 Shorten some address references at link time, when possible; uses the
12841 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12842 ld, Using ld}, for a fuller description.
12846 Generate code for the H8/300H@.
12850 Generate code for the H8S@.
12854 Generate code for the H8S and H8/300H in the normal mode. This switch
12855 must be used either with @option{-mh} or @option{-ms}.
12859 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12863 Extended registers are stored on stack before execution of function
12864 with monitor attribute. Default option is @option{-mexr}.
12865 This option is valid only for H8S targets.
12869 Extended registers are not stored on stack before execution of function
12870 with monitor attribute. Default option is @option{-mno-exr}.
12871 This option is valid only for H8S targets.
12875 Make @code{int} data 32 bits by default.
12878 @opindex malign-300
12879 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12880 The default for the H8/300H and H8S is to align longs and floats on
12882 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
12883 This option has no effect on the H8/300.
12887 @subsection HPPA Options
12888 @cindex HPPA Options
12890 These @samp{-m} options are defined for the HPPA family of computers:
12893 @item -march=@var{architecture-type}
12895 Generate code for the specified architecture. The choices for
12896 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12897 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12898 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12899 architecture option for your machine. Code compiled for lower numbered
12900 architectures runs on higher numbered architectures, but not the
12903 @item -mpa-risc-1-0
12904 @itemx -mpa-risc-1-1
12905 @itemx -mpa-risc-2-0
12906 @opindex mpa-risc-1-0
12907 @opindex mpa-risc-1-1
12908 @opindex mpa-risc-2-0
12909 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12912 @opindex mbig-switch
12913 Generate code suitable for big switch tables. Use this option only if
12914 the assembler/linker complain about out of range branches within a switch
12917 @item -mjump-in-delay
12918 @opindex mjump-in-delay
12919 Fill delay slots of function calls with unconditional jump instructions
12920 by modifying the return pointer for the function call to be the target
12921 of the conditional jump.
12923 @item -mdisable-fpregs
12924 @opindex mdisable-fpregs
12925 Prevent floating-point registers from being used in any manner. This is
12926 necessary for compiling kernels that perform lazy context switching of
12927 floating-point registers. If you use this option and attempt to perform
12928 floating-point operations, the compiler aborts.
12930 @item -mdisable-indexing
12931 @opindex mdisable-indexing
12932 Prevent the compiler from using indexing address modes. This avoids some
12933 rather obscure problems when compiling MIG generated code under MACH@.
12935 @item -mno-space-regs
12936 @opindex mno-space-regs
12937 Generate code that assumes the target has no space registers. This allows
12938 GCC to generate faster indirect calls and use unscaled index address modes.
12940 Such code is suitable for level 0 PA systems and kernels.
12942 @item -mfast-indirect-calls
12943 @opindex mfast-indirect-calls
12944 Generate code that assumes calls never cross space boundaries. This
12945 allows GCC to emit code that performs faster indirect calls.
12947 This option does not work in the presence of shared libraries or nested
12950 @item -mfixed-range=@var{register-range}
12951 @opindex mfixed-range
12952 Generate code treating the given register range as fixed registers.
12953 A fixed register is one that the register allocator can not use. This is
12954 useful when compiling kernel code. A register range is specified as
12955 two registers separated by a dash. Multiple register ranges can be
12956 specified separated by a comma.
12958 @item -mlong-load-store
12959 @opindex mlong-load-store
12960 Generate 3-instruction load and store sequences as sometimes required by
12961 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12964 @item -mportable-runtime
12965 @opindex mportable-runtime
12966 Use the portable calling conventions proposed by HP for ELF systems.
12970 Enable the use of assembler directives only GAS understands.
12972 @item -mschedule=@var{cpu-type}
12974 Schedule code according to the constraints for the machine type
12975 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12976 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12977 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12978 proper scheduling option for your machine. The default scheduling is
12982 @opindex mlinker-opt
12983 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12984 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12985 linkers in which they give bogus error messages when linking some programs.
12988 @opindex msoft-float
12989 Generate output containing library calls for floating point.
12990 @strong{Warning:} the requisite libraries are not available for all HPPA
12991 targets. Normally the facilities of the machine's usual C compiler are
12992 used, but this cannot be done directly in cross-compilation. You must make
12993 your own arrangements to provide suitable library functions for
12996 @option{-msoft-float} changes the calling convention in the output file;
12997 therefore, it is only useful if you compile @emph{all} of a program with
12998 this option. In particular, you need to compile @file{libgcc.a}, the
12999 library that comes with GCC, with @option{-msoft-float} in order for
13004 Generate the predefine, @code{_SIO}, for server IO@. The default is
13005 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
13006 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
13007 options are available under HP-UX and HI-UX@.
13011 Use options specific to GNU @command{ld}.
13012 This passes @option{-shared} to @command{ld} when
13013 building a shared library. It is the default when GCC is configured,
13014 explicitly or implicitly, with the GNU linker. This option does not
13015 affect which @command{ld} is called; it only changes what parameters
13016 are passed to that @command{ld}.
13017 The @command{ld} that is called is determined by the
13018 @option{--with-ld} configure option, GCC's program search path, and
13019 finally by the user's @env{PATH}. The linker used by GCC can be printed
13020 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
13021 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
13025 Use options specific to HP @command{ld}.
13026 This passes @option{-b} to @command{ld} when building
13027 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
13028 links. It is the default when GCC is configured, explicitly or
13029 implicitly, with the HP linker. This option does not affect
13030 which @command{ld} is called; it only changes what parameters are passed to that
13032 The @command{ld} that is called is determined by the @option{--with-ld}
13033 configure option, GCC's program search path, and finally by the user's
13034 @env{PATH}. The linker used by GCC can be printed using @samp{which
13035 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
13036 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
13039 @opindex mno-long-calls
13040 Generate code that uses long call sequences. This ensures that a call
13041 is always able to reach linker generated stubs. The default is to generate
13042 long calls only when the distance from the call site to the beginning
13043 of the function or translation unit, as the case may be, exceeds a
13044 predefined limit set by the branch type being used. The limits for
13045 normal calls are 7,600,000 and 240,000 bytes, respectively for the
13046 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
13049 Distances are measured from the beginning of functions when using the
13050 @option{-ffunction-sections} option, or when using the @option{-mgas}
13051 and @option{-mno-portable-runtime} options together under HP-UX with
13054 It is normally not desirable to use this option as it degrades
13055 performance. However, it may be useful in large applications,
13056 particularly when partial linking is used to build the application.
13058 The types of long calls used depends on the capabilities of the
13059 assembler and linker, and the type of code being generated. The
13060 impact on systems that support long absolute calls, and long pic
13061 symbol-difference or pc-relative calls should be relatively small.
13062 However, an indirect call is used on 32-bit ELF systems in pic code
13063 and it is quite long.
13065 @item -munix=@var{unix-std}
13067 Generate compiler predefines and select a startfile for the specified
13068 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
13069 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
13070 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
13071 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
13072 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
13075 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
13076 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
13077 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
13078 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
13079 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
13080 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
13082 It is @emph{important} to note that this option changes the interfaces
13083 for various library routines. It also affects the operational behavior
13084 of the C library. Thus, @emph{extreme} care is needed in using this
13087 Library code that is intended to operate with more than one UNIX
13088 standard must test, set and restore the variable @var{__xpg4_extended_mask}
13089 as appropriate. Most GNU software doesn't provide this capability.
13093 Suppress the generation of link options to search libdld.sl when the
13094 @option{-static} option is specified on HP-UX 10 and later.
13098 The HP-UX implementation of setlocale in libc has a dependency on
13099 libdld.sl. There isn't an archive version of libdld.sl. Thus,
13100 when the @option{-static} option is specified, special link options
13101 are needed to resolve this dependency.
13103 On HP-UX 10 and later, the GCC driver adds the necessary options to
13104 link with libdld.sl when the @option{-static} option is specified.
13105 This causes the resulting binary to be dynamic. On the 64-bit port,
13106 the linkers generate dynamic binaries by default in any case. The
13107 @option{-nolibdld} option can be used to prevent the GCC driver from
13108 adding these link options.
13112 Add support for multithreading with the @dfn{dce thread} library
13113 under HP-UX@. This option sets flags for both the preprocessor and
13117 @node i386 and x86-64 Options
13118 @subsection Intel 386 and AMD x86-64 Options
13119 @cindex i386 Options
13120 @cindex x86-64 Options
13121 @cindex Intel 386 Options
13122 @cindex AMD x86-64 Options
13124 These @samp{-m} options are defined for the i386 and x86-64 family of
13129 @item -march=@var{cpu-type}
13131 Generate instructions for the machine type @var{cpu-type}. In contrast to
13132 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
13133 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
13134 to generate code that may not run at all on processors other than the one
13135 indicated. Specifying @option{-march=@var{cpu-type}} implies
13136 @option{-mtune=@var{cpu-type}}.
13138 The choices for @var{cpu-type} are:
13142 This selects the CPU to generate code for at compilation time by determining
13143 the processor type of the compiling machine. Using @option{-march=native}
13144 enables all instruction subsets supported by the local machine (hence
13145 the result might not run on different machines). Using @option{-mtune=native}
13146 produces code optimized for the local machine under the constraints
13147 of the selected instruction set.
13150 Original Intel i386 CPU@.
13153 Intel i486 CPU@. (No scheduling is implemented for this chip.)
13157 Intel Pentium CPU with no MMX support.
13160 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
13163 Intel Pentium Pro CPU@.
13166 When used with @option{-march}, the Pentium Pro
13167 instruction set is used, so the code runs on all i686 family chips.
13168 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
13171 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
13176 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
13180 Intel Pentium M; low-power version of Intel Pentium III CPU
13181 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
13185 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
13188 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
13192 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
13193 SSE2 and SSE3 instruction set support.
13196 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13197 instruction set support.
13200 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
13201 and SSE4.2 instruction set support.
13204 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13205 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
13208 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13209 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
13213 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13214 instruction set support.
13217 AMD K6 CPU with MMX instruction set support.
13221 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
13224 @itemx athlon-tbird
13225 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
13231 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
13232 instruction set support.
13238 Processors based on the AMD K8 core with x86-64 instruction set support,
13239 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
13240 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
13241 instruction set extensions.)
13244 @itemx opteron-sse3
13245 @itemx athlon64-sse3
13246 Improved versions of AMD K8 cores with SSE3 instruction set support.
13250 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
13251 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
13252 instruction set extensions.)
13255 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
13256 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
13257 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
13259 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
13260 supersets BMI, TBM, F16C, FMA, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE,
13261 SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
13265 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
13266 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
13267 instruction set extensions.)
13270 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
13274 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
13275 instruction set support.
13278 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
13279 implemented for this chip.)
13282 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
13284 implemented for this chip.)
13287 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
13290 @item -mtune=@var{cpu-type}
13292 Tune to @var{cpu-type} everything applicable about the generated code, except
13293 for the ABI and the set of available instructions.
13294 While picking a specific @var{cpu-type} schedules things appropriately
13295 for that particular chip, the compiler does not generate any code that
13296 cannot run on the default machine type unless you use a
13297 @option{-march=@var{cpu-type}} option.
13298 For example, if GCC is configured for i686-pc-linux-gnu
13299 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
13300 but still runs on i686 machines.
13302 The choices for @var{cpu-type} are the same as for @option{-march}.
13303 In addition, @option{-mtune} supports an extra choice for @var{cpu-type}:
13307 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
13308 If you know the CPU on which your code will run, then you should use
13309 the corresponding @option{-mtune} or @option{-march} option instead of
13310 @option{-mtune=generic}. But, if you do not know exactly what CPU users
13311 of your application will have, then you should use this option.
13313 As new processors are deployed in the marketplace, the behavior of this
13314 option will change. Therefore, if you upgrade to a newer version of
13315 GCC, code generation controlled by this option will change to reflect
13317 that are most common at the time that version of GCC is released.
13319 There is no @option{-march=generic} option because @option{-march}
13320 indicates the instruction set the compiler can use, and there is no
13321 generic instruction set applicable to all processors. In contrast,
13322 @option{-mtune} indicates the processor (or, in this case, collection of
13323 processors) for which the code is optimized.
13326 @item -mcpu=@var{cpu-type}
13328 A deprecated synonym for @option{-mtune}.
13330 @item -mfpmath=@var{unit}
13332 Generate floating-point arithmetic for selected unit @var{unit}. The choices
13333 for @var{unit} are:
13337 Use the standard 387 floating-point coprocessor present on the majority of chips and
13338 emulated otherwise. Code compiled with this option runs almost everywhere.
13339 The temporary results are computed in 80-bit precision instead of the precision
13340 specified by the type, resulting in slightly different results compared to most
13341 of other chips. See @option{-ffloat-store} for more detailed description.
13343 This is the default choice for i386 compiler.
13346 Use scalar floating-point instructions present in the SSE instruction set.
13347 This instruction set is supported by Pentium III and newer chips,
13348 and in the AMD line
13349 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
13350 instruction set supports only single-precision arithmetic, thus the double and
13351 extended-precision arithmetic are still done using 387. A later version, present
13352 only in Pentium 4 and AMD x86-64 chips, supports double-precision
13355 For the i386 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
13356 or @option{-msse2} switches to enable SSE extensions and make this option
13357 effective. For the x86-64 compiler, these extensions are enabled by default.
13359 The resulting code should be considerably faster in the majority of cases and avoid
13360 the numerical instability problems of 387 code, but may break some existing
13361 code that expects temporaries to be 80 bits.
13363 This is the default choice for the x86-64 compiler.
13368 Attempt to utilize both instruction sets at once. This effectively doubles the
13369 amount of available registers, and on chips with separate execution units for
13370 387 and SSE the execution resources too. Use this option with care, as it is
13371 still experimental, because the GCC register allocator does not model separate
13372 functional units well, resulting in unstable performance.
13375 @item -masm=@var{dialect}
13376 @opindex masm=@var{dialect}
13377 Output assembly instructions using selected @var{dialect}. Supported
13378 choices are @samp{intel} or @samp{att} (the default). Darwin does
13379 not support @samp{intel}.
13382 @itemx -mno-ieee-fp
13384 @opindex mno-ieee-fp
13385 Control whether or not the compiler uses IEEE floating-point
13386 comparisons. These correctly handle the case where the result of a
13387 comparison is unordered.
13390 @opindex msoft-float
13391 Generate output containing library calls for floating point.
13393 @strong{Warning:} the requisite libraries are not part of GCC@.
13394 Normally the facilities of the machine's usual C compiler are used, but
13395 this can't be done directly in cross-compilation. You must make your
13396 own arrangements to provide suitable library functions for
13399 On machines where a function returns floating-point results in the 80387
13400 register stack, some floating-point opcodes may be emitted even if
13401 @option{-msoft-float} is used.
13403 @item -mno-fp-ret-in-387
13404 @opindex mno-fp-ret-in-387
13405 Do not use the FPU registers for return values of functions.
13407 The usual calling convention has functions return values of types
13408 @code{float} and @code{double} in an FPU register, even if there
13409 is no FPU@. The idea is that the operating system should emulate
13412 The option @option{-mno-fp-ret-in-387} causes such values to be returned
13413 in ordinary CPU registers instead.
13415 @item -mno-fancy-math-387
13416 @opindex mno-fancy-math-387
13417 Some 387 emulators do not support the @code{sin}, @code{cos} and
13418 @code{sqrt} instructions for the 387. Specify this option to avoid
13419 generating those instructions. This option is the default on FreeBSD,
13420 OpenBSD and NetBSD@. This option is overridden when @option{-march}
13421 indicates that the target CPU always has an FPU and so the
13422 instruction does not need emulation. These
13423 instructions are not generated unless you also use the
13424 @option{-funsafe-math-optimizations} switch.
13426 @item -malign-double
13427 @itemx -mno-align-double
13428 @opindex malign-double
13429 @opindex mno-align-double
13430 Control whether GCC aligns @code{double}, @code{long double}, and
13431 @code{long long} variables on a two-word boundary or a one-word
13432 boundary. Aligning @code{double} variables on a two-word boundary
13433 produces code that runs somewhat faster on a Pentium at the
13434 expense of more memory.
13436 On x86-64, @option{-malign-double} is enabled by default.
13438 @strong{Warning:} if you use the @option{-malign-double} switch,
13439 structures containing the above types are aligned differently than
13440 the published application binary interface specifications for the 386
13441 and are not binary compatible with structures in code compiled
13442 without that switch.
13444 @item -m96bit-long-double
13445 @itemx -m128bit-long-double
13446 @opindex m96bit-long-double
13447 @opindex m128bit-long-double
13448 These switches control the size of @code{long double} type. The i386
13449 application binary interface specifies the size to be 96 bits,
13450 so @option{-m96bit-long-double} is the default in 32-bit mode.
13452 Modern architectures (Pentium and newer) prefer @code{long double}
13453 to be aligned to an 8- or 16-byte boundary. In arrays or structures
13454 conforming to the ABI, this is not possible. So specifying
13455 @option{-m128bit-long-double} aligns @code{long double}
13456 to a 16-byte boundary by padding the @code{long double} with an additional
13459 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
13460 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
13462 Notice that neither of these options enable any extra precision over the x87
13463 standard of 80 bits for a @code{long double}.
13465 @strong{Warning:} if you override the default value for your target ABI, this
13466 changes the size of
13467 structures and arrays containing @code{long double} variables,
13468 as well as modifying the function calling convention for functions taking
13469 @code{long double}. Hence they are not binary-compatible
13470 with code compiled without that switch.
13472 @item -mlarge-data-threshold=@var{threshold}
13473 @opindex mlarge-data-threshold
13474 When @option{-mcmodel=medium} is specified, data objects larger than
13475 @var{threshold} are placed in the large data section. This value must be the
13476 same across all objects linked into the binary, and defaults to 65535.
13480 Use a different function-calling convention, in which functions that
13481 take a fixed number of arguments return with the @code{ret @var{num}}
13482 instruction, which pops their arguments while returning. This saves one
13483 instruction in the caller since there is no need to pop the arguments
13486 You can specify that an individual function is called with this calling
13487 sequence with the function attribute @samp{stdcall}. You can also
13488 override the @option{-mrtd} option by using the function attribute
13489 @samp{cdecl}. @xref{Function Attributes}.
13491 @strong{Warning:} this calling convention is incompatible with the one
13492 normally used on Unix, so you cannot use it if you need to call
13493 libraries compiled with the Unix compiler.
13495 Also, you must provide function prototypes for all functions that
13496 take variable numbers of arguments (including @code{printf});
13497 otherwise incorrect code is generated for calls to those
13500 In addition, seriously incorrect code results if you call a
13501 function with too many arguments. (Normally, extra arguments are
13502 harmlessly ignored.)
13504 @item -mregparm=@var{num}
13506 Control how many registers are used to pass integer arguments. By
13507 default, no registers are used to pass arguments, and at most 3
13508 registers can be used. You can control this behavior for a specific
13509 function by using the function attribute @samp{regparm}.
13510 @xref{Function Attributes}.
13512 @strong{Warning:} if you use this switch, and
13513 @var{num} is nonzero, then you must build all modules with the same
13514 value, including any libraries. This includes the system libraries and
13518 @opindex msseregparm
13519 Use SSE register passing conventions for float and double arguments
13520 and return values. You can control this behavior for a specific
13521 function by using the function attribute @samp{sseregparm}.
13522 @xref{Function Attributes}.
13524 @strong{Warning:} if you use this switch then you must build all
13525 modules with the same value, including any libraries. This includes
13526 the system libraries and startup modules.
13528 @item -mvect8-ret-in-mem
13529 @opindex mvect8-ret-in-mem
13530 Return 8-byte vectors in memory instead of MMX registers. This is the
13531 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
13532 Studio compilers until version 12. Later compiler versions (starting
13533 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
13534 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
13535 you need to remain compatible with existing code produced by those
13536 previous compiler versions or older versions of GCC@.
13545 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
13546 is specified, the significands of results of floating-point operations are
13547 rounded to 24 bits (single precision); @option{-mpc64} rounds the
13548 significands of results of floating-point operations to 53 bits (double
13549 precision) and @option{-mpc80} rounds the significands of results of
13550 floating-point operations to 64 bits (extended double precision), which is
13551 the default. When this option is used, floating-point operations in higher
13552 precisions are not available to the programmer without setting the FPU
13553 control word explicitly.
13555 Setting the rounding of floating-point operations to less than the default
13556 80 bits can speed some programs by 2% or more. Note that some mathematical
13557 libraries assume that extended-precision (80-bit) floating-point operations
13558 are enabled by default; routines in such libraries could suffer significant
13559 loss of accuracy, typically through so-called ``catastrophic cancellation'',
13560 when this option is used to set the precision to less than extended precision.
13562 @item -mstackrealign
13563 @opindex mstackrealign
13564 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
13565 option generates an alternate prologue and epilogue that realigns the
13566 run-time stack if necessary. This supports mixing legacy codes that keep
13567 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
13568 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
13569 applicable to individual functions.
13571 @item -mpreferred-stack-boundary=@var{num}
13572 @opindex mpreferred-stack-boundary
13573 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
13574 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
13575 the default is 4 (16 bytes or 128 bits).
13577 @item -mincoming-stack-boundary=@var{num}
13578 @opindex mincoming-stack-boundary
13579 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
13580 boundary. If @option{-mincoming-stack-boundary} is not specified,
13581 the one specified by @option{-mpreferred-stack-boundary} is used.
13583 On Pentium and Pentium Pro, @code{double} and @code{long double} values
13584 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
13585 suffer significant run time performance penalties. On Pentium III, the
13586 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
13587 properly if it is not 16-byte aligned.
13589 To ensure proper alignment of this values on the stack, the stack boundary
13590 must be as aligned as that required by any value stored on the stack.
13591 Further, every function must be generated such that it keeps the stack
13592 aligned. Thus calling a function compiled with a higher preferred
13593 stack boundary from a function compiled with a lower preferred stack
13594 boundary most likely misaligns the stack. It is recommended that
13595 libraries that use callbacks always use the default setting.
13597 This extra alignment does consume extra stack space, and generally
13598 increases code size. Code that is sensitive to stack space usage, such
13599 as embedded systems and operating system kernels, may want to reduce the
13600 preferred alignment to @option{-mpreferred-stack-boundary=2}.
13629 @itemx -mno-fsgsbase
13666 These switches enable or disable the use of instructions in the MMX, SSE,
13667 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
13668 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT, RTM or 3DNow!@:
13669 extended instruction sets.
13670 These extensions are also available as built-in functions: see
13671 @ref{X86 Built-in Functions}, for details of the functions enabled and
13672 disabled by these switches.
13674 To generate SSE/SSE2 instructions automatically from floating-point
13675 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
13677 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
13678 generates new AVX instructions or AVX equivalence for all SSEx instructions
13681 These options enable GCC to use these extended instructions in
13682 generated code, even without @option{-mfpmath=sse}. Applications that
13683 perform run-time CPU detection must compile separate files for each
13684 supported architecture, using the appropriate flags. In particular,
13685 the file containing the CPU detection code should be compiled without
13690 This option instructs GCC to emit a @code{cld} instruction in the prologue
13691 of functions that use string instructions. String instructions depend on
13692 the DF flag to select between autoincrement or autodecrement mode. While the
13693 ABI specifies the DF flag to be cleared on function entry, some operating
13694 systems violate this specification by not clearing the DF flag in their
13695 exception dispatchers. The exception handler can be invoked with the DF flag
13696 set, which leads to wrong direction mode when string instructions are used.
13697 This option can be enabled by default on 32-bit x86 targets by configuring
13698 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
13699 instructions can be suppressed with the @option{-mno-cld} compiler option
13703 @opindex mvzeroupper
13704 This option instructs GCC to emit a @code{vzeroupper} instruction
13705 before a transfer of control flow out of the function to minimize
13706 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
13711 This option enables GCC to generate @code{CMPXCHG16B} instructions.
13712 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
13713 (or oword) data types.
13714 This is useful for high-resolution counters that can be updated
13715 by multiple processors (or cores). This instruction is generated as part of
13716 atomic built-in functions: see @ref{__sync Builtins} or
13717 @ref{__atomic Builtins} for details.
13721 This option enables generation of @code{SAHF} instructions in 64-bit code.
13722 Early Intel Pentium 4 CPUs with Intel 64 support,
13723 prior to the introduction of Pentium 4 G1 step in December 2005,
13724 lacked the @code{LAHF} and @code{SAHF} instructions
13725 which were supported by AMD64.
13726 These are load and store instructions, respectively, for certain status flags.
13727 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
13728 @code{drem}, and @code{remainder} built-in functions;
13729 see @ref{Other Builtins} for details.
13733 This option enables use of the @code{movbe} instruction to implement
13734 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
13738 This option enables built-in functions @code{__builtin_ia32_crc32qi},
13739 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
13740 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
13744 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
13745 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
13746 with an additional Newton-Raphson step
13747 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
13748 (and their vectorized
13749 variants) for single-precision floating-point arguments. These instructions
13750 are generated only when @option{-funsafe-math-optimizations} is enabled
13751 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
13752 Note that while the throughput of the sequence is higher than the throughput
13753 of the non-reciprocal instruction, the precision of the sequence can be
13754 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
13756 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
13757 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
13758 combination), and doesn't need @option{-mrecip}.
13760 Also note that GCC emits the above sequence with additional Newton-Raphson step
13761 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
13762 already with @option{-ffast-math} (or the above option combination), and
13763 doesn't need @option{-mrecip}.
13765 @item -mrecip=@var{opt}
13766 @opindex mrecip=opt
13767 This option controls which reciprocal estimate instructions
13768 may be used. @var{opt} is a comma-separated list of options, which may
13769 be preceded by a @samp{!} to invert the option:
13773 Enable all estimate instructions.
13776 Enable the default instructions, equivalent to @option{-mrecip}.
13779 Disable all estimate instructions, equivalent to @option{-mno-recip}.
13782 Enable the approximation for scalar division.
13785 Enable the approximation for vectorized division.
13788 Enable the approximation for scalar square root.
13791 Enable the approximation for vectorized square root.
13794 So, for example, @option{-mrecip=all,!sqrt} enables
13795 all of the reciprocal approximations, except for square root.
13797 @item -mveclibabi=@var{type}
13798 @opindex mveclibabi
13799 Specifies the ABI type to use for vectorizing intrinsics using an
13800 external library. Supported values for @var{type} are @samp{svml}
13801 for the Intel short
13802 vector math library and @samp{acml} for the AMD math core library.
13803 To use this option, both @option{-ftree-vectorize} and
13804 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
13805 ABI-compatible library must be specified at link time.
13807 GCC currently emits calls to @code{vmldExp2},
13808 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
13809 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
13810 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
13811 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
13812 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
13813 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
13814 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
13815 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
13816 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
13817 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
13818 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
13819 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
13820 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
13821 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
13822 when @option{-mveclibabi=acml} is used.
13824 @item -mabi=@var{name}
13826 Generate code for the specified calling convention. Permissible values
13827 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
13828 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
13829 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
13830 You can control this behavior for a specific function by
13831 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
13832 @xref{Function Attributes}.
13834 @item -mtls-dialect=@var{type}
13835 @opindex mtls-dialect
13836 Generate code to access thread-local storage using the @samp{gnu} or
13837 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
13838 @samp{gnu2} is more efficient, but it may add compile- and run-time
13839 requirements that cannot be satisfied on all systems.
13842 @itemx -mno-push-args
13843 @opindex mpush-args
13844 @opindex mno-push-args
13845 Use PUSH operations to store outgoing parameters. This method is shorter
13846 and usually equally fast as method using SUB/MOV operations and is enabled
13847 by default. In some cases disabling it may improve performance because of
13848 improved scheduling and reduced dependencies.
13850 @item -maccumulate-outgoing-args
13851 @opindex maccumulate-outgoing-args
13852 If enabled, the maximum amount of space required for outgoing arguments is
13853 computed in the function prologue. This is faster on most modern CPUs
13854 because of reduced dependencies, improved scheduling and reduced stack usage
13855 when the preferred stack boundary is not equal to 2. The drawback is a notable
13856 increase in code size. This switch implies @option{-mno-push-args}.
13860 Support thread-safe exception handling on MinGW. Programs that rely
13861 on thread-safe exception handling must compile and link all code with the
13862 @option{-mthreads} option. When compiling, @option{-mthreads} defines
13863 @code{-D_MT}; when linking, it links in a special thread helper library
13864 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
13866 @item -mno-align-stringops
13867 @opindex mno-align-stringops
13868 Do not align the destination of inlined string operations. This switch reduces
13869 code size and improves performance in case the destination is already aligned,
13870 but GCC doesn't know about it.
13872 @item -minline-all-stringops
13873 @opindex minline-all-stringops
13874 By default GCC inlines string operations only when the destination is
13875 known to be aligned to least a 4-byte boundary.
13876 This enables more inlining and increases code
13877 size, but may improve performance of code that depends on fast
13878 @code{memcpy}, @code{strlen},
13879 and @code{memset} for short lengths.
13881 @item -minline-stringops-dynamically
13882 @opindex minline-stringops-dynamically
13883 For string operations of unknown size, use run-time checks with
13884 inline code for small blocks and a library call for large blocks.
13886 @item -mstringop-strategy=@var{alg}
13887 @opindex mstringop-strategy=@var{alg}
13888 Override the internal decision heuristic for the particular algorithm to use
13889 for inlining string operations. The allowed values for @var{alg} are:
13895 Expand using i386 @code{rep} prefix of the specified size.
13899 @itemx unrolled_loop
13900 Expand into an inline loop.
13903 Always use a library call.
13906 @item -momit-leaf-frame-pointer
13907 @opindex momit-leaf-frame-pointer
13908 Don't keep the frame pointer in a register for leaf functions. This
13909 avoids the instructions to save, set up, and restore frame pointers and
13910 makes an extra register available in leaf functions. The option
13911 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
13912 which might make debugging harder.
13914 @item -mtls-direct-seg-refs
13915 @itemx -mno-tls-direct-seg-refs
13916 @opindex mtls-direct-seg-refs
13917 Controls whether TLS variables may be accessed with offsets from the
13918 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
13919 or whether the thread base pointer must be added. Whether or not this
13920 is valid depends on the operating system, and whether it maps the
13921 segment to cover the entire TLS area.
13923 For systems that use the GNU C Library, the default is on.
13926 @itemx -mno-sse2avx
13928 Specify that the assembler should encode SSE instructions with VEX
13929 prefix. The option @option{-mavx} turns this on by default.
13934 If profiling is active (@option{-pg}), put the profiling
13935 counter call before the prologue.
13936 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13937 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13940 @itemx -mno-8bit-idiv
13942 On some processors, like Intel Atom, 8-bit unsigned integer divide is
13943 much faster than 32-bit/64-bit integer divide. This option generates a
13944 run-time check. If both dividend and divisor are within range of 0
13945 to 255, 8-bit unsigned integer divide is used instead of
13946 32-bit/64-bit integer divide.
13948 @item -mavx256-split-unaligned-load
13949 @itemx -mavx256-split-unaligned-store
13950 @opindex avx256-split-unaligned-load
13951 @opindex avx256-split-unaligned-store
13952 Split 32-byte AVX unaligned load and store.
13956 These @samp{-m} switches are supported in addition to the above
13957 on x86-64 processors in 64-bit environments.
13966 Generate code for a 32-bit or 64-bit environment.
13967 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
13969 generates code that runs on any i386 system.
13971 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
13972 types to 64 bits, and generates code for the x86-64 architecture.
13973 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
13974 and @option{-mdynamic-no-pic} options.
13976 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
13978 generates code for the x86-64 architecture.
13980 @item -mno-red-zone
13981 @opindex mno-red-zone
13982 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
13983 by the x86-64 ABI; it is a 128-byte area beyond the location of the
13984 stack pointer that is not modified by signal or interrupt handlers
13985 and therefore can be used for temporary data without adjusting the stack
13986 pointer. The flag @option{-mno-red-zone} disables this red zone.
13988 @item -mcmodel=small
13989 @opindex mcmodel=small
13990 Generate code for the small code model: the program and its symbols must
13991 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13992 Programs can be statically or dynamically linked. This is the default
13995 @item -mcmodel=kernel
13996 @opindex mcmodel=kernel
13997 Generate code for the kernel code model. The kernel runs in the
13998 negative 2 GB of the address space.
13999 This model has to be used for Linux kernel code.
14001 @item -mcmodel=medium
14002 @opindex mcmodel=medium
14003 Generate code for the medium model: the program is linked in the lower 2
14004 GB of the address space. Small symbols are also placed there. Symbols
14005 with sizes larger than @option{-mlarge-data-threshold} are put into
14006 large data or BSS sections and can be located above 2GB. Programs can
14007 be statically or dynamically linked.
14009 @item -mcmodel=large
14010 @opindex mcmodel=large
14011 Generate code for the large model. This model makes no assumptions
14012 about addresses and sizes of sections.
14014 @item -maddress-mode=long
14015 @opindex maddress-mode=long
14016 Generate code for long address mode. This is only supported for 64-bit
14017 and x32 environments. It is the default address mode for 64-bit
14020 @item -maddress-mode=short
14021 @opindex maddress-mode=short
14022 Generate code for short address mode. This is only supported for 32-bit
14023 and x32 environments. It is the default address mode for 32-bit and
14027 @node i386 and x86-64 Windows Options
14028 @subsection i386 and x86-64 Windows Options
14029 @cindex i386 and x86-64 Windows Options
14031 These additional options are available for Microsoft Windows targets:
14037 specifies that a console application is to be generated, by
14038 instructing the linker to set the PE header subsystem type
14039 required for console applications.
14040 This option is available for Cygwin and MinGW targets and is
14041 enabled by default on those targets.
14045 This option is available for Cygwin and MinGW targets. It
14046 specifies that a DLL---a dynamic link library---is to be
14047 generated, enabling the selection of the required runtime
14048 startup object and entry point.
14050 @item -mnop-fun-dllimport
14051 @opindex mnop-fun-dllimport
14052 This option is available for Cygwin and MinGW targets. It
14053 specifies that the @code{dllimport} attribute should be ignored.
14057 This option is available for MinGW targets. It specifies
14058 that MinGW-specific thread support is to be used.
14062 This option is available for MinGW-w64 targets. It causes
14063 the @code{UNICODE} preprocessor macro to be predefined, and
14064 chooses Unicode-capable runtime startup code.
14068 This option is available for Cygwin and MinGW targets. It
14069 specifies that the typical Microsoft Windows predefined macros are to
14070 be set in the pre-processor, but does not influence the choice
14071 of runtime library/startup code.
14075 This option is available for Cygwin and MinGW targets. It
14076 specifies that a GUI application is to be generated by
14077 instructing the linker to set the PE header subsystem type
14080 @item -fno-set-stack-executable
14081 @opindex fno-set-stack-executable
14082 This option is available for MinGW targets. It specifies that
14083 the executable flag for the stack used by nested functions isn't
14084 set. This is necessary for binaries running in kernel mode of
14085 Microsoft Windows, as there the User32 API, which is used to set executable
14086 privileges, isn't available.
14088 @item -fwritable-relocated-rdata
14089 @opindex fno-writable-relocated-rdata
14090 This option is available for MinGW and Cygwin targets. It specifies
14091 that relocated-data in read-only section is put into .data
14092 section. This is a necessary for older runtimes not supporting
14093 modification of .rdata sections for pseudo-relocation.
14095 @item -mpe-aligned-commons
14096 @opindex mpe-aligned-commons
14097 This option is available for Cygwin and MinGW targets. It
14098 specifies that the GNU extension to the PE file format that
14099 permits the correct alignment of COMMON variables should be
14100 used when generating code. It is enabled by default if
14101 GCC detects that the target assembler found during configuration
14102 supports the feature.
14105 See also under @ref{i386 and x86-64 Options} for standard options.
14107 @node IA-64 Options
14108 @subsection IA-64 Options
14109 @cindex IA-64 Options
14111 These are the @samp{-m} options defined for the Intel IA-64 architecture.
14115 @opindex mbig-endian
14116 Generate code for a big-endian target. This is the default for HP-UX@.
14118 @item -mlittle-endian
14119 @opindex mlittle-endian
14120 Generate code for a little-endian target. This is the default for AIX5
14126 @opindex mno-gnu-as
14127 Generate (or don't) code for the GNU assembler. This is the default.
14128 @c Also, this is the default if the configure option @option{--with-gnu-as}
14134 @opindex mno-gnu-ld
14135 Generate (or don't) code for the GNU linker. This is the default.
14136 @c Also, this is the default if the configure option @option{--with-gnu-ld}
14141 Generate code that does not use a global pointer register. The result
14142 is not position independent code, and violates the IA-64 ABI@.
14144 @item -mvolatile-asm-stop
14145 @itemx -mno-volatile-asm-stop
14146 @opindex mvolatile-asm-stop
14147 @opindex mno-volatile-asm-stop
14148 Generate (or don't) a stop bit immediately before and after volatile asm
14151 @item -mregister-names
14152 @itemx -mno-register-names
14153 @opindex mregister-names
14154 @opindex mno-register-names
14155 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
14156 the stacked registers. This may make assembler output more readable.
14162 Disable (or enable) optimizations that use the small data section. This may
14163 be useful for working around optimizer bugs.
14165 @item -mconstant-gp
14166 @opindex mconstant-gp
14167 Generate code that uses a single constant global pointer value. This is
14168 useful when compiling kernel code.
14172 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
14173 This is useful when compiling firmware code.
14175 @item -minline-float-divide-min-latency
14176 @opindex minline-float-divide-min-latency
14177 Generate code for inline divides of floating-point values
14178 using the minimum latency algorithm.
14180 @item -minline-float-divide-max-throughput
14181 @opindex minline-float-divide-max-throughput
14182 Generate code for inline divides of floating-point values
14183 using the maximum throughput algorithm.
14185 @item -mno-inline-float-divide
14186 @opindex mno-inline-float-divide
14187 Do not generate inline code for divides of floating-point values.
14189 @item -minline-int-divide-min-latency
14190 @opindex minline-int-divide-min-latency
14191 Generate code for inline divides of integer values
14192 using the minimum latency algorithm.
14194 @item -minline-int-divide-max-throughput
14195 @opindex minline-int-divide-max-throughput
14196 Generate code for inline divides of integer values
14197 using the maximum throughput algorithm.
14199 @item -mno-inline-int-divide
14200 @opindex mno-inline-int-divide
14201 Do not generate inline code for divides of integer values.
14203 @item -minline-sqrt-min-latency
14204 @opindex minline-sqrt-min-latency
14205 Generate code for inline square roots
14206 using the minimum latency algorithm.
14208 @item -minline-sqrt-max-throughput
14209 @opindex minline-sqrt-max-throughput
14210 Generate code for inline square roots
14211 using the maximum throughput algorithm.
14213 @item -mno-inline-sqrt
14214 @opindex mno-inline-sqrt
14215 Do not generate inline code for @code{sqrt}.
14218 @itemx -mno-fused-madd
14219 @opindex mfused-madd
14220 @opindex mno-fused-madd
14221 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
14222 instructions. The default is to use these instructions.
14224 @item -mno-dwarf2-asm
14225 @itemx -mdwarf2-asm
14226 @opindex mno-dwarf2-asm
14227 @opindex mdwarf2-asm
14228 Don't (or do) generate assembler code for the DWARF 2 line number debugging
14229 info. This may be useful when not using the GNU assembler.
14231 @item -mearly-stop-bits
14232 @itemx -mno-early-stop-bits
14233 @opindex mearly-stop-bits
14234 @opindex mno-early-stop-bits
14235 Allow stop bits to be placed earlier than immediately preceding the
14236 instruction that triggered the stop bit. This can improve instruction
14237 scheduling, but does not always do so.
14239 @item -mfixed-range=@var{register-range}
14240 @opindex mfixed-range
14241 Generate code treating the given register range as fixed registers.
14242 A fixed register is one that the register allocator cannot use. This is
14243 useful when compiling kernel code. A register range is specified as
14244 two registers separated by a dash. Multiple register ranges can be
14245 specified separated by a comma.
14247 @item -mtls-size=@var{tls-size}
14249 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
14252 @item -mtune=@var{cpu-type}
14254 Tune the instruction scheduling for a particular CPU, Valid values are
14255 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
14256 and @samp{mckinley}.
14262 Generate code for a 32-bit or 64-bit environment.
14263 The 32-bit environment sets int, long and pointer to 32 bits.
14264 The 64-bit environment sets int to 32 bits and long and pointer
14265 to 64 bits. These are HP-UX specific flags.
14267 @item -mno-sched-br-data-spec
14268 @itemx -msched-br-data-spec
14269 @opindex mno-sched-br-data-spec
14270 @opindex msched-br-data-spec
14271 (Dis/En)able data speculative scheduling before reload.
14272 This results in generation of @code{ld.a} instructions and
14273 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
14274 The default is 'disable'.
14276 @item -msched-ar-data-spec
14277 @itemx -mno-sched-ar-data-spec
14278 @opindex msched-ar-data-spec
14279 @opindex mno-sched-ar-data-spec
14280 (En/Dis)able data speculative scheduling after reload.
14281 This results in generation of @code{ld.a} instructions and
14282 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
14283 The default is 'enable'.
14285 @item -mno-sched-control-spec
14286 @itemx -msched-control-spec
14287 @opindex mno-sched-control-spec
14288 @opindex msched-control-spec
14289 (Dis/En)able control speculative scheduling. This feature is
14290 available only during region scheduling (i.e.@: before reload).
14291 This results in generation of the @code{ld.s} instructions and
14292 the corresponding check instructions @code{chk.s}.
14293 The default is 'disable'.
14295 @item -msched-br-in-data-spec
14296 @itemx -mno-sched-br-in-data-spec
14297 @opindex msched-br-in-data-spec
14298 @opindex mno-sched-br-in-data-spec
14299 (En/Dis)able speculative scheduling of the instructions that
14300 are dependent on the data speculative loads before reload.
14301 This is effective only with @option{-msched-br-data-spec} enabled.
14302 The default is 'enable'.
14304 @item -msched-ar-in-data-spec
14305 @itemx -mno-sched-ar-in-data-spec
14306 @opindex msched-ar-in-data-spec
14307 @opindex mno-sched-ar-in-data-spec
14308 (En/Dis)able speculative scheduling of the instructions that
14309 are dependent on the data speculative loads after reload.
14310 This is effective only with @option{-msched-ar-data-spec} enabled.
14311 The default is 'enable'.
14313 @item -msched-in-control-spec
14314 @itemx -mno-sched-in-control-spec
14315 @opindex msched-in-control-spec
14316 @opindex mno-sched-in-control-spec
14317 (En/Dis)able speculative scheduling of the instructions that
14318 are dependent on the control speculative loads.
14319 This is effective only with @option{-msched-control-spec} enabled.
14320 The default is 'enable'.
14322 @item -mno-sched-prefer-non-data-spec-insns
14323 @itemx -msched-prefer-non-data-spec-insns
14324 @opindex mno-sched-prefer-non-data-spec-insns
14325 @opindex msched-prefer-non-data-spec-insns
14326 If enabled, data-speculative instructions are chosen for schedule
14327 only if there are no other choices at the moment. This makes
14328 the use of the data speculation much more conservative.
14329 The default is 'disable'.
14331 @item -mno-sched-prefer-non-control-spec-insns
14332 @itemx -msched-prefer-non-control-spec-insns
14333 @opindex mno-sched-prefer-non-control-spec-insns
14334 @opindex msched-prefer-non-control-spec-insns
14335 If enabled, control-speculative instructions are chosen for schedule
14336 only if there are no other choices at the moment. This makes
14337 the use of the control speculation much more conservative.
14338 The default is 'disable'.
14340 @item -mno-sched-count-spec-in-critical-path
14341 @itemx -msched-count-spec-in-critical-path
14342 @opindex mno-sched-count-spec-in-critical-path
14343 @opindex msched-count-spec-in-critical-path
14344 If enabled, speculative dependencies are considered during
14345 computation of the instructions priorities. This makes the use of the
14346 speculation a bit more conservative.
14347 The default is 'disable'.
14349 @item -msched-spec-ldc
14350 @opindex msched-spec-ldc
14351 Use a simple data speculation check. This option is on by default.
14353 @item -msched-control-spec-ldc
14354 @opindex msched-spec-ldc
14355 Use a simple check for control speculation. This option is on by default.
14357 @item -msched-stop-bits-after-every-cycle
14358 @opindex msched-stop-bits-after-every-cycle
14359 Place a stop bit after every cycle when scheduling. This option is on
14362 @item -msched-fp-mem-deps-zero-cost
14363 @opindex msched-fp-mem-deps-zero-cost
14364 Assume that floating-point stores and loads are not likely to cause a conflict
14365 when placed into the same instruction group. This option is disabled by
14368 @item -msel-sched-dont-check-control-spec
14369 @opindex msel-sched-dont-check-control-spec
14370 Generate checks for control speculation in selective scheduling.
14371 This flag is disabled by default.
14373 @item -msched-max-memory-insns=@var{max-insns}
14374 @opindex msched-max-memory-insns
14375 Limit on the number of memory insns per instruction group, giving lower
14376 priority to subsequent memory insns attempting to schedule in the same
14377 instruction group. Frequently useful to prevent cache bank conflicts.
14378 The default value is 1.
14380 @item -msched-max-memory-insns-hard-limit
14381 @opindex msched-max-memory-insns-hard-limit
14382 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
14383 disallowing more than that number in an instruction group.
14384 Otherwise, the limit is ``soft'', meaning that non-memory operations
14385 are preferred when the limit is reached, but memory operations may still
14391 @subsection LM32 Options
14392 @cindex LM32 options
14394 These @option{-m} options are defined for the Lattice Mico32 architecture:
14397 @item -mbarrel-shift-enabled
14398 @opindex mbarrel-shift-enabled
14399 Enable barrel-shift instructions.
14401 @item -mdivide-enabled
14402 @opindex mdivide-enabled
14403 Enable divide and modulus instructions.
14405 @item -mmultiply-enabled
14406 @opindex multiply-enabled
14407 Enable multiply instructions.
14409 @item -msign-extend-enabled
14410 @opindex msign-extend-enabled
14411 Enable sign extend instructions.
14413 @item -muser-enabled
14414 @opindex muser-enabled
14415 Enable user-defined instructions.
14420 @subsection M32C Options
14421 @cindex M32C options
14424 @item -mcpu=@var{name}
14426 Select the CPU for which code is generated. @var{name} may be one of
14427 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
14428 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
14429 the M32C/80 series.
14433 Specifies that the program will be run on the simulator. This causes
14434 an alternate runtime library to be linked in which supports, for
14435 example, file I/O@. You must not use this option when generating
14436 programs that will run on real hardware; you must provide your own
14437 runtime library for whatever I/O functions are needed.
14439 @item -memregs=@var{number}
14441 Specifies the number of memory-based pseudo-registers GCC uses
14442 during code generation. These pseudo-registers are used like real
14443 registers, so there is a tradeoff between GCC's ability to fit the
14444 code into available registers, and the performance penalty of using
14445 memory instead of registers. Note that all modules in a program must
14446 be compiled with the same value for this option. Because of that, you
14447 must not use this option with the default runtime libraries gcc
14452 @node M32R/D Options
14453 @subsection M32R/D Options
14454 @cindex M32R/D options
14456 These @option{-m} options are defined for Renesas M32R/D architectures:
14461 Generate code for the M32R/2@.
14465 Generate code for the M32R/X@.
14469 Generate code for the M32R@. This is the default.
14471 @item -mmodel=small
14472 @opindex mmodel=small
14473 Assume all objects live in the lower 16MB of memory (so that their addresses
14474 can be loaded with the @code{ld24} instruction), and assume all subroutines
14475 are reachable with the @code{bl} instruction.
14476 This is the default.
14478 The addressability of a particular object can be set with the
14479 @code{model} attribute.
14481 @item -mmodel=medium
14482 @opindex mmodel=medium
14483 Assume objects may be anywhere in the 32-bit address space (the compiler
14484 generates @code{seth/add3} instructions to load their addresses), and
14485 assume all subroutines are reachable with the @code{bl} instruction.
14487 @item -mmodel=large
14488 @opindex mmodel=large
14489 Assume objects may be anywhere in the 32-bit address space (the compiler
14490 generates @code{seth/add3} instructions to load their addresses), and
14491 assume subroutines may not be reachable with the @code{bl} instruction
14492 (the compiler generates the much slower @code{seth/add3/jl}
14493 instruction sequence).
14496 @opindex msdata=none
14497 Disable use of the small data area. Variables are put into
14498 one of @samp{.data}, @samp{.bss}, or @samp{.rodata} (unless the
14499 @code{section} attribute has been specified).
14500 This is the default.
14502 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
14503 Objects may be explicitly put in the small data area with the
14504 @code{section} attribute using one of these sections.
14506 @item -msdata=sdata
14507 @opindex msdata=sdata
14508 Put small global and static data in the small data area, but do not
14509 generate special code to reference them.
14512 @opindex msdata=use
14513 Put small global and static data in the small data area, and generate
14514 special instructions to reference them.
14518 @cindex smaller data references
14519 Put global and static objects less than or equal to @var{num} bytes
14520 into the small data or bss sections instead of the normal data or bss
14521 sections. The default value of @var{num} is 8.
14522 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
14523 for this option to have any effect.
14525 All modules should be compiled with the same @option{-G @var{num}} value.
14526 Compiling with different values of @var{num} may or may not work; if it
14527 doesn't the linker gives an error message---incorrect code is not
14532 Makes the M32R specific code in the compiler display some statistics
14533 that might help in debugging programs.
14535 @item -malign-loops
14536 @opindex malign-loops
14537 Align all loops to a 32-byte boundary.
14539 @item -mno-align-loops
14540 @opindex mno-align-loops
14541 Do not enforce a 32-byte alignment for loops. This is the default.
14543 @item -missue-rate=@var{number}
14544 @opindex missue-rate=@var{number}
14545 Issue @var{number} instructions per cycle. @var{number} can only be 1
14548 @item -mbranch-cost=@var{number}
14549 @opindex mbranch-cost=@var{number}
14550 @var{number} can only be 1 or 2. If it is 1 then branches are
14551 preferred over conditional code, if it is 2, then the opposite applies.
14553 @item -mflush-trap=@var{number}
14554 @opindex mflush-trap=@var{number}
14555 Specifies the trap number to use to flush the cache. The default is
14556 12. Valid numbers are between 0 and 15 inclusive.
14558 @item -mno-flush-trap
14559 @opindex mno-flush-trap
14560 Specifies that the cache cannot be flushed by using a trap.
14562 @item -mflush-func=@var{name}
14563 @opindex mflush-func=@var{name}
14564 Specifies the name of the operating system function to call to flush
14565 the cache. The default is @emph{_flush_cache}, but a function call
14566 is only used if a trap is not available.
14568 @item -mno-flush-func
14569 @opindex mno-flush-func
14570 Indicates that there is no OS function for flushing the cache.
14574 @node M680x0 Options
14575 @subsection M680x0 Options
14576 @cindex M680x0 options
14578 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
14579 The default settings depend on which architecture was selected when
14580 the compiler was configured; the defaults for the most common choices
14584 @item -march=@var{arch}
14586 Generate code for a specific M680x0 or ColdFire instruction set
14587 architecture. Permissible values of @var{arch} for M680x0
14588 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
14589 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
14590 architectures are selected according to Freescale's ISA classification
14591 and the permissible values are: @samp{isaa}, @samp{isaaplus},
14592 @samp{isab} and @samp{isac}.
14594 GCC defines a macro @samp{__mcf@var{arch}__} whenever it is generating
14595 code for a ColdFire target. The @var{arch} in this macro is one of the
14596 @option{-march} arguments given above.
14598 When used together, @option{-march} and @option{-mtune} select code
14599 that runs on a family of similar processors but that is optimized
14600 for a particular microarchitecture.
14602 @item -mcpu=@var{cpu}
14604 Generate code for a specific M680x0 or ColdFire processor.
14605 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
14606 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
14607 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
14608 below, which also classifies the CPUs into families:
14610 @multitable @columnfractions 0.20 0.80
14611 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
14612 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51ag} @samp{51cn} @samp{51em} @samp{51je} @samp{51jf} @samp{51jg} @samp{51jm} @samp{51mm} @samp{51qe} @samp{51qm}
14613 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
14614 @item @samp{5206e} @tab @samp{5206e}
14615 @item @samp{5208} @tab @samp{5207} @samp{5208}
14616 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
14617 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
14618 @item @samp{5216} @tab @samp{5214} @samp{5216}
14619 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
14620 @item @samp{5225} @tab @samp{5224} @samp{5225}
14621 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
14622 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
14623 @item @samp{5249} @tab @samp{5249}
14624 @item @samp{5250} @tab @samp{5250}
14625 @item @samp{5271} @tab @samp{5270} @samp{5271}
14626 @item @samp{5272} @tab @samp{5272}
14627 @item @samp{5275} @tab @samp{5274} @samp{5275}
14628 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
14629 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
14630 @item @samp{5307} @tab @samp{5307}
14631 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
14632 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
14633 @item @samp{5407} @tab @samp{5407}
14634 @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}
14637 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
14638 @var{arch} is compatible with @var{cpu}. Other combinations of
14639 @option{-mcpu} and @option{-march} are rejected.
14641 GCC defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
14642 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
14643 where the value of @var{family} is given by the table above.
14645 @item -mtune=@var{tune}
14647 Tune the code for a particular microarchitecture, within the
14648 constraints set by @option{-march} and @option{-mcpu}.
14649 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
14650 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
14651 and @samp{cpu32}. The ColdFire microarchitectures
14652 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
14654 You can also use @option{-mtune=68020-40} for code that needs
14655 to run relatively well on 68020, 68030 and 68040 targets.
14656 @option{-mtune=68020-60} is similar but includes 68060 targets
14657 as well. These two options select the same tuning decisions as
14658 @option{-m68020-40} and @option{-m68020-60} respectively.
14660 GCC defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
14661 when tuning for 680x0 architecture @var{arch}. It also defines
14662 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
14663 option is used. If GCC is tuning for a range of architectures,
14664 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
14665 it defines the macros for every architecture in the range.
14667 GCC also defines the macro @samp{__m@var{uarch}__} when tuning for
14668 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
14669 of the arguments given above.
14675 Generate output for a 68000. This is the default
14676 when the compiler is configured for 68000-based systems.
14677 It is equivalent to @option{-march=68000}.
14679 Use this option for microcontrollers with a 68000 or EC000 core,
14680 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
14684 Generate output for a 68010. This is the default
14685 when the compiler is configured for 68010-based systems.
14686 It is equivalent to @option{-march=68010}.
14692 Generate output for a 68020. This is the default
14693 when the compiler is configured for 68020-based systems.
14694 It is equivalent to @option{-march=68020}.
14698 Generate output for a 68030. This is the default when the compiler is
14699 configured for 68030-based systems. It is equivalent to
14700 @option{-march=68030}.
14704 Generate output for a 68040. This is the default when the compiler is
14705 configured for 68040-based systems. It is equivalent to
14706 @option{-march=68040}.
14708 This option inhibits the use of 68881/68882 instructions that have to be
14709 emulated by software on the 68040. Use this option if your 68040 does not
14710 have code to emulate those instructions.
14714 Generate output for a 68060. This is the default when the compiler is
14715 configured for 68060-based systems. It is equivalent to
14716 @option{-march=68060}.
14718 This option inhibits the use of 68020 and 68881/68882 instructions that
14719 have to be emulated by software on the 68060. Use this option if your 68060
14720 does not have code to emulate those instructions.
14724 Generate output for a CPU32. This is the default
14725 when the compiler is configured for CPU32-based systems.
14726 It is equivalent to @option{-march=cpu32}.
14728 Use this option for microcontrollers with a
14729 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
14730 68336, 68340, 68341, 68349 and 68360.
14734 Generate output for a 520X ColdFire CPU@. This is the default
14735 when the compiler is configured for 520X-based systems.
14736 It is equivalent to @option{-mcpu=5206}, and is now deprecated
14737 in favor of that option.
14739 Use this option for microcontroller with a 5200 core, including
14740 the MCF5202, MCF5203, MCF5204 and MCF5206.
14744 Generate output for a 5206e ColdFire CPU@. The option is now
14745 deprecated in favor of the equivalent @option{-mcpu=5206e}.
14749 Generate output for a member of the ColdFire 528X family.
14750 The option is now deprecated in favor of the equivalent
14751 @option{-mcpu=528x}.
14755 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
14756 in favor of the equivalent @option{-mcpu=5307}.
14760 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
14761 in favor of the equivalent @option{-mcpu=5407}.
14765 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
14766 This includes use of hardware floating-point instructions.
14767 The option is equivalent to @option{-mcpu=547x}, and is now
14768 deprecated in favor of that option.
14772 Generate output for a 68040, without using any of the new instructions.
14773 This results in code that can run relatively efficiently on either a
14774 68020/68881 or a 68030 or a 68040. The generated code does use the
14775 68881 instructions that are emulated on the 68040.
14777 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
14781 Generate output for a 68060, without using any of the new instructions.
14782 This results in code that can run relatively efficiently on either a
14783 68020/68881 or a 68030 or a 68040. The generated code does use the
14784 68881 instructions that are emulated on the 68060.
14786 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
14790 @opindex mhard-float
14792 Generate floating-point instructions. This is the default for 68020
14793 and above, and for ColdFire devices that have an FPU@. It defines the
14794 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
14795 on ColdFire targets.
14798 @opindex msoft-float
14799 Do not generate floating-point instructions; use library calls instead.
14800 This is the default for 68000, 68010, and 68832 targets. It is also
14801 the default for ColdFire devices that have no FPU.
14807 Generate (do not generate) ColdFire hardware divide and remainder
14808 instructions. If @option{-march} is used without @option{-mcpu},
14809 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
14810 architectures. Otherwise, the default is taken from the target CPU
14811 (either the default CPU, or the one specified by @option{-mcpu}). For
14812 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
14813 @option{-mcpu=5206e}.
14815 GCC defines the macro @samp{__mcfhwdiv__} when this option is enabled.
14819 Consider type @code{int} to be 16 bits wide, like @code{short int}.
14820 Additionally, parameters passed on the stack are also aligned to a
14821 16-bit boundary even on targets whose API mandates promotion to 32-bit.
14825 Do not consider type @code{int} to be 16 bits wide. This is the default.
14828 @itemx -mno-bitfield
14829 @opindex mnobitfield
14830 @opindex mno-bitfield
14831 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
14832 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
14836 Do use the bit-field instructions. The @option{-m68020} option implies
14837 @option{-mbitfield}. This is the default if you use a configuration
14838 designed for a 68020.
14842 Use a different function-calling convention, in which functions
14843 that take a fixed number of arguments return with the @code{rtd}
14844 instruction, which pops their arguments while returning. This
14845 saves one instruction in the caller since there is no need to pop
14846 the arguments there.
14848 This calling convention is incompatible with the one normally
14849 used on Unix, so you cannot use it if you need to call libraries
14850 compiled with the Unix compiler.
14852 Also, you must provide function prototypes for all functions that
14853 take variable numbers of arguments (including @code{printf});
14854 otherwise incorrect code is generated for calls to those
14857 In addition, seriously incorrect code results if you call a
14858 function with too many arguments. (Normally, extra arguments are
14859 harmlessly ignored.)
14861 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
14862 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
14866 Do not use the calling conventions selected by @option{-mrtd}.
14867 This is the default.
14870 @itemx -mno-align-int
14871 @opindex malign-int
14872 @opindex mno-align-int
14873 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
14874 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
14875 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
14876 Aligning variables on 32-bit boundaries produces code that runs somewhat
14877 faster on processors with 32-bit busses at the expense of more memory.
14879 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
14880 aligns structures containing the above types differently than
14881 most published application binary interface specifications for the m68k.
14885 Use the pc-relative addressing mode of the 68000 directly, instead of
14886 using a global offset table. At present, this option implies @option{-fpic},
14887 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
14888 not presently supported with @option{-mpcrel}, though this could be supported for
14889 68020 and higher processors.
14891 @item -mno-strict-align
14892 @itemx -mstrict-align
14893 @opindex mno-strict-align
14894 @opindex mstrict-align
14895 Do not (do) assume that unaligned memory references are handled by
14899 Generate code that allows the data segment to be located in a different
14900 area of memory from the text segment. This allows for execute in place in
14901 an environment without virtual memory management. This option implies
14904 @item -mno-sep-data
14905 Generate code that assumes that the data segment follows the text segment.
14906 This is the default.
14908 @item -mid-shared-library
14909 Generate code that supports shared libraries via the library ID method.
14910 This allows for execute in place and shared libraries in an environment
14911 without virtual memory management. This option implies @option{-fPIC}.
14913 @item -mno-id-shared-library
14914 Generate code that doesn't assume ID based shared libraries are being used.
14915 This is the default.
14917 @item -mshared-library-id=n
14918 Specifies the identification number of the ID-based shared library being
14919 compiled. Specifying a value of 0 generates more compact code; specifying
14920 other values forces the allocation of that number to the current
14921 library, but is no more space- or time-efficient than omitting this option.
14927 When generating position-independent code for ColdFire, generate code
14928 that works if the GOT has more than 8192 entries. This code is
14929 larger and slower than code generated without this option. On M680x0
14930 processors, this option is not needed; @option{-fPIC} suffices.
14932 GCC normally uses a single instruction to load values from the GOT@.
14933 While this is relatively efficient, it only works if the GOT
14934 is smaller than about 64k. Anything larger causes the linker
14935 to report an error such as:
14937 @cindex relocation truncated to fit (ColdFire)
14939 relocation truncated to fit: R_68K_GOT16O foobar
14942 If this happens, you should recompile your code with @option{-mxgot}.
14943 It should then work with very large GOTs. However, code generated with
14944 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14945 the value of a global symbol.
14947 Note that some linkers, including newer versions of the GNU linker,
14948 can create multiple GOTs and sort GOT entries. If you have such a linker,
14949 you should only need to use @option{-mxgot} when compiling a single
14950 object file that accesses more than 8192 GOT entries. Very few do.
14952 These options have no effect unless GCC is generating
14953 position-independent code.
14957 @node MCore Options
14958 @subsection MCore Options
14959 @cindex MCore options
14961 These are the @samp{-m} options defined for the Motorola M*Core
14967 @itemx -mno-hardlit
14969 @opindex mno-hardlit
14970 Inline constants into the code stream if it can be done in two
14971 instructions or less.
14977 Use the divide instruction. (Enabled by default).
14979 @item -mrelax-immediate
14980 @itemx -mno-relax-immediate
14981 @opindex mrelax-immediate
14982 @opindex mno-relax-immediate
14983 Allow arbitrary sized immediates in bit operations.
14985 @item -mwide-bitfields
14986 @itemx -mno-wide-bitfields
14987 @opindex mwide-bitfields
14988 @opindex mno-wide-bitfields
14989 Always treat bit-fields as int-sized.
14991 @item -m4byte-functions
14992 @itemx -mno-4byte-functions
14993 @opindex m4byte-functions
14994 @opindex mno-4byte-functions
14995 Force all functions to be aligned to a 4-byte boundary.
14997 @item -mcallgraph-data
14998 @itemx -mno-callgraph-data
14999 @opindex mcallgraph-data
15000 @opindex mno-callgraph-data
15001 Emit callgraph information.
15004 @itemx -mno-slow-bytes
15005 @opindex mslow-bytes
15006 @opindex mno-slow-bytes
15007 Prefer word access when reading byte quantities.
15009 @item -mlittle-endian
15010 @itemx -mbig-endian
15011 @opindex mlittle-endian
15012 @opindex mbig-endian
15013 Generate code for a little-endian target.
15019 Generate code for the 210 processor.
15023 Assume that runtime support has been provided and so omit the
15024 simulator library (@file{libsim.a)} from the linker command line.
15026 @item -mstack-increment=@var{size}
15027 @opindex mstack-increment
15028 Set the maximum amount for a single stack increment operation. Large
15029 values can increase the speed of programs that contain functions
15030 that need a large amount of stack space, but they can also trigger a
15031 segmentation fault if the stack is extended too much. The default
15037 @subsection MeP Options
15038 @cindex MeP options
15044 Enables the @code{abs} instruction, which is the absolute difference
15045 between two registers.
15049 Enables all the optional instructions---average, multiply, divide, bit
15050 operations, leading zero, absolute difference, min/max, clip, and
15056 Enables the @code{ave} instruction, which computes the average of two
15059 @item -mbased=@var{n}
15061 Variables of size @var{n} bytes or smaller are placed in the
15062 @code{.based} section by default. Based variables use the @code{$tp}
15063 register as a base register, and there is a 128-byte limit to the
15064 @code{.based} section.
15068 Enables the bit operation instructions---bit test (@code{btstm}), set
15069 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
15070 test-and-set (@code{tas}).
15072 @item -mc=@var{name}
15074 Selects which section constant data is placed in. @var{name} may
15075 be @code{tiny}, @code{near}, or @code{far}.
15079 Enables the @code{clip} instruction. Note that @code{-mclip} is not
15080 useful unless you also provide @code{-mminmax}.
15082 @item -mconfig=@var{name}
15084 Selects one of the build-in core configurations. Each MeP chip has
15085 one or more modules in it; each module has a core CPU and a variety of
15086 coprocessors, optional instructions, and peripherals. The
15087 @code{MeP-Integrator} tool, not part of GCC, provides these
15088 configurations through this option; using this option is the same as
15089 using all the corresponding command-line options. The default
15090 configuration is @code{default}.
15094 Enables the coprocessor instructions. By default, this is a 32-bit
15095 coprocessor. Note that the coprocessor is normally enabled via the
15096 @code{-mconfig=} option.
15100 Enables the 32-bit coprocessor's instructions.
15104 Enables the 64-bit coprocessor's instructions.
15108 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
15112 Causes constant variables to be placed in the @code{.near} section.
15116 Enables the @code{div} and @code{divu} instructions.
15120 Generate big-endian code.
15124 Generate little-endian code.
15126 @item -mio-volatile
15127 @opindex mio-volatile
15128 Tells the compiler that any variable marked with the @code{io}
15129 attribute is to be considered volatile.
15133 Causes variables to be assigned to the @code{.far} section by default.
15137 Enables the @code{leadz} (leading zero) instruction.
15141 Causes variables to be assigned to the @code{.near} section by default.
15145 Enables the @code{min} and @code{max} instructions.
15149 Enables the multiplication and multiply-accumulate instructions.
15153 Disables all the optional instructions enabled by @code{-mall-opts}.
15157 Enables the @code{repeat} and @code{erepeat} instructions, used for
15158 low-overhead looping.
15162 Causes all variables to default to the @code{.tiny} section. Note
15163 that there is a 65536-byte limit to this section. Accesses to these
15164 variables use the @code{%gp} base register.
15168 Enables the saturation instructions. Note that the compiler does not
15169 currently generate these itself, but this option is included for
15170 compatibility with other tools, like @code{as}.
15174 Link the SDRAM-based runtime instead of the default ROM-based runtime.
15178 Link the simulator runtime libraries.
15182 Link the simulator runtime libraries, excluding built-in support
15183 for reset and exception vectors and tables.
15187 Causes all functions to default to the @code{.far} section. Without
15188 this option, functions default to the @code{.near} section.
15190 @item -mtiny=@var{n}
15192 Variables that are @var{n} bytes or smaller are allocated to the
15193 @code{.tiny} section. These variables use the @code{$gp} base
15194 register. The default for this option is 4, but note that there's a
15195 65536-byte limit to the @code{.tiny} section.
15199 @node MicroBlaze Options
15200 @subsection MicroBlaze Options
15201 @cindex MicroBlaze Options
15206 @opindex msoft-float
15207 Use software emulation for floating point (default).
15210 @opindex mhard-float
15211 Use hardware floating-point instructions.
15215 Do not optimize block moves, use @code{memcpy}.
15217 @item -mno-clearbss
15218 @opindex mno-clearbss
15219 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
15221 @item -mcpu=@var{cpu-type}
15223 Use features of and schedule code for given CPU.
15224 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
15225 where @var{X} is a major version, @var{YY} is the minor version, and
15226 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
15227 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
15229 @item -mxl-soft-mul
15230 @opindex mxl-soft-mul
15231 Use software multiply emulation (default).
15233 @item -mxl-soft-div
15234 @opindex mxl-soft-div
15235 Use software emulation for divides (default).
15237 @item -mxl-barrel-shift
15238 @opindex mxl-barrel-shift
15239 Use the hardware barrel shifter.
15241 @item -mxl-pattern-compare
15242 @opindex mxl-pattern-compare
15243 Use pattern compare instructions.
15245 @item -msmall-divides
15246 @opindex msmall-divides
15247 Use table lookup optimization for small signed integer divisions.
15249 @item -mxl-stack-check
15250 @opindex mxl-stack-check
15251 This option is deprecated. Use @option{-fstack-check} instead.
15254 @opindex mxl-gp-opt
15255 Use GP-relative @code{.sdata}/@code{.sbss} sections.
15257 @item -mxl-multiply-high
15258 @opindex mxl-multiply-high
15259 Use multiply high instructions for high part of 32x32 multiply.
15261 @item -mxl-float-convert
15262 @opindex mxl-float-convert
15263 Use hardware floating-point conversion instructions.
15265 @item -mxl-float-sqrt
15266 @opindex mxl-float-sqrt
15267 Use hardware floating-point square root instruction.
15269 @item -mxl-mode-@var{app-model}
15270 Select application model @var{app-model}. Valid models are
15273 normal executable (default), uses startup code @file{crt0.o}.
15276 for use with Xilinx Microprocessor Debugger (XMD) based
15277 software intrusive debug agent called xmdstub. This uses startup file
15278 @file{crt1.o} and sets the start address of the program to be 0x800.
15281 for applications that are loaded using a bootloader.
15282 This model uses startup file @file{crt2.o} which does not contain a processor
15283 reset vector handler. This is suitable for transferring control on a
15284 processor reset to the bootloader rather than the application.
15287 for applications that do not require any of the
15288 MicroBlaze vectors. This option may be useful for applications running
15289 within a monitoring application. This model uses @file{crt3.o} as a startup file.
15292 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
15293 @option{-mxl-mode-@var{app-model}}.
15298 @subsection MIPS Options
15299 @cindex MIPS options
15305 Generate big-endian code.
15309 Generate little-endian code. This is the default for @samp{mips*el-*-*}
15312 @item -march=@var{arch}
15314 Generate code that runs on @var{arch}, which can be the name of a
15315 generic MIPS ISA, or the name of a particular processor.
15317 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
15318 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
15319 The processor names are:
15320 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
15321 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
15322 @samp{5kc}, @samp{5kf},
15324 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
15325 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
15326 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
15327 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
15328 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
15329 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
15331 @samp{octeon}, @samp{octeon+}, @samp{octeon2},
15333 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
15334 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
15335 @samp{rm7000}, @samp{rm9000},
15336 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
15339 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
15340 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
15341 @samp{xlr} and @samp{xlp}.
15342 The special value @samp{from-abi} selects the
15343 most compatible architecture for the selected ABI (that is,
15344 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
15346 The native Linux/GNU toolchain also supports the value @samp{native},
15347 which selects the best architecture option for the host processor.
15348 @option{-march=native} has no effect if GCC does not recognize
15351 In processor names, a final @samp{000} can be abbreviated as @samp{k}
15352 (for example, @option{-march=r2k}). Prefixes are optional, and
15353 @samp{vr} may be written @samp{r}.
15355 Names of the form @samp{@var{n}f2_1} refer to processors with
15356 FPUs clocked at half the rate of the core, names of the form
15357 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
15358 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
15359 processors with FPUs clocked a ratio of 3:2 with respect to the core.
15360 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
15361 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
15362 accepted as synonyms for @samp{@var{n}f1_1}.
15364 GCC defines two macros based on the value of this option. The first
15365 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
15366 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
15367 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
15368 For example, @option{-march=r2000} sets @samp{_MIPS_ARCH}
15369 to @samp{"r2000"} and defines the macro @samp{_MIPS_ARCH_R2000}.
15371 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
15372 above. In other words, it has the full prefix and does not
15373 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
15374 the macro names the resolved architecture (either @samp{"mips1"} or
15375 @samp{"mips3"}). It names the default architecture when no
15376 @option{-march} option is given.
15378 @item -mtune=@var{arch}
15380 Optimize for @var{arch}. Among other things, this option controls
15381 the way instructions are scheduled, and the perceived cost of arithmetic
15382 operations. The list of @var{arch} values is the same as for
15385 When this option is not used, GCC optimizes for the processor
15386 specified by @option{-march}. By using @option{-march} and
15387 @option{-mtune} together, it is possible to generate code that
15388 runs on a family of processors, but optimize the code for one
15389 particular member of that family.
15391 @option{-mtune} defines the macros @samp{_MIPS_TUNE} and
15392 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
15393 @option{-march} ones described above.
15397 Equivalent to @option{-march=mips1}.
15401 Equivalent to @option{-march=mips2}.
15405 Equivalent to @option{-march=mips3}.
15409 Equivalent to @option{-march=mips4}.
15413 Equivalent to @option{-march=mips32}.
15417 Equivalent to @option{-march=mips32r2}.
15421 Equivalent to @option{-march=mips64}.
15425 Equivalent to @option{-march=mips64r2}.
15430 @opindex mno-mips16
15431 Generate (do not generate) MIPS16 code. If GCC is targeting a
15432 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
15434 MIPS16 code generation can also be controlled on a per-function basis
15435 by means of @code{mips16} and @code{nomips16} attributes.
15436 @xref{Function Attributes}, for more information.
15438 @item -mflip-mips16
15439 @opindex mflip-mips16
15440 Generate MIPS16 code on alternating functions. This option is provided
15441 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
15442 not intended for ordinary use in compiling user code.
15444 @item -minterlink-mips16
15445 @itemx -mno-interlink-mips16
15446 @opindex minterlink-mips16
15447 @opindex mno-interlink-mips16
15448 Require (do not require) that non-MIPS16 code be link-compatible with
15451 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
15452 it must either use a call or an indirect jump. @option{-minterlink-mips16}
15453 therefore disables direct jumps unless GCC knows that the target of the
15454 jump is not MIPS16.
15466 Generate code for the given ABI@.
15468 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
15469 generates 64-bit code when you select a 64-bit architecture, but you
15470 can use @option{-mgp32} to get 32-bit code instead.
15472 For information about the O64 ABI, see
15473 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
15475 GCC supports a variant of the o32 ABI in which floating-point registers
15476 are 64 rather than 32 bits wide. You can select this combination with
15477 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
15478 and @code{mfhc1} instructions and is therefore only supported for
15479 MIPS32R2 processors.
15481 The register assignments for arguments and return values remain the
15482 same, but each scalar value is passed in a single 64-bit register
15483 rather than a pair of 32-bit registers. For example, scalar
15484 floating-point values are returned in @samp{$f0} only, not a
15485 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
15486 remains the same, but all 64 bits are saved.
15489 @itemx -mno-abicalls
15491 @opindex mno-abicalls
15492 Generate (do not generate) code that is suitable for SVR4-style
15493 dynamic objects. @option{-mabicalls} is the default for SVR4-based
15498 Generate (do not generate) code that is fully position-independent,
15499 and that can therefore be linked into shared libraries. This option
15500 only affects @option{-mabicalls}.
15502 All @option{-mabicalls} code has traditionally been position-independent,
15503 regardless of options like @option{-fPIC} and @option{-fpic}. However,
15504 as an extension, the GNU toolchain allows executables to use absolute
15505 accesses for locally-binding symbols. It can also use shorter GP
15506 initialization sequences and generate direct calls to locally-defined
15507 functions. This mode is selected by @option{-mno-shared}.
15509 @option{-mno-shared} depends on binutils 2.16 or higher and generates
15510 objects that can only be linked by the GNU linker. However, the option
15511 does not affect the ABI of the final executable; it only affects the ABI
15512 of relocatable objects. Using @option{-mno-shared} generally makes
15513 executables both smaller and quicker.
15515 @option{-mshared} is the default.
15521 Assume (do not assume) that the static and dynamic linkers
15522 support PLTs and copy relocations. This option only affects
15523 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
15524 has no effect without @option{-msym32}.
15526 You can make @option{-mplt} the default by configuring
15527 GCC with @option{--with-mips-plt}. The default is
15528 @option{-mno-plt} otherwise.
15534 Lift (do not lift) the usual restrictions on the size of the global
15537 GCC normally uses a single instruction to load values from the GOT@.
15538 While this is relatively efficient, it only works if the GOT
15539 is smaller than about 64k. Anything larger causes the linker
15540 to report an error such as:
15542 @cindex relocation truncated to fit (MIPS)
15544 relocation truncated to fit: R_MIPS_GOT16 foobar
15547 If this happens, you should recompile your code with @option{-mxgot}.
15548 This works with very large GOTs, although the code is also
15549 less efficient, since it takes three instructions to fetch the
15550 value of a global symbol.
15552 Note that some linkers can create multiple GOTs. If you have such a
15553 linker, you should only need to use @option{-mxgot} when a single object
15554 file accesses more than 64k's worth of GOT entries. Very few do.
15556 These options have no effect unless GCC is generating position
15561 Assume that general-purpose registers are 32 bits wide.
15565 Assume that general-purpose registers are 64 bits wide.
15569 Assume that floating-point registers are 32 bits wide.
15573 Assume that floating-point registers are 64 bits wide.
15576 @opindex mhard-float
15577 Use floating-point coprocessor instructions.
15580 @opindex msoft-float
15581 Do not use floating-point coprocessor instructions. Implement
15582 floating-point calculations using library calls instead.
15584 @item -msingle-float
15585 @opindex msingle-float
15586 Assume that the floating-point coprocessor only supports single-precision
15589 @item -mdouble-float
15590 @opindex mdouble-float
15591 Assume that the floating-point coprocessor supports double-precision
15592 operations. This is the default.
15598 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
15599 implement atomic memory built-in functions. When neither option is
15600 specified, GCC uses the instructions if the target architecture
15603 @option{-mllsc} is useful if the runtime environment can emulate the
15604 instructions and @option{-mno-llsc} can be useful when compiling for
15605 nonstandard ISAs. You can make either option the default by
15606 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
15607 respectively. @option{--with-llsc} is the default for some
15608 configurations; see the installation documentation for details.
15614 Use (do not use) revision 1 of the MIPS DSP ASE@.
15615 @xref{MIPS DSP Built-in Functions}. This option defines the
15616 preprocessor macro @samp{__mips_dsp}. It also defines
15617 @samp{__mips_dsp_rev} to 1.
15623 Use (do not use) revision 2 of the MIPS DSP ASE@.
15624 @xref{MIPS DSP Built-in Functions}. This option defines the
15625 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
15626 It also defines @samp{__mips_dsp_rev} to 2.
15629 @itemx -mno-smartmips
15630 @opindex msmartmips
15631 @opindex mno-smartmips
15632 Use (do not use) the MIPS SmartMIPS ASE.
15634 @item -mpaired-single
15635 @itemx -mno-paired-single
15636 @opindex mpaired-single
15637 @opindex mno-paired-single
15638 Use (do not use) paired-single floating-point instructions.
15639 @xref{MIPS Paired-Single Support}. This option requires
15640 hardware floating-point support to be enabled.
15646 Use (do not use) MIPS Digital Media Extension instructions.
15647 This option can only be used when generating 64-bit code and requires
15648 hardware floating-point support to be enabled.
15653 @opindex mno-mips3d
15654 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
15655 The option @option{-mips3d} implies @option{-mpaired-single}.
15661 Use (do not use) MT Multithreading instructions.
15665 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
15666 an explanation of the default and the way that the pointer size is
15671 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
15673 The default size of @code{int}s, @code{long}s and pointers depends on
15674 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
15675 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
15676 32-bit @code{long}s. Pointers are the same size as @code{long}s,
15677 or the same size as integer registers, whichever is smaller.
15683 Assume (do not assume) that all symbols have 32-bit values, regardless
15684 of the selected ABI@. This option is useful in combination with
15685 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
15686 to generate shorter and faster references to symbolic addresses.
15690 Put definitions of externally-visible data in a small data section
15691 if that data is no bigger than @var{num} bytes. GCC can then access
15692 the data more efficiently; see @option{-mgpopt} for details.
15694 The default @option{-G} option depends on the configuration.
15696 @item -mlocal-sdata
15697 @itemx -mno-local-sdata
15698 @opindex mlocal-sdata
15699 @opindex mno-local-sdata
15700 Extend (do not extend) the @option{-G} behavior to local data too,
15701 such as to static variables in C@. @option{-mlocal-sdata} is the
15702 default for all configurations.
15704 If the linker complains that an application is using too much small data,
15705 you might want to try rebuilding the less performance-critical parts with
15706 @option{-mno-local-sdata}. You might also want to build large
15707 libraries with @option{-mno-local-sdata}, so that the libraries leave
15708 more room for the main program.
15710 @item -mextern-sdata
15711 @itemx -mno-extern-sdata
15712 @opindex mextern-sdata
15713 @opindex mno-extern-sdata
15714 Assume (do not assume) that externally-defined data is in
15715 a small data section if the size of that data is within the @option{-G} limit.
15716 @option{-mextern-sdata} is the default for all configurations.
15718 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
15719 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
15720 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
15721 is placed in a small data section. If @var{Var} is defined by another
15722 module, you must either compile that module with a high-enough
15723 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
15724 definition. If @var{Var} is common, you must link the application
15725 with a high-enough @option{-G} setting.
15727 The easiest way of satisfying these restrictions is to compile
15728 and link every module with the same @option{-G} option. However,
15729 you may wish to build a library that supports several different
15730 small data limits. You can do this by compiling the library with
15731 the highest supported @option{-G} setting and additionally using
15732 @option{-mno-extern-sdata} to stop the library from making assumptions
15733 about externally-defined data.
15739 Use (do not use) GP-relative accesses for symbols that are known to be
15740 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
15741 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
15744 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
15745 might not hold the value of @code{_gp}. For example, if the code is
15746 part of a library that might be used in a boot monitor, programs that
15747 call boot monitor routines pass an unknown value in @code{$gp}.
15748 (In such situations, the boot monitor itself is usually compiled
15749 with @option{-G0}.)
15751 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
15752 @option{-mno-extern-sdata}.
15754 @item -membedded-data
15755 @itemx -mno-embedded-data
15756 @opindex membedded-data
15757 @opindex mno-embedded-data
15758 Allocate variables to the read-only data section first if possible, then
15759 next in the small data section if possible, otherwise in data. This gives
15760 slightly slower code than the default, but reduces the amount of RAM required
15761 when executing, and thus may be preferred for some embedded systems.
15763 @item -muninit-const-in-rodata
15764 @itemx -mno-uninit-const-in-rodata
15765 @opindex muninit-const-in-rodata
15766 @opindex mno-uninit-const-in-rodata
15767 Put uninitialized @code{const} variables in the read-only data section.
15768 This option is only meaningful in conjunction with @option{-membedded-data}.
15770 @item -mcode-readable=@var{setting}
15771 @opindex mcode-readable
15772 Specify whether GCC may generate code that reads from executable sections.
15773 There are three possible settings:
15776 @item -mcode-readable=yes
15777 Instructions may freely access executable sections. This is the
15780 @item -mcode-readable=pcrel
15781 MIPS16 PC-relative load instructions can access executable sections,
15782 but other instructions must not do so. This option is useful on 4KSc
15783 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
15784 It is also useful on processors that can be configured to have a dual
15785 instruction/data SRAM interface and that, like the M4K, automatically
15786 redirect PC-relative loads to the instruction RAM.
15788 @item -mcode-readable=no
15789 Instructions must not access executable sections. This option can be
15790 useful on targets that are configured to have a dual instruction/data
15791 SRAM interface but that (unlike the M4K) do not automatically redirect
15792 PC-relative loads to the instruction RAM.
15795 @item -msplit-addresses
15796 @itemx -mno-split-addresses
15797 @opindex msplit-addresses
15798 @opindex mno-split-addresses
15799 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
15800 relocation operators. This option has been superseded by
15801 @option{-mexplicit-relocs} but is retained for backwards compatibility.
15803 @item -mexplicit-relocs
15804 @itemx -mno-explicit-relocs
15805 @opindex mexplicit-relocs
15806 @opindex mno-explicit-relocs
15807 Use (do not use) assembler relocation operators when dealing with symbolic
15808 addresses. The alternative, selected by @option{-mno-explicit-relocs},
15809 is to use assembler macros instead.
15811 @option{-mexplicit-relocs} is the default if GCC was configured
15812 to use an assembler that supports relocation operators.
15814 @item -mcheck-zero-division
15815 @itemx -mno-check-zero-division
15816 @opindex mcheck-zero-division
15817 @opindex mno-check-zero-division
15818 Trap (do not trap) on integer division by zero.
15820 The default is @option{-mcheck-zero-division}.
15822 @item -mdivide-traps
15823 @itemx -mdivide-breaks
15824 @opindex mdivide-traps
15825 @opindex mdivide-breaks
15826 MIPS systems check for division by zero by generating either a
15827 conditional trap or a break instruction. Using traps results in
15828 smaller code, but is only supported on MIPS II and later. Also, some
15829 versions of the Linux kernel have a bug that prevents trap from
15830 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
15831 allow conditional traps on architectures that support them and
15832 @option{-mdivide-breaks} to force the use of breaks.
15834 The default is usually @option{-mdivide-traps}, but this can be
15835 overridden at configure time using @option{--with-divide=breaks}.
15836 Divide-by-zero checks can be completely disabled using
15837 @option{-mno-check-zero-division}.
15842 @opindex mno-memcpy
15843 Force (do not force) the use of @code{memcpy()} for non-trivial block
15844 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
15845 most constant-sized copies.
15848 @itemx -mno-long-calls
15849 @opindex mlong-calls
15850 @opindex mno-long-calls
15851 Disable (do not disable) use of the @code{jal} instruction. Calling
15852 functions using @code{jal} is more efficient but requires the caller
15853 and callee to be in the same 256 megabyte segment.
15855 This option has no effect on abicalls code. The default is
15856 @option{-mno-long-calls}.
15862 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
15863 instructions, as provided by the R4650 ISA@.
15866 @itemx -mno-fused-madd
15867 @opindex mfused-madd
15868 @opindex mno-fused-madd
15869 Enable (disable) use of the floating-point multiply-accumulate
15870 instructions, when they are available. The default is
15871 @option{-mfused-madd}.
15873 When multiply-accumulate instructions are used, the intermediate
15874 product is calculated to infinite precision and is not subject to
15875 the FCSR Flush to Zero bit. This may be undesirable in some
15880 Tell the MIPS assembler to not run its preprocessor over user
15881 assembler files (with a @samp{.s} suffix) when assembling them.
15886 @opindex mno-fix-24k
15887 Work around the 24K E48 (lost data on stores during refill) errata.
15888 The workarounds are implemented by the assembler rather than by GCC@.
15891 @itemx -mno-fix-r4000
15892 @opindex mfix-r4000
15893 @opindex mno-fix-r4000
15894 Work around certain R4000 CPU errata:
15897 A double-word or a variable shift may give an incorrect result if executed
15898 immediately after starting an integer division.
15900 A double-word or a variable shift may give an incorrect result if executed
15901 while an integer multiplication is in progress.
15903 An integer division may give an incorrect result if started in a delay slot
15904 of a taken branch or a jump.
15908 @itemx -mno-fix-r4400
15909 @opindex mfix-r4400
15910 @opindex mno-fix-r4400
15911 Work around certain R4400 CPU errata:
15914 A double-word or a variable shift may give an incorrect result if executed
15915 immediately after starting an integer division.
15919 @itemx -mno-fix-r10000
15920 @opindex mfix-r10000
15921 @opindex mno-fix-r10000
15922 Work around certain R10000 errata:
15925 @code{ll}/@code{sc} sequences may not behave atomically on revisions
15926 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15929 This option can only be used if the target architecture supports
15930 branch-likely instructions. @option{-mfix-r10000} is the default when
15931 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15935 @itemx -mno-fix-vr4120
15936 @opindex mfix-vr4120
15937 Work around certain VR4120 errata:
15940 @code{dmultu} does not always produce the correct result.
15942 @code{div} and @code{ddiv} do not always produce the correct result if one
15943 of the operands is negative.
15945 The workarounds for the division errata rely on special functions in
15946 @file{libgcc.a}. At present, these functions are only provided by
15947 the @code{mips64vr*-elf} configurations.
15949 Other VR4120 errata require a nop to be inserted between certain pairs of
15950 instructions. These errata are handled by the assembler, not by GCC itself.
15953 @opindex mfix-vr4130
15954 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15955 workarounds are implemented by the assembler rather than by GCC,
15956 although GCC avoids using @code{mflo} and @code{mfhi} if the
15957 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15958 instructions are available instead.
15961 @itemx -mno-fix-sb1
15963 Work around certain SB-1 CPU core errata.
15964 (This flag currently works around the SB-1 revision 2
15965 ``F1'' and ``F2'' floating-point errata.)
15967 @item -mr10k-cache-barrier=@var{setting}
15968 @opindex mr10k-cache-barrier
15969 Specify whether GCC should insert cache barriers to avoid the
15970 side-effects of speculation on R10K processors.
15972 In common with many processors, the R10K tries to predict the outcome
15973 of a conditional branch and speculatively executes instructions from
15974 the ``taken'' branch. It later aborts these instructions if the
15975 predicted outcome is wrong. However, on the R10K, even aborted
15976 instructions can have side effects.
15978 This problem only affects kernel stores and, depending on the system,
15979 kernel loads. As an example, a speculatively-executed store may load
15980 the target memory into cache and mark the cache line as dirty, even if
15981 the store itself is later aborted. If a DMA operation writes to the
15982 same area of memory before the ``dirty'' line is flushed, the cached
15983 data overwrites the DMA-ed data. See the R10K processor manual
15984 for a full description, including other potential problems.
15986 One workaround is to insert cache barrier instructions before every memory
15987 access that might be speculatively executed and that might have side
15988 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15989 controls GCC's implementation of this workaround. It assumes that
15990 aborted accesses to any byte in the following regions does not have
15995 the memory occupied by the current function's stack frame;
15998 the memory occupied by an incoming stack argument;
16001 the memory occupied by an object with a link-time-constant address.
16004 It is the kernel's responsibility to ensure that speculative
16005 accesses to these regions are indeed safe.
16007 If the input program contains a function declaration such as:
16013 then the implementation of @code{foo} must allow @code{j foo} and
16014 @code{jal foo} to be executed speculatively. GCC honors this
16015 restriction for functions it compiles itself. It expects non-GCC
16016 functions (such as hand-written assembly code) to do the same.
16018 The option has three forms:
16021 @item -mr10k-cache-barrier=load-store
16022 Insert a cache barrier before a load or store that might be
16023 speculatively executed and that might have side effects even
16026 @item -mr10k-cache-barrier=store
16027 Insert a cache barrier before a store that might be speculatively
16028 executed and that might have side effects even if aborted.
16030 @item -mr10k-cache-barrier=none
16031 Disable the insertion of cache barriers. This is the default setting.
16034 @item -mflush-func=@var{func}
16035 @itemx -mno-flush-func
16036 @opindex mflush-func
16037 Specifies the function to call to flush the I and D caches, or to not
16038 call any such function. If called, the function must take the same
16039 arguments as the common @code{_flush_func()}, that is, the address of the
16040 memory range for which the cache is being flushed, the size of the
16041 memory range, and the number 3 (to flush both caches). The default
16042 depends on the target GCC was configured for, but commonly is either
16043 @samp{_flush_func} or @samp{__cpu_flush}.
16045 @item mbranch-cost=@var{num}
16046 @opindex mbranch-cost
16047 Set the cost of branches to roughly @var{num} ``simple'' instructions.
16048 This cost is only a heuristic and is not guaranteed to produce
16049 consistent results across releases. A zero cost redundantly selects
16050 the default, which is based on the @option{-mtune} setting.
16052 @item -mbranch-likely
16053 @itemx -mno-branch-likely
16054 @opindex mbranch-likely
16055 @opindex mno-branch-likely
16056 Enable or disable use of Branch Likely instructions, regardless of the
16057 default for the selected architecture. By default, Branch Likely
16058 instructions may be generated if they are supported by the selected
16059 architecture. An exception is for the MIPS32 and MIPS64 architectures
16060 and processors that implement those architectures; for those, Branch
16061 Likely instructions are not be generated by default because the MIPS32
16062 and MIPS64 architectures specifically deprecate their use.
16064 @item -mfp-exceptions
16065 @itemx -mno-fp-exceptions
16066 @opindex mfp-exceptions
16067 Specifies whether FP exceptions are enabled. This affects how
16068 FP instructions are scheduled for some processors.
16069 The default is that FP exceptions are
16072 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
16073 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
16076 @item -mvr4130-align
16077 @itemx -mno-vr4130-align
16078 @opindex mvr4130-align
16079 The VR4130 pipeline is two-way superscalar, but can only issue two
16080 instructions together if the first one is 8-byte aligned. When this
16081 option is enabled, GCC aligns pairs of instructions that it
16082 thinks should execute in parallel.
16084 This option only has an effect when optimizing for the VR4130.
16085 It normally makes code faster, but at the expense of making it bigger.
16086 It is enabled by default at optimization level @option{-O3}.
16091 Enable (disable) generation of @code{synci} instructions on
16092 architectures that support it. The @code{synci} instructions (if
16093 enabled) are generated when @code{__builtin___clear_cache()} is
16096 This option defaults to @code{-mno-synci}, but the default can be
16097 overridden by configuring with @code{--with-synci}.
16099 When compiling code for single processor systems, it is generally safe
16100 to use @code{synci}. However, on many multi-core (SMP) systems, it
16101 does not invalidate the instruction caches on all cores and may lead
16102 to undefined behavior.
16104 @item -mrelax-pic-calls
16105 @itemx -mno-relax-pic-calls
16106 @opindex mrelax-pic-calls
16107 Try to turn PIC calls that are normally dispatched via register
16108 @code{$25} into direct calls. This is only possible if the linker can
16109 resolve the destination at link-time and if the destination is within
16110 range for a direct call.
16112 @option{-mrelax-pic-calls} is the default if GCC was configured to use
16113 an assembler and a linker that supports the @code{.reloc} assembly
16114 directive and @code{-mexplicit-relocs} is in effect. With
16115 @code{-mno-explicit-relocs}, this optimization can be performed by the
16116 assembler and the linker alone without help from the compiler.
16118 @item -mmcount-ra-address
16119 @itemx -mno-mcount-ra-address
16120 @opindex mmcount-ra-address
16121 @opindex mno-mcount-ra-address
16122 Emit (do not emit) code that allows @code{_mcount} to modify the
16123 calling function's return address. When enabled, this option extends
16124 the usual @code{_mcount} interface with a new @var{ra-address}
16125 parameter, which has type @code{intptr_t *} and is passed in register
16126 @code{$12}. @code{_mcount} can then modify the return address by
16127 doing both of the following:
16130 Returning the new address in register @code{$31}.
16132 Storing the new address in @code{*@var{ra-address}},
16133 if @var{ra-address} is nonnull.
16136 The default is @option{-mno-mcount-ra-address}.
16141 @subsection MMIX Options
16142 @cindex MMIX Options
16144 These options are defined for the MMIX:
16148 @itemx -mno-libfuncs
16150 @opindex mno-libfuncs
16151 Specify that intrinsic library functions are being compiled, passing all
16152 values in registers, no matter the size.
16155 @itemx -mno-epsilon
16157 @opindex mno-epsilon
16158 Generate floating-point comparison instructions that compare with respect
16159 to the @code{rE} epsilon register.
16161 @item -mabi=mmixware
16163 @opindex mabi=mmixware
16165 Generate code that passes function parameters and return values that (in
16166 the called function) are seen as registers @code{$0} and up, as opposed to
16167 the GNU ABI which uses global registers @code{$231} and up.
16169 @item -mzero-extend
16170 @itemx -mno-zero-extend
16171 @opindex mzero-extend
16172 @opindex mno-zero-extend
16173 When reading data from memory in sizes shorter than 64 bits, use (do not
16174 use) zero-extending load instructions by default, rather than
16175 sign-extending ones.
16178 @itemx -mno-knuthdiv
16180 @opindex mno-knuthdiv
16181 Make the result of a division yielding a remainder have the same sign as
16182 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
16183 remainder follows the sign of the dividend. Both methods are
16184 arithmetically valid, the latter being almost exclusively used.
16186 @item -mtoplevel-symbols
16187 @itemx -mno-toplevel-symbols
16188 @opindex mtoplevel-symbols
16189 @opindex mno-toplevel-symbols
16190 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
16191 code can be used with the @code{PREFIX} assembly directive.
16195 Generate an executable in the ELF format, rather than the default
16196 @samp{mmo} format used by the @command{mmix} simulator.
16198 @item -mbranch-predict
16199 @itemx -mno-branch-predict
16200 @opindex mbranch-predict
16201 @opindex mno-branch-predict
16202 Use (do not use) the probable-branch instructions, when static branch
16203 prediction indicates a probable branch.
16205 @item -mbase-addresses
16206 @itemx -mno-base-addresses
16207 @opindex mbase-addresses
16208 @opindex mno-base-addresses
16209 Generate (do not generate) code that uses @emph{base addresses}. Using a
16210 base address automatically generates a request (handled by the assembler
16211 and the linker) for a constant to be set up in a global register. The
16212 register is used for one or more base address requests within the range 0
16213 to 255 from the value held in the register. The generally leads to short
16214 and fast code, but the number of different data items that can be
16215 addressed is limited. This means that a program that uses lots of static
16216 data may require @option{-mno-base-addresses}.
16218 @item -msingle-exit
16219 @itemx -mno-single-exit
16220 @opindex msingle-exit
16221 @opindex mno-single-exit
16222 Force (do not force) generated code to have a single exit point in each
16226 @node MN10300 Options
16227 @subsection MN10300 Options
16228 @cindex MN10300 options
16230 These @option{-m} options are defined for Matsushita MN10300 architectures:
16235 Generate code to avoid bugs in the multiply instructions for the MN10300
16236 processors. This is the default.
16238 @item -mno-mult-bug
16239 @opindex mno-mult-bug
16240 Do not generate code to avoid bugs in the multiply instructions for the
16241 MN10300 processors.
16245 Generate code using features specific to the AM33 processor.
16249 Do not generate code using features specific to the AM33 processor. This
16254 Generate code using features specific to the AM33/2.0 processor.
16258 Generate code using features specific to the AM34 processor.
16260 @item -mtune=@var{cpu-type}
16262 Use the timing characteristics of the indicated CPU type when
16263 scheduling instructions. This does not change the targeted processor
16264 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
16265 @samp{am33-2} or @samp{am34}.
16267 @item -mreturn-pointer-on-d0
16268 @opindex mreturn-pointer-on-d0
16269 When generating a function that returns a pointer, return the pointer
16270 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
16271 only in @code{a0}, and attempts to call such functions without a prototype
16272 result in errors. Note that this option is on by default; use
16273 @option{-mno-return-pointer-on-d0} to disable it.
16277 Do not link in the C run-time initialization object file.
16281 Indicate to the linker that it should perform a relaxation optimization pass
16282 to shorten branches, calls and absolute memory addresses. This option only
16283 has an effect when used on the command line for the final link step.
16285 This option makes symbolic debugging impossible.
16289 Allow the compiler to generate @emph{Long Instruction Word}
16290 instructions if the target is the @samp{AM33} or later. This is the
16291 default. This option defines the preprocessor macro @samp{__LIW__}.
16295 Do not allow the compiler to generate @emph{Long Instruction Word}
16296 instructions. This option defines the preprocessor macro
16301 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
16302 instructions if the target is the @samp{AM33} or later. This is the
16303 default. This option defines the preprocessor macro @samp{__SETLB__}.
16307 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
16308 instructions. This option defines the preprocessor macro
16309 @samp{__NO_SETLB__}.
16313 @node PDP-11 Options
16314 @subsection PDP-11 Options
16315 @cindex PDP-11 Options
16317 These options are defined for the PDP-11:
16322 Use hardware FPP floating point. This is the default. (FIS floating
16323 point on the PDP-11/40 is not supported.)
16326 @opindex msoft-float
16327 Do not use hardware floating point.
16331 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
16335 Return floating-point results in memory. This is the default.
16339 Generate code for a PDP-11/40.
16343 Generate code for a PDP-11/45. This is the default.
16347 Generate code for a PDP-11/10.
16349 @item -mbcopy-builtin
16350 @opindex mbcopy-builtin
16351 Use inline @code{movmemhi} patterns for copying memory. This is the
16356 Do not use inline @code{movmemhi} patterns for copying memory.
16362 Use 16-bit @code{int}. This is the default.
16368 Use 32-bit @code{int}.
16371 @itemx -mno-float32
16373 @opindex mno-float32
16374 Use 64-bit @code{float}. This is the default.
16377 @itemx -mno-float64
16379 @opindex mno-float64
16380 Use 32-bit @code{float}.
16384 Use @code{abshi2} pattern. This is the default.
16388 Do not use @code{abshi2} pattern.
16390 @item -mbranch-expensive
16391 @opindex mbranch-expensive
16392 Pretend that branches are expensive. This is for experimenting with
16393 code generation only.
16395 @item -mbranch-cheap
16396 @opindex mbranch-cheap
16397 Do not pretend that branches are expensive. This is the default.
16401 Use Unix assembler syntax. This is the default when configured for
16402 @samp{pdp11-*-bsd}.
16406 Use DEC assembler syntax. This is the default when configured for any
16407 PDP-11 target other than @samp{pdp11-*-bsd}.
16410 @node picoChip Options
16411 @subsection picoChip Options
16412 @cindex picoChip options
16414 These @samp{-m} options are defined for picoChip implementations:
16418 @item -mae=@var{ae_type}
16420 Set the instruction set, register set, and instruction scheduling
16421 parameters for array element type @var{ae_type}. Supported values
16422 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
16424 @option{-mae=ANY} selects a completely generic AE type. Code
16425 generated with this option runs on any of the other AE types. The
16426 code is not as efficient as it would be if compiled for a specific
16427 AE type, and some types of operation (e.g., multiplication) do not
16428 work properly on all types of AE.
16430 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
16431 for compiled code, and is the default.
16433 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
16434 option may suffer from poor performance of byte (char) manipulation,
16435 since the DSP AE does not provide hardware support for byte load/stores.
16437 @item -msymbol-as-address
16438 Enable the compiler to directly use a symbol name as an address in a
16439 load/store instruction, without first loading it into a
16440 register. Typically, the use of this option generates larger
16441 programs, which run faster than when the option isn't used. However, the
16442 results vary from program to program, so it is left as a user option,
16443 rather than being permanently enabled.
16445 @item -mno-inefficient-warnings
16446 Disables warnings about the generation of inefficient code. These
16447 warnings can be generated, for example, when compiling code that
16448 performs byte-level memory operations on the MAC AE type. The MAC AE has
16449 no hardware support for byte-level memory operations, so all byte
16450 load/stores must be synthesized from word load/store operations. This is
16451 inefficient and a warning is generated to indicate
16452 that you should rewrite the code to avoid byte operations, or to target
16453 an AE type that has the necessary hardware support. This option disables
16458 @node PowerPC Options
16459 @subsection PowerPC Options
16460 @cindex PowerPC options
16462 These are listed under @xref{RS/6000 and PowerPC Options}.
16465 @subsection RL78 Options
16466 @cindex RL78 Options
16472 Links in additional target libraries to support operation within a
16479 Specifies the type of hardware multiplication support to be used. The
16480 default is @code{none}, which uses software multiplication functions.
16481 The @code{g13} option is for the hardware multiply/divide peripheral
16482 only on the RL78/G13 targets. The @code{rl78} option is for the
16483 standard hardware multiplication defined in the RL78 software manual.
16487 @node RS/6000 and PowerPC Options
16488 @subsection IBM RS/6000 and PowerPC Options
16489 @cindex RS/6000 and PowerPC Options
16490 @cindex IBM RS/6000 and PowerPC Options
16492 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
16499 @itemx -mno-powerpc
16500 @itemx -mpowerpc-gpopt
16501 @itemx -mno-powerpc-gpopt
16502 @itemx -mpowerpc-gfxopt
16503 @itemx -mno-powerpc-gfxopt
16506 @itemx -mno-powerpc64
16510 @itemx -mno-popcntb
16512 @itemx -mno-popcntd
16521 @itemx -mno-hard-dfp
16525 @opindex mno-power2
16527 @opindex mno-powerpc
16528 @opindex mpowerpc-gpopt
16529 @opindex mno-powerpc-gpopt
16530 @opindex mpowerpc-gfxopt
16531 @opindex mno-powerpc-gfxopt
16532 @opindex mpowerpc64
16533 @opindex mno-powerpc64
16537 @opindex mno-popcntb
16539 @opindex mno-popcntd
16545 @opindex mno-mfpgpr
16547 @opindex mno-hard-dfp
16548 GCC supports two related instruction set architectures for the
16549 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
16550 instructions supported by the @samp{rios} chip set used in the original
16551 RS/6000 systems and the @dfn{PowerPC} instruction set is the
16552 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
16553 the IBM 4xx, 6xx, and follow-on microprocessors.
16555 Neither architecture is a subset of the other. However there is a
16556 large common subset of instructions supported by both. An MQ
16557 register is included in processors supporting the POWER architecture.
16559 You use these options to specify which instructions are available on the
16560 processor you are using. The default value of these options is
16561 determined when configuring GCC@. Specifying the
16562 @option{-mcpu=@var{cpu_type}} overrides the specification of these
16563 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
16564 rather than the options listed above.
16566 The @option{-mpower} option allows GCC to generate instructions that
16567 are found only in the POWER architecture and to use the MQ register.
16568 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
16569 to generate instructions that are present in the POWER2 architecture but
16570 not the original POWER architecture.
16572 The @option{-mpowerpc} option allows GCC to generate instructions that
16573 are found only in the 32-bit subset of the PowerPC architecture.
16574 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
16575 GCC to use the optional PowerPC architecture instructions in the
16576 General Purpose group, including floating-point square root. Specifying
16577 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
16578 use the optional PowerPC architecture instructions in the Graphics
16579 group, including floating-point select.
16581 The @option{-mmfcrf} option allows GCC to generate the move from
16582 condition register field instruction implemented on the POWER4
16583 processor and other processors that support the PowerPC V2.01
16585 The @option{-mpopcntb} option allows GCC to generate the popcount and
16586 double-precision FP reciprocal estimate instruction implemented on the
16587 POWER5 processor and other processors that support the PowerPC V2.02
16589 The @option{-mpopcntd} option allows GCC to generate the popcount
16590 instruction implemented on the POWER7 processor and other processors
16591 that support the PowerPC V2.06 architecture.
16592 The @option{-mfprnd} option allows GCC to generate the FP round to
16593 integer instructions implemented on the POWER5+ processor and other
16594 processors that support the PowerPC V2.03 architecture.
16595 The @option{-mcmpb} option allows GCC to generate the compare bytes
16596 instruction implemented on the POWER6 processor and other processors
16597 that support the PowerPC V2.05 architecture.
16598 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
16599 general-purpose register instructions implemented on the POWER6X
16600 processor and other processors that support the extended PowerPC V2.05
16602 The @option{-mhard-dfp} option allows GCC to generate the decimal
16603 floating-point instructions implemented on some POWER processors.
16605 The @option{-mpowerpc64} option allows GCC to generate the additional
16606 64-bit instructions that are found in the full PowerPC64 architecture
16607 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
16608 @option{-mno-powerpc64}.
16610 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
16611 uses only the instructions in the common subset of both
16612 architectures plus some special AIX common-mode calls, and does not use
16613 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
16614 permits GCC to use any instruction from either architecture and to
16615 allow use of the MQ register; specify this for the Motorola MPC601.
16617 @item -mnew-mnemonics
16618 @itemx -mold-mnemonics
16619 @opindex mnew-mnemonics
16620 @opindex mold-mnemonics
16621 Select which mnemonics to use in the generated assembler code. With
16622 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
16623 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
16624 assembler mnemonics defined for the POWER architecture. Instructions
16625 defined in only one architecture have only one mnemonic; GCC uses that
16626 mnemonic irrespective of which of these options is specified.
16628 GCC defaults to the mnemonics appropriate for the architecture in
16629 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
16630 value of these option. Unless you are building a cross-compiler, you
16631 should normally not specify either @option{-mnew-mnemonics} or
16632 @option{-mold-mnemonics}, but should instead accept the default.
16634 @item -mcpu=@var{cpu_type}
16636 Set architecture type, register usage, choice of mnemonics, and
16637 instruction scheduling parameters for machine type @var{cpu_type}.
16638 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
16639 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
16640 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
16641 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
16642 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
16643 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
16644 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
16645 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
16646 @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
16647 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
16648 @samp{power6x}, @samp{power7}, @samp{common}, @samp{powerpc},
16649 @samp{powerpc64}, @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc},
16652 @option{-mcpu=common} selects a completely generic processor. Code
16653 generated under this option runs on any POWER or PowerPC processor.
16654 GCC uses only the instructions in the common subset of both
16655 architectures, and does not use the MQ register. GCC assumes a generic
16656 processor model for scheduling purposes.
16658 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
16659 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
16660 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
16661 types, with an appropriate, generic processor model assumed for
16662 scheduling purposes.
16664 The other options specify a specific processor. Code generated under
16665 those options runs best on that processor, and may not run at all on
16668 The @option{-mcpu} options automatically enable or disable the
16671 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
16672 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
16673 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
16674 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
16676 The particular options set for any particular CPU varies between
16677 compiler versions, depending on what setting seems to produce optimal
16678 code for that CPU; it doesn't necessarily reflect the actual hardware's
16679 capabilities. If you wish to set an individual option to a particular
16680 value, you may specify it after the @option{-mcpu} option, like
16681 @option{-mcpu=970 -mno-altivec}.
16683 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
16684 not enabled or disabled by the @option{-mcpu} option at present because
16685 AIX does not have full support for these options. You may still
16686 enable or disable them individually if you're sure it'll work in your
16689 @item -mtune=@var{cpu_type}
16691 Set the instruction scheduling parameters for machine type
16692 @var{cpu_type}, but do not set the architecture type, register usage, or
16693 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} does. The same
16694 values for @var{cpu_type} are used for @option{-mtune} as for
16695 @option{-mcpu}. If both are specified, the code generated uses the
16696 architecture, registers, and mnemonics set by @option{-mcpu}, but the
16697 scheduling parameters set by @option{-mtune}.
16699 @item -mcmodel=small
16700 @opindex mcmodel=small
16701 Generate PowerPC64 code for the small model: The TOC is limited to
16704 @item -mcmodel=medium
16705 @opindex mcmodel=medium
16706 Generate PowerPC64 code for the medium model: The TOC and other static
16707 data may be up to a total of 4G in size.
16709 @item -mcmodel=large
16710 @opindex mcmodel=large
16711 Generate PowerPC64 code for the large model: The TOC may be up to 4G
16712 in size. Other data and code is only limited by the 64-bit address
16716 @itemx -mno-altivec
16718 @opindex mno-altivec
16719 Generate code that uses (does not use) AltiVec instructions, and also
16720 enable the use of built-in functions that allow more direct access to
16721 the AltiVec instruction set. You may also need to set
16722 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
16728 @opindex mno-vrsave
16729 Generate VRSAVE instructions when generating AltiVec code.
16731 @item -mgen-cell-microcode
16732 @opindex mgen-cell-microcode
16733 Generate Cell microcode instructions
16735 @item -mwarn-cell-microcode
16736 @opindex mwarn-cell-microcode
16737 Warning when a Cell microcode instruction is going to emitted. An example
16738 of a Cell microcode instruction is a variable shift.
16741 @opindex msecure-plt
16742 Generate code that allows @command{ld} and @command{ld.so}
16743 to build executables and shared
16744 libraries with non-executable @code{.plt} and @code{.got} sections.
16746 32-bit SYSV ABI option.
16750 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
16752 requires @code{.plt} and @code{.got}
16753 sections that are both writable and executable.
16754 This is a PowerPC 32-bit SYSV ABI option.
16760 This switch enables or disables the generation of ISEL instructions.
16762 @item -misel=@var{yes/no}
16763 This switch has been deprecated. Use @option{-misel} and
16764 @option{-mno-isel} instead.
16770 This switch enables or disables the generation of SPE simd
16776 @opindex mno-paired
16777 This switch enables or disables the generation of PAIRED simd
16780 @item -mspe=@var{yes/no}
16781 This option has been deprecated. Use @option{-mspe} and
16782 @option{-mno-spe} instead.
16788 Generate code that uses (does not use) vector/scalar (VSX)
16789 instructions, and also enable the use of built-in functions that allow
16790 more direct access to the VSX instruction set.
16792 @item -mfloat-gprs=@var{yes/single/double/no}
16793 @itemx -mfloat-gprs
16794 @opindex mfloat-gprs
16795 This switch enables or disables the generation of floating-point
16796 operations on the general-purpose registers for architectures that
16799 The argument @var{yes} or @var{single} enables the use of
16800 single-precision floating-point operations.
16802 The argument @var{double} enables the use of single and
16803 double-precision floating-point operations.
16805 The argument @var{no} disables floating-point operations on the
16806 general-purpose registers.
16808 This option is currently only available on the MPC854x.
16814 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
16815 targets (including GNU/Linux). The 32-bit environment sets int, long
16816 and pointer to 32 bits and generates code that runs on any PowerPC
16817 variant. The 64-bit environment sets int to 32 bits and long and
16818 pointer to 64 bits, and generates code for PowerPC64, as for
16819 @option{-mpowerpc64}.
16822 @itemx -mno-fp-in-toc
16823 @itemx -mno-sum-in-toc
16824 @itemx -mminimal-toc
16826 @opindex mno-fp-in-toc
16827 @opindex mno-sum-in-toc
16828 @opindex mminimal-toc
16829 Modify generation of the TOC (Table Of Contents), which is created for
16830 every executable file. The @option{-mfull-toc} option is selected by
16831 default. In that case, GCC allocates at least one TOC entry for
16832 each unique non-automatic variable reference in your program. GCC
16833 also places floating-point constants in the TOC@. However, only
16834 16,384 entries are available in the TOC@.
16836 If you receive a linker error message that saying you have overflowed
16837 the available TOC space, you can reduce the amount of TOC space used
16838 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
16839 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
16840 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
16841 generate code to calculate the sum of an address and a constant at
16842 run time instead of putting that sum into the TOC@. You may specify one
16843 or both of these options. Each causes GCC to produce very slightly
16844 slower and larger code at the expense of conserving TOC space.
16846 If you still run out of space in the TOC even when you specify both of
16847 these options, specify @option{-mminimal-toc} instead. This option causes
16848 GCC to make only one TOC entry for every file. When you specify this
16849 option, GCC produces code that is slower and larger but which
16850 uses extremely little TOC space. You may wish to use this option
16851 only on files that contain less frequently-executed code.
16857 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
16858 @code{long} type, and the infrastructure needed to support them.
16859 Specifying @option{-maix64} implies @option{-mpowerpc64} and
16860 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
16861 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
16864 @itemx -mno-xl-compat
16865 @opindex mxl-compat
16866 @opindex mno-xl-compat
16867 Produce code that conforms more closely to IBM XL compiler semantics
16868 when using AIX-compatible ABI@. Pass floating-point arguments to
16869 prototyped functions beyond the register save area (RSA) on the stack
16870 in addition to argument FPRs. Do not assume that most significant
16871 double in 128-bit long double value is properly rounded when comparing
16872 values and converting to double. Use XL symbol names for long double
16875 The AIX calling convention was extended but not initially documented to
16876 handle an obscure K&R C case of calling a function that takes the
16877 address of its arguments with fewer arguments than declared. IBM XL
16878 compilers access floating-point arguments that do not fit in the
16879 RSA from the stack when a subroutine is compiled without
16880 optimization. Because always storing floating-point arguments on the
16881 stack is inefficient and rarely needed, this option is not enabled by
16882 default and only is necessary when calling subroutines compiled by IBM
16883 XL compilers without optimization.
16887 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
16888 application written to use message passing with special startup code to
16889 enable the application to run. The system must have PE installed in the
16890 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
16891 must be overridden with the @option{-specs=} option to specify the
16892 appropriate directory location. The Parallel Environment does not
16893 support threads, so the @option{-mpe} option and the @option{-pthread}
16894 option are incompatible.
16896 @item -malign-natural
16897 @itemx -malign-power
16898 @opindex malign-natural
16899 @opindex malign-power
16900 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
16901 @option{-malign-natural} overrides the ABI-defined alignment of larger
16902 types, such as floating-point doubles, on their natural size-based boundary.
16903 The option @option{-malign-power} instructs GCC to follow the ABI-specified
16904 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
16906 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
16910 @itemx -mhard-float
16911 @opindex msoft-float
16912 @opindex mhard-float
16913 Generate code that does not use (uses) the floating-point register set.
16914 Software floating-point emulation is provided if you use the
16915 @option{-msoft-float} option, and pass the option to GCC when linking.
16917 @item -msingle-float
16918 @itemx -mdouble-float
16919 @opindex msingle-float
16920 @opindex mdouble-float
16921 Generate code for single- or double-precision floating-point operations.
16922 @option{-mdouble-float} implies @option{-msingle-float}.
16925 @opindex msimple-fpu
16926 Do not generate @code{sqrt} and @code{div} instructions for hardware
16927 floating-point unit.
16929 @item -mfpu=@var{name}
16931 Specify type of floating-point unit. Valid values for @var{name} are
16932 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
16933 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
16934 @samp{sp_full} (equivalent to @option{-msingle-float}),
16935 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
16938 @opindex mxilinx-fpu
16939 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
16942 @itemx -mno-multiple
16944 @opindex mno-multiple
16945 Generate code that uses (does not use) the load multiple word
16946 instructions and the store multiple word instructions. These
16947 instructions are generated by default on POWER systems, and not
16948 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
16949 PowerPC systems, since those instructions do not work when the
16950 processor is in little-endian mode. The exceptions are PPC740 and
16951 PPC750 which permit these instructions in little-endian mode.
16956 @opindex mno-string
16957 Generate code that uses (does not use) the load string instructions
16958 and the store string word instructions to save multiple registers and
16959 do small block moves. These instructions are generated by default on
16960 POWER systems, and not generated on PowerPC systems. Do not use
16961 @option{-mstring} on little-endian PowerPC systems, since those
16962 instructions do not work when the processor is in little-endian mode.
16963 The exceptions are PPC740 and PPC750 which permit these instructions
16964 in little-endian mode.
16969 @opindex mno-update
16970 Generate code that uses (does not use) the load or store instructions
16971 that update the base register to the address of the calculated memory
16972 location. These instructions are generated by default. If you use
16973 @option{-mno-update}, there is a small window between the time that the
16974 stack pointer is updated and the address of the previous frame is
16975 stored, which means code that walks the stack frame across interrupts or
16976 signals may get corrupted data.
16978 @item -mavoid-indexed-addresses
16979 @itemx -mno-avoid-indexed-addresses
16980 @opindex mavoid-indexed-addresses
16981 @opindex mno-avoid-indexed-addresses
16982 Generate code that tries to avoid (not avoid) the use of indexed load
16983 or store instructions. These instructions can incur a performance
16984 penalty on Power6 processors in certain situations, such as when
16985 stepping through large arrays that cross a 16M boundary. This option
16986 is enabled by default when targeting Power6 and disabled otherwise.
16989 @itemx -mno-fused-madd
16990 @opindex mfused-madd
16991 @opindex mno-fused-madd
16992 Generate code that uses (does not use) the floating-point multiply and
16993 accumulate instructions. These instructions are generated by default
16994 if hardware floating point is used. The machine-dependent
16995 @option{-mfused-madd} option is now mapped to the machine-independent
16996 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16997 mapped to @option{-ffp-contract=off}.
17003 Generate code that uses (does not use) the half-word multiply and
17004 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
17005 These instructions are generated by default when targeting those
17012 Generate code that uses (does not use) the string-search @samp{dlmzb}
17013 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
17014 generated by default when targeting those processors.
17016 @item -mno-bit-align
17018 @opindex mno-bit-align
17019 @opindex mbit-align
17020 On System V.4 and embedded PowerPC systems do not (do) force structures
17021 and unions that contain bit-fields to be aligned to the base type of the
17024 For example, by default a structure containing nothing but 8
17025 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
17026 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
17027 the structure is aligned to a 1-byte boundary and is 1 byte in
17030 @item -mno-strict-align
17031 @itemx -mstrict-align
17032 @opindex mno-strict-align
17033 @opindex mstrict-align
17034 On System V.4 and embedded PowerPC systems do not (do) assume that
17035 unaligned memory references are handled by the system.
17037 @item -mrelocatable
17038 @itemx -mno-relocatable
17039 @opindex mrelocatable
17040 @opindex mno-relocatable
17041 Generate code that allows (does not allow) a static executable to be
17042 relocated to a different address at run time. A simple embedded
17043 PowerPC system loader should relocate the entire contents of
17044 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
17045 a table of 32-bit addresses generated by this option. For this to
17046 work, all objects linked together must be compiled with
17047 @option{-mrelocatable} or @option{-mrelocatable-lib}.
17048 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
17050 @item -mrelocatable-lib
17051 @itemx -mno-relocatable-lib
17052 @opindex mrelocatable-lib
17053 @opindex mno-relocatable-lib
17054 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
17055 @code{.fixup} section to allow static executables to be relocated at
17056 run time, but @option{-mrelocatable-lib} does not use the smaller stack
17057 alignment of @option{-mrelocatable}. Objects compiled with
17058 @option{-mrelocatable-lib} may be linked with objects compiled with
17059 any combination of the @option{-mrelocatable} options.
17065 On System V.4 and embedded PowerPC systems do not (do) assume that
17066 register 2 contains a pointer to a global area pointing to the addresses
17067 used in the program.
17070 @itemx -mlittle-endian
17072 @opindex mlittle-endian
17073 On System V.4 and embedded PowerPC systems compile code for the
17074 processor in little-endian mode. The @option{-mlittle-endian} option is
17075 the same as @option{-mlittle}.
17078 @itemx -mbig-endian
17080 @opindex mbig-endian
17081 On System V.4 and embedded PowerPC systems compile code for the
17082 processor in big-endian mode. The @option{-mbig-endian} option is
17083 the same as @option{-mbig}.
17085 @item -mdynamic-no-pic
17086 @opindex mdynamic-no-pic
17087 On Darwin and Mac OS X systems, compile code so that it is not
17088 relocatable, but that its external references are relocatable. The
17089 resulting code is suitable for applications, but not shared
17092 @item -msingle-pic-base
17093 @opindex msingle-pic-base
17094 Treat the register used for PIC addressing as read-only, rather than
17095 loading it in the prologue for each function. The runtime system is
17096 responsible for initializing this register with an appropriate value
17097 before execution begins.
17099 @item -mprioritize-restricted-insns=@var{priority}
17100 @opindex mprioritize-restricted-insns
17101 This option controls the priority that is assigned to
17102 dispatch-slot restricted instructions during the second scheduling
17103 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
17104 or @samp{2} to assign no, highest, or second-highest (respectively)
17105 priority to dispatch-slot restricted
17108 @item -msched-costly-dep=@var{dependence_type}
17109 @opindex msched-costly-dep
17110 This option controls which dependences are considered costly
17111 by the target during instruction scheduling. The argument
17112 @var{dependence_type} takes one of the following values:
17116 No dependence is costly.
17119 All dependences are costly.
17121 @item @samp{true_store_to_load}
17122 A true dependence from store to load is costly.
17124 @item @samp{store_to_load}
17125 Any dependence from store to load is costly.
17128 Any dependence for which the latency is greater than or equal to
17129 @var{number} is costly.
17132 @item -minsert-sched-nops=@var{scheme}
17133 @opindex minsert-sched-nops
17134 This option controls which NOP insertion scheme is used during
17135 the second scheduling pass. The argument @var{scheme} takes one of the
17143 Pad with NOPs any dispatch group that has vacant issue slots,
17144 according to the scheduler's grouping.
17146 @item @samp{regroup_exact}
17147 Insert NOPs to force costly dependent insns into
17148 separate groups. Insert exactly as many NOPs as needed to force an insn
17149 to a new group, according to the estimated processor grouping.
17152 Insert NOPs to force costly dependent insns into
17153 separate groups. Insert @var{number} NOPs to force an insn to a new group.
17157 @opindex mcall-sysv
17158 On System V.4 and embedded PowerPC systems compile code using calling
17159 conventions that adheres to the March 1995 draft of the System V
17160 Application Binary Interface, PowerPC processor supplement. This is the
17161 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
17163 @item -mcall-sysv-eabi
17165 @opindex mcall-sysv-eabi
17166 @opindex mcall-eabi
17167 Specify both @option{-mcall-sysv} and @option{-meabi} options.
17169 @item -mcall-sysv-noeabi
17170 @opindex mcall-sysv-noeabi
17171 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
17173 @item -mcall-aixdesc
17175 On System V.4 and embedded PowerPC systems compile code for the AIX
17179 @opindex mcall-linux
17180 On System V.4 and embedded PowerPC systems compile code for the
17181 Linux-based GNU system.
17183 @item -mcall-freebsd
17184 @opindex mcall-freebsd
17185 On System V.4 and embedded PowerPC systems compile code for the
17186 FreeBSD operating system.
17188 @item -mcall-netbsd
17189 @opindex mcall-netbsd
17190 On System V.4 and embedded PowerPC systems compile code for the
17191 NetBSD operating system.
17193 @item -mcall-openbsd
17194 @opindex mcall-netbsd
17195 On System V.4 and embedded PowerPC systems compile code for the
17196 OpenBSD operating system.
17198 @item -maix-struct-return
17199 @opindex maix-struct-return
17200 Return all structures in memory (as specified by the AIX ABI)@.
17202 @item -msvr4-struct-return
17203 @opindex msvr4-struct-return
17204 Return structures smaller than 8 bytes in registers (as specified by the
17207 @item -mabi=@var{abi-type}
17209 Extend the current ABI with a particular extension, or remove such extension.
17210 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
17211 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
17215 Extend the current ABI with SPE ABI extensions. This does not change
17216 the default ABI, instead it adds the SPE ABI extensions to the current
17220 @opindex mabi=no-spe
17221 Disable Booke SPE ABI extensions for the current ABI@.
17223 @item -mabi=ibmlongdouble
17224 @opindex mabi=ibmlongdouble
17225 Change the current ABI to use IBM extended-precision long double.
17226 This is a PowerPC 32-bit SYSV ABI option.
17228 @item -mabi=ieeelongdouble
17229 @opindex mabi=ieeelongdouble
17230 Change the current ABI to use IEEE extended-precision long double.
17231 This is a PowerPC 32-bit Linux ABI option.
17234 @itemx -mno-prototype
17235 @opindex mprototype
17236 @opindex mno-prototype
17237 On System V.4 and embedded PowerPC systems assume that all calls to
17238 variable argument functions are properly prototyped. Otherwise, the
17239 compiler must insert an instruction before every non prototyped call to
17240 set or clear bit 6 of the condition code register (@var{CR}) to
17241 indicate whether floating-point values are passed in the floating-point
17242 registers in case the function takes variable arguments. With
17243 @option{-mprototype}, only calls to prototyped variable argument functions
17244 set or clear the bit.
17248 On embedded PowerPC systems, assume that the startup module is called
17249 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
17250 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
17255 On embedded PowerPC systems, assume that the startup module is called
17256 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
17261 On embedded PowerPC systems, assume that the startup module is called
17262 @file{crt0.o} and the standard C libraries are @file{libads.a} and
17265 @item -myellowknife
17266 @opindex myellowknife
17267 On embedded PowerPC systems, assume that the startup module is called
17268 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
17273 On System V.4 and embedded PowerPC systems, specify that you are
17274 compiling for a VxWorks system.
17278 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
17279 header to indicate that @samp{eabi} extended relocations are used.
17285 On System V.4 and embedded PowerPC systems do (do not) adhere to the
17286 Embedded Applications Binary Interface (eabi) which is a set of
17287 modifications to the System V.4 specifications. Selecting @option{-meabi}
17288 means that the stack is aligned to an 8-byte boundary, a function
17289 @code{__eabi} is called to from @code{main} to set up the eabi
17290 environment, and the @option{-msdata} option can use both @code{r2} and
17291 @code{r13} to point to two separate small data areas. Selecting
17292 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
17293 do not call an initialization function from @code{main}, and the
17294 @option{-msdata} option only uses @code{r13} to point to a single
17295 small data area. The @option{-meabi} option is on by default if you
17296 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
17299 @opindex msdata=eabi
17300 On System V.4 and embedded PowerPC systems, put small initialized
17301 @code{const} global and static data in the @samp{.sdata2} section, which
17302 is pointed to by register @code{r2}. Put small initialized
17303 non-@code{const} global and static data in the @samp{.sdata} section,
17304 which is pointed to by register @code{r13}. Put small uninitialized
17305 global and static data in the @samp{.sbss} section, which is adjacent to
17306 the @samp{.sdata} section. The @option{-msdata=eabi} option is
17307 incompatible with the @option{-mrelocatable} option. The
17308 @option{-msdata=eabi} option also sets the @option{-memb} option.
17311 @opindex msdata=sysv
17312 On System V.4 and embedded PowerPC systems, put small global and static
17313 data in the @samp{.sdata} section, which is pointed to by register
17314 @code{r13}. Put small uninitialized global and static data in the
17315 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
17316 The @option{-msdata=sysv} option is incompatible with the
17317 @option{-mrelocatable} option.
17319 @item -msdata=default
17321 @opindex msdata=default
17323 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
17324 compile code the same as @option{-msdata=eabi}, otherwise compile code the
17325 same as @option{-msdata=sysv}.
17328 @opindex msdata=data
17329 On System V.4 and embedded PowerPC systems, put small global
17330 data in the @samp{.sdata} section. Put small uninitialized global
17331 data in the @samp{.sbss} section. Do not use register @code{r13}
17332 to address small data however. This is the default behavior unless
17333 other @option{-msdata} options are used.
17337 @opindex msdata=none
17339 On embedded PowerPC systems, put all initialized global and static data
17340 in the @samp{.data} section, and all uninitialized data in the
17341 @samp{.bss} section.
17343 @item -mblock-move-inline-limit=@var{num}
17344 @opindex mblock-move-inline-limit
17345 Inline all block moves (such as calls to @code{memcpy} or structure
17346 copies) less than or equal to @var{num} bytes. The minimum value for
17347 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
17348 targets. The default value is target-specific.
17352 @cindex smaller data references (PowerPC)
17353 @cindex .sdata/.sdata2 references (PowerPC)
17354 On embedded PowerPC systems, put global and static items less than or
17355 equal to @var{num} bytes into the small data or bss sections instead of
17356 the normal data or bss section. By default, @var{num} is 8. The
17357 @option{-G @var{num}} switch is also passed to the linker.
17358 All modules should be compiled with the same @option{-G @var{num}} value.
17361 @itemx -mno-regnames
17363 @opindex mno-regnames
17364 On System V.4 and embedded PowerPC systems do (do not) emit register
17365 names in the assembly language output using symbolic forms.
17368 @itemx -mno-longcall
17370 @opindex mno-longcall
17371 By default assume that all calls are far away so that a longer more
17372 expensive calling sequence is required. This is required for calls
17373 further than 32 megabytes (33,554,432 bytes) from the current location.
17374 A short call is generated if the compiler knows
17375 the call cannot be that far away. This setting can be overridden by
17376 the @code{shortcall} function attribute, or by @code{#pragma
17379 Some linkers are capable of detecting out-of-range calls and generating
17380 glue code on the fly. On these systems, long calls are unnecessary and
17381 generate slower code. As of this writing, the AIX linker can do this,
17382 as can the GNU linker for PowerPC/64. It is planned to add this feature
17383 to the GNU linker for 32-bit PowerPC systems as well.
17385 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
17386 callee, L42}, plus a @dfn{branch island} (glue code). The two target
17387 addresses represent the callee and the branch island. The
17388 Darwin/PPC linker prefers the first address and generates a @code{bl
17389 callee} if the PPC @code{bl} instruction reaches the callee directly;
17390 otherwise, the linker generates @code{bl L42} to call the branch
17391 island. The branch island is appended to the body of the
17392 calling function; it computes the full 32-bit address of the callee
17395 On Mach-O (Darwin) systems, this option directs the compiler emit to
17396 the glue for every direct call, and the Darwin linker decides whether
17397 to use or discard it.
17399 In the future, GCC may ignore all longcall specifications
17400 when the linker is known to generate glue.
17402 @item -mtls-markers
17403 @itemx -mno-tls-markers
17404 @opindex mtls-markers
17405 @opindex mno-tls-markers
17406 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
17407 specifying the function argument. The relocation allows the linker to
17408 reliably associate function call with argument setup instructions for
17409 TLS optimization, which in turn allows GCC to better schedule the
17414 Adds support for multithreading with the @dfn{pthreads} library.
17415 This option sets flags for both the preprocessor and linker.
17420 This option enables use of the reciprocal estimate and
17421 reciprocal square root estimate instructions with additional
17422 Newton-Raphson steps to increase precision instead of doing a divide or
17423 square root and divide for floating-point arguments. You should use
17424 the @option{-ffast-math} option when using @option{-mrecip} (or at
17425 least @option{-funsafe-math-optimizations},
17426 @option{-finite-math-only}, @option{-freciprocal-math} and
17427 @option{-fno-trapping-math}). Note that while the throughput of the
17428 sequence is generally higher than the throughput of the non-reciprocal
17429 instruction, the precision of the sequence can be decreased by up to 2
17430 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
17433 @item -mrecip=@var{opt}
17434 @opindex mrecip=opt
17435 This option allows to control which reciprocal estimate instructions
17436 may be used. @var{opt} is a comma separated list of options, which may
17437 be preceded by a @code{!} to invert the option:
17438 @code{all}: enable all estimate instructions,
17439 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
17440 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
17441 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
17442 @code{divf}: enable the single-precision reciprocal approximation instructions;
17443 @code{divd}: enable the double-precision reciprocal approximation instructions;
17444 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
17445 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
17446 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
17448 So, for example, @option{-mrecip=all,!rsqrtd} enables
17449 all of the reciprocal estimate instructions, except for the
17450 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
17451 which handle the double-precision reciprocal square root calculations.
17453 @item -mrecip-precision
17454 @itemx -mno-recip-precision
17455 @opindex mrecip-precision
17456 Assume (do not assume) that the reciprocal estimate instructions
17457 provide higher-precision estimates than is mandated by the PowerPC
17458 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
17459 automatically selects @option{-mrecip-precision}. The double-precision
17460 square root estimate instructions are not generated by
17461 default on low-precision machines, since they do not provide an
17462 estimate that converges after three steps.
17464 @item -mveclibabi=@var{type}
17465 @opindex mveclibabi
17466 Specifies the ABI type to use for vectorizing intrinsics using an
17467 external library. The only type supported at present is @code{mass},
17468 which specifies to use IBM's Mathematical Acceleration Subsystem
17469 (MASS) libraries for vectorizing intrinsics using external libraries.
17470 GCC currently emits calls to @code{acosd2}, @code{acosf4},
17471 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
17472 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
17473 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
17474 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
17475 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
17476 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
17477 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
17478 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
17479 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
17480 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
17481 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
17482 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
17483 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
17484 for power7. Both @option{-ftree-vectorize} and
17485 @option{-funsafe-math-optimizations} must also be enabled. The MASS
17486 libraries must be specified at link time.
17491 Generate (do not generate) the @code{friz} instruction when the
17492 @option{-funsafe-math-optimizations} option is used to optimize
17493 rounding of floating-point values to 64-bit integer and back to floating
17494 point. The @code{friz} instruction does not return the same value if
17495 the floating-point number is too large to fit in an integer.
17497 @item -mpointers-to-nested-functions
17498 @itemx -mno-pointers-to-nested-functions
17499 @opindex mpointers-to-nested-functions
17500 Generate (do not generate) code to load up the static chain register
17501 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
17502 systems where a function pointer points to a 3-word descriptor giving
17503 the function address, TOC value to be loaded in register @var{r2}, and
17504 static chain value to be loaded in register @var{r11}. The
17505 @option{-mpointers-to-nested-functions} is on by default. You cannot
17506 call through pointers to nested functions or pointers
17507 to functions compiled in other languages that use the static chain if
17508 you use the @option{-mno-pointers-to-nested-functions}.
17510 @item -msave-toc-indirect
17511 @itemx -mno-save-toc-indirect
17512 @opindex msave-toc-indirect
17513 Generate (do not generate) code to save the TOC value in the reserved
17514 stack location in the function prologue if the function calls through
17515 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
17516 saved in the prologue, it is saved just before the call through the
17517 pointer. The @option{-mno-save-toc-indirect} option is the default.
17521 @subsection RX Options
17524 These command-line options are defined for RX targets:
17527 @item -m64bit-doubles
17528 @itemx -m32bit-doubles
17529 @opindex m64bit-doubles
17530 @opindex m32bit-doubles
17531 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
17532 or 32 bits (@option{-m32bit-doubles}) in size. The default is
17533 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
17534 works on 32-bit values, which is why the default is
17535 @option{-m32bit-doubles}.
17541 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
17542 floating-point hardware. The default is enabled for the @var{RX600}
17543 series and disabled for the @var{RX200} series.
17545 Floating-point instructions are only generated for 32-bit floating-point
17546 values, however, so the FPU hardware is not used for doubles if the
17547 @option{-m64bit-doubles} option is used.
17549 @emph{Note} If the @option{-fpu} option is enabled then
17550 @option{-funsafe-math-optimizations} is also enabled automatically.
17551 This is because the RX FPU instructions are themselves unsafe.
17553 @item -mcpu=@var{name}
17555 Selects the type of RX CPU to be targeted. Currently three types are
17556 supported, the generic @var{RX600} and @var{RX200} series hardware and
17557 the specific @var{RX610} CPU. The default is @var{RX600}.
17559 The only difference between @var{RX600} and @var{RX610} is that the
17560 @var{RX610} does not support the @code{MVTIPL} instruction.
17562 The @var{RX200} series does not have a hardware floating-point unit
17563 and so @option{-nofpu} is enabled by default when this type is
17566 @item -mbig-endian-data
17567 @itemx -mlittle-endian-data
17568 @opindex mbig-endian-data
17569 @opindex mlittle-endian-data
17570 Store data (but not code) in the big-endian format. The default is
17571 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
17574 @item -msmall-data-limit=@var{N}
17575 @opindex msmall-data-limit
17576 Specifies the maximum size in bytes of global and static variables
17577 which can be placed into the small data area. Using the small data
17578 area can lead to smaller and faster code, but the size of area is
17579 limited and it is up to the programmer to ensure that the area does
17580 not overflow. Also when the small data area is used one of the RX's
17581 registers (usually @code{r13}) is reserved for use pointing to this
17582 area, so it is no longer available for use by the compiler. This
17583 could result in slower and/or larger code if variables are pushed onto
17584 the stack instead of being held in this register.
17586 Note, common variables (variables that have not been initialized) and
17587 constants are not placed into the small data area as they are assigned
17588 to other sections in the output executable.
17590 The default value is zero, which disables this feature. Note, this
17591 feature is not enabled by default with higher optimization levels
17592 (@option{-O2} etc) because of the potentially detrimental effects of
17593 reserving a register. It is up to the programmer to experiment and
17594 discover whether this feature is of benefit to their program. See the
17595 description of the @option{-mpid} option for a description of how the
17596 actual register to hold the small data area pointer is chosen.
17602 Use the simulator runtime. The default is to use the libgloss board
17605 @item -mas100-syntax
17606 @itemx -mno-as100-syntax
17607 @opindex mas100-syntax
17608 @opindex mno-as100-syntax
17609 When generating assembler output use a syntax that is compatible with
17610 Renesas's AS100 assembler. This syntax can also be handled by the GAS
17611 assembler but it has some restrictions so generating it is not the
17614 @item -mmax-constant-size=@var{N}
17615 @opindex mmax-constant-size
17616 Specifies the maximum size, in bytes, of a constant that can be used as
17617 an operand in a RX instruction. Although the RX instruction set does
17618 allow constants of up to 4 bytes in length to be used in instructions,
17619 a longer value equates to a longer instruction. Thus in some
17620 circumstances it can be beneficial to restrict the size of constants
17621 that are used in instructions. Constants that are too big are instead
17622 placed into a constant pool and referenced via register indirection.
17624 The value @var{N} can be between 0 and 4. A value of 0 (the default)
17625 or 4 means that constants of any size are allowed.
17629 Enable linker relaxation. Linker relaxation is a process whereby the
17630 linker attempts to reduce the size of a program by finding shorter
17631 versions of various instructions. Disabled by default.
17633 @item -mint-register=@var{N}
17634 @opindex mint-register
17635 Specify the number of registers to reserve for fast interrupt handler
17636 functions. The value @var{N} can be between 0 and 4. A value of 1
17637 means that register @code{r13} is reserved for the exclusive use
17638 of fast interrupt handlers. A value of 2 reserves @code{r13} and
17639 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
17640 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
17641 A value of 0, the default, does not reserve any registers.
17643 @item -msave-acc-in-interrupts
17644 @opindex msave-acc-in-interrupts
17645 Specifies that interrupt handler functions should preserve the
17646 accumulator register. This is only necessary if normal code might use
17647 the accumulator register, for example because it performs 64-bit
17648 multiplications. The default is to ignore the accumulator as this
17649 makes the interrupt handlers faster.
17655 Enables the generation of position independent data. When enabled any
17656 access to constant data is done via an offset from a base address
17657 held in a register. This allows the location of constant data to be
17658 determined at run time without requiring the executable to be
17659 relocated, which is a benefit to embedded applications with tight
17660 memory constraints. Data that can be modified is not affected by this
17663 Note, using this feature reserves a register, usually @code{r13}, for
17664 the constant data base address. This can result in slower and/or
17665 larger code, especially in complicated functions.
17667 The actual register chosen to hold the constant data base address
17668 depends upon whether the @option{-msmall-data-limit} and/or the
17669 @option{-mint-register} command-line options are enabled. Starting
17670 with register @code{r13} and proceeding downwards, registers are
17671 allocated first to satisfy the requirements of @option{-mint-register},
17672 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
17673 is possible for the small data area register to be @code{r8} if both
17674 @option{-mint-register=4} and @option{-mpid} are specified on the
17677 By default this feature is not enabled. The default can be restored
17678 via the @option{-mno-pid} command-line option.
17682 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
17683 has special significance to the RX port when used with the
17684 @code{interrupt} function attribute. This attribute indicates a
17685 function intended to process fast interrupts. GCC ensures
17686 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
17687 and/or @code{r13} and only provided that the normal use of the
17688 corresponding registers have been restricted via the
17689 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
17692 @node S/390 and zSeries Options
17693 @subsection S/390 and zSeries Options
17694 @cindex S/390 and zSeries Options
17696 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
17700 @itemx -msoft-float
17701 @opindex mhard-float
17702 @opindex msoft-float
17703 Use (do not use) the hardware floating-point instructions and registers
17704 for floating-point operations. When @option{-msoft-float} is specified,
17705 functions in @file{libgcc.a} are used to perform floating-point
17706 operations. When @option{-mhard-float} is specified, the compiler
17707 generates IEEE floating-point instructions. This is the default.
17710 @itemx -mno-hard-dfp
17712 @opindex mno-hard-dfp
17713 Use (do not use) the hardware decimal-floating-point instructions for
17714 decimal-floating-point operations. When @option{-mno-hard-dfp} is
17715 specified, functions in @file{libgcc.a} are used to perform
17716 decimal-floating-point operations. When @option{-mhard-dfp} is
17717 specified, the compiler generates decimal-floating-point hardware
17718 instructions. This is the default for @option{-march=z9-ec} or higher.
17720 @item -mlong-double-64
17721 @itemx -mlong-double-128
17722 @opindex mlong-double-64
17723 @opindex mlong-double-128
17724 These switches control the size of @code{long double} type. A size
17725 of 64 bits makes the @code{long double} type equivalent to the @code{double}
17726 type. This is the default.
17729 @itemx -mno-backchain
17730 @opindex mbackchain
17731 @opindex mno-backchain
17732 Store (do not store) the address of the caller's frame as backchain pointer
17733 into the callee's stack frame.
17734 A backchain may be needed to allow debugging using tools that do not understand
17735 DWARF 2 call frame information.
17736 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
17737 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
17738 the backchain is placed into the topmost word of the 96/160 byte register
17741 In general, code compiled with @option{-mbackchain} is call-compatible with
17742 code compiled with @option{-mmo-backchain}; however, use of the backchain
17743 for debugging purposes usually requires that the whole binary is built with
17744 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
17745 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17746 to build a linux kernel use @option{-msoft-float}.
17748 The default is to not maintain the backchain.
17750 @item -mpacked-stack
17751 @itemx -mno-packed-stack
17752 @opindex mpacked-stack
17753 @opindex mno-packed-stack
17754 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
17755 specified, the compiler uses the all fields of the 96/160 byte register save
17756 area only for their default purpose; unused fields still take up stack space.
17757 When @option{-mpacked-stack} is specified, register save slots are densely
17758 packed at the top of the register save area; unused space is reused for other
17759 purposes, allowing for more efficient use of the available stack space.
17760 However, when @option{-mbackchain} is also in effect, the topmost word of
17761 the save area is always used to store the backchain, and the return address
17762 register is always saved two words below the backchain.
17764 As long as the stack frame backchain is not used, code generated with
17765 @option{-mpacked-stack} is call-compatible with code generated with
17766 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
17767 S/390 or zSeries generated code that uses the stack frame backchain at run
17768 time, not just for debugging purposes. Such code is not call-compatible
17769 with code compiled with @option{-mpacked-stack}. Also, note that the
17770 combination of @option{-mbackchain},
17771 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17772 to build a linux kernel use @option{-msoft-float}.
17774 The default is to not use the packed stack layout.
17777 @itemx -mno-small-exec
17778 @opindex msmall-exec
17779 @opindex mno-small-exec
17780 Generate (or do not generate) code using the @code{bras} instruction
17781 to do subroutine calls.
17782 This only works reliably if the total executable size does not
17783 exceed 64k. The default is to use the @code{basr} instruction instead,
17784 which does not have this limitation.
17790 When @option{-m31} is specified, generate code compliant to the
17791 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
17792 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
17793 particular to generate 64-bit instructions. For the @samp{s390}
17794 targets, the default is @option{-m31}, while the @samp{s390x}
17795 targets default to @option{-m64}.
17801 When @option{-mzarch} is specified, generate code using the
17802 instructions available on z/Architecture.
17803 When @option{-mesa} is specified, generate code using the
17804 instructions available on ESA/390. Note that @option{-mesa} is
17805 not possible with @option{-m64}.
17806 When generating code compliant to the GNU/Linux for S/390 ABI,
17807 the default is @option{-mesa}. When generating code compliant
17808 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
17814 Generate (or do not generate) code using the @code{mvcle} instruction
17815 to perform block moves. When @option{-mno-mvcle} is specified,
17816 use a @code{mvc} loop instead. This is the default unless optimizing for
17823 Print (or do not print) additional debug information when compiling.
17824 The default is to not print debug information.
17826 @item -march=@var{cpu-type}
17828 Generate code that runs on @var{cpu-type}, which is the name of a system
17829 representing a certain processor type. Possible values for
17830 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
17831 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
17832 When generating code using the instructions available on z/Architecture,
17833 the default is @option{-march=z900}. Otherwise, the default is
17834 @option{-march=g5}.
17836 @item -mtune=@var{cpu-type}
17838 Tune to @var{cpu-type} everything applicable about the generated code,
17839 except for the ABI and the set of available instructions.
17840 The list of @var{cpu-type} values is the same as for @option{-march}.
17841 The default is the value used for @option{-march}.
17844 @itemx -mno-tpf-trace
17845 @opindex mtpf-trace
17846 @opindex mno-tpf-trace
17847 Generate code that adds (does not add) in TPF OS specific branches to trace
17848 routines in the operating system. This option is off by default, even
17849 when compiling for the TPF OS@.
17852 @itemx -mno-fused-madd
17853 @opindex mfused-madd
17854 @opindex mno-fused-madd
17855 Generate code that uses (does not use) the floating-point multiply and
17856 accumulate instructions. These instructions are generated by default if
17857 hardware floating point is used.
17859 @item -mwarn-framesize=@var{framesize}
17860 @opindex mwarn-framesize
17861 Emit a warning if the current function exceeds the given frame size. Because
17862 this is a compile-time check it doesn't need to be a real problem when the program
17863 runs. It is intended to identify functions that most probably cause
17864 a stack overflow. It is useful to be used in an environment with limited stack
17865 size e.g.@: the linux kernel.
17867 @item -mwarn-dynamicstack
17868 @opindex mwarn-dynamicstack
17869 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
17870 arrays. This is generally a bad idea with a limited stack size.
17872 @item -mstack-guard=@var{stack-guard}
17873 @itemx -mstack-size=@var{stack-size}
17874 @opindex mstack-guard
17875 @opindex mstack-size
17876 If these options are provided the s390 back end emits additional instructions in
17877 the function prologue which trigger a trap if the stack size is @var{stack-guard}
17878 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
17879 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
17880 the frame size of the compiled function is chosen.
17881 These options are intended to be used to help debugging stack overflow problems.
17882 The additionally emitted code causes only little overhead and hence can also be
17883 used in production like systems without greater performance degradation. The given
17884 values have to be exact powers of 2 and @var{stack-size} has to be greater than
17885 @var{stack-guard} without exceeding 64k.
17886 In order to be efficient the extra code makes the assumption that the stack starts
17887 at an address aligned to the value given by @var{stack-size}.
17888 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
17891 @node Score Options
17892 @subsection Score Options
17893 @cindex Score Options
17895 These options are defined for Score implementations:
17900 Compile code for big-endian mode. This is the default.
17904 Compile code for little-endian mode.
17908 Disable generate bcnz instruction.
17912 Enable generate unaligned load and store instruction.
17916 Enable the use of multiply-accumulate instructions. Disabled by default.
17920 Specify the SCORE5 as the target architecture.
17924 Specify the SCORE5U of the target architecture.
17928 Specify the SCORE7 as the target architecture. This is the default.
17932 Specify the SCORE7D as the target architecture.
17936 @subsection SH Options
17938 These @samp{-m} options are defined for the SH implementations:
17943 Generate code for the SH1.
17947 Generate code for the SH2.
17950 Generate code for the SH2e.
17954 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
17955 that the floating-point unit is not used.
17957 @item -m2a-single-only
17958 @opindex m2a-single-only
17959 Generate code for the SH2a-FPU, in such a way that no double-precision
17960 floating-point operations are used.
17963 @opindex m2a-single
17964 Generate code for the SH2a-FPU assuming the floating-point unit is in
17965 single-precision mode by default.
17969 Generate code for the SH2a-FPU assuming the floating-point unit is in
17970 double-precision mode by default.
17974 Generate code for the SH3.
17978 Generate code for the SH3e.
17982 Generate code for the SH4 without a floating-point unit.
17984 @item -m4-single-only
17985 @opindex m4-single-only
17986 Generate code for the SH4 with a floating-point unit that only
17987 supports single-precision arithmetic.
17991 Generate code for the SH4 assuming the floating-point unit is in
17992 single-precision mode by default.
17996 Generate code for the SH4.
18000 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
18001 floating-point unit is not used.
18003 @item -m4a-single-only
18004 @opindex m4a-single-only
18005 Generate code for the SH4a, in such a way that no double-precision
18006 floating-point operations are used.
18009 @opindex m4a-single
18010 Generate code for the SH4a assuming the floating-point unit is in
18011 single-precision mode by default.
18015 Generate code for the SH4a.
18019 Same as @option{-m4a-nofpu}, except that it implicitly passes
18020 @option{-dsp} to the assembler. GCC doesn't generate any DSP
18021 instructions at the moment.
18025 Compile code for the processor in big-endian mode.
18029 Compile code for the processor in little-endian mode.
18033 Align doubles at 64-bit boundaries. Note that this changes the calling
18034 conventions, and thus some functions from the standard C library do
18035 not work unless you recompile it first with @option{-mdalign}.
18039 Shorten some address references at link time, when possible; uses the
18040 linker option @option{-relax}.
18044 Use 32-bit offsets in @code{switch} tables. The default is to use
18049 Enable the use of bit manipulation instructions on SH2A.
18053 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
18054 alignment constraints.
18058 Comply with the calling conventions defined by Renesas.
18062 Comply with the calling conventions defined by Renesas.
18066 Comply with the calling conventions defined for GCC before the Renesas
18067 conventions were available. This option is the default for all
18068 targets of the SH toolchain.
18071 @opindex mnomacsave
18072 Mark the @code{MAC} register as call-clobbered, even if
18073 @option{-mhitachi} is given.
18077 Increase IEEE compliance of floating-point code.
18078 At the moment, this is equivalent to @option{-fno-finite-math-only}.
18079 When generating 16-bit SH opcodes, getting IEEE-conforming results for
18080 comparisons of NANs / infinities incurs extra overhead in every
18081 floating-point comparison, therefore the default is set to
18082 @option{-ffinite-math-only}.
18084 @item -minline-ic_invalidate
18085 @opindex minline-ic_invalidate
18086 Inline code to invalidate instruction cache entries after setting up
18087 nested function trampolines.
18088 This option has no effect if @option{-musermode} is in effect and the selected
18089 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
18091 If the selected code generation option does not allow the use of the @code{icbi}
18092 instruction, and @option{-musermode} is not in effect, the inlined code
18093 manipulates the instruction cache address array directly with an associative
18094 write. This not only requires privileged mode at run time, but it also
18095 fails if the cache line had been mapped via the TLB and has become unmapped.
18099 Dump instruction size and location in the assembly code.
18102 @opindex mpadstruct
18103 This option is deprecated. It pads structures to multiple of 4 bytes,
18104 which is incompatible with the SH ABI@.
18106 @item -msoft-atomic
18107 @opindex msoft-atomic
18108 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
18109 built-in functions. The generated atomic sequences require support from the
18110 interrupt / exception handling code of the system and are only suitable for
18111 single-core systems. They will not operate correctly on multi-core systems.
18112 This option is enabled by default when the target is @code{sh-*-linux*}.
18113 When the target is SH4A, this option will also partially utilize the hardware
18114 atomic instructions @code{movli.l} and @code{movco.l} to create more
18116 For details on the atomic built-in functions see @ref{__atomic Builtins}.
18118 @item -mhard-atomic
18119 @opindex hard-atomic
18120 Generate hardware atomic sequences for the atomic built-in functions. This
18121 is only available on SH4A and is suitable for multi-core systems. Code
18122 compiled with this option will also be compatible with gUSA aware
18123 interrupt / exception handling systems. In contrast to the
18124 @option{-msoft-atomic} option this will only use the instructions
18125 @code{movli.l} and @code{movco.l} to create atomic sequences.
18128 @opindex menable-tas
18129 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
18130 Notice that depending on the particular hardware and software configuration
18131 this can degrade overall performance due to the operand cache line flushes
18132 that are implied by the @code{tas.b} instruction. On multi-core SH4A
18133 processors the @code{tas.b} instruction must be used with caution since it
18134 can result in data corruption for certain cache configurations.
18138 Optimize for space instead of speed. Implied by @option{-Os}.
18141 @opindex mprefergot
18142 When generating position-independent code, emit function calls using
18143 the Global Offset Table instead of the Procedure Linkage Table.
18147 Don't generate privileged mode only code. This option
18148 implies @option{-mno-inline-ic_invalidate}
18149 if the inlined code would not work in user mode.
18150 This is the default when the target is @code{sh-*-linux*}.
18152 @item -multcost=@var{number}
18153 @opindex multcost=@var{number}
18154 Set the cost to assume for a multiply insn.
18156 @item -mdiv=@var{strategy}
18157 @opindex mdiv=@var{strategy}
18158 Set the division strategy to use for SHmedia code. @var{strategy} must be
18164 Performs the operation in floating point. This has a very high latency,
18165 but needs only a few instructions, so it might be a good choice if
18166 your code has enough easily-exploitable ILP to allow the compiler to
18167 schedule the floating-point instructions together with other instructions.
18168 Division by zero causes a floating-point exception.
18171 Uses integer operations to calculate the inverse of the divisor,
18172 and then multiplies the dividend with the inverse. This strategy allows
18173 CSE and hoisting of the inverse calculation. Division by zero calculates
18174 an unspecified result, but does not trap.
18177 A variant of @samp{inv} where, if no CSE or hoisting opportunities
18178 have been found, or if the entire operation has been hoisted to the same
18179 place, the last stages of the inverse calculation are intertwined with the
18180 final multiply to reduce the overall latency, at the expense of using a few
18181 more instructions, and thus offering fewer scheduling opportunities with
18185 Calls a library function that usually implements the @samp{inv:minlat}
18187 This gives high code density for @code{m5-*media-nofpu} compilations.
18190 Uses a different entry point of the same library function, where it
18191 assumes that a pointer to a lookup table has already been set up, which
18192 exposes the pointer load to CSE and code hoisting optimizations.
18197 Use the @samp{inv} algorithm for initial
18198 code generation, but if the code stays unoptimized, revert to the @samp{call},
18199 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
18200 potentially-trapping side effect of division by zero is carried by a
18201 separate instruction, so it is possible that all the integer instructions
18202 are hoisted out, but the marker for the side effect stays where it is.
18203 A recombination to floating-point operations or a call is not possible
18208 Variants of the @samp{inv:minlat} strategy. In the case
18209 that the inverse calculation is not separated from the multiply, they speed
18210 up division where the dividend fits into 20 bits (plus sign where applicable)
18211 by inserting a test to skip a number of operations in this case; this test
18212 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
18213 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
18217 @item -maccumulate-outgoing-args
18218 @opindex maccumulate-outgoing-args
18219 Reserve space once for outgoing arguments in the function prologue rather
18220 than around each call. Generally beneficial for performance and size. Also
18221 needed for unwinding to avoid changing the stack frame around conditional code.
18223 @item -mdivsi3_libfunc=@var{name}
18224 @opindex mdivsi3_libfunc=@var{name}
18225 Set the name of the library function used for 32-bit signed division to
18227 This only affects the name used in the @samp{call} and @samp{inv:call}
18228 division strategies, and the compiler still expects the same
18229 sets of input/output/clobbered registers as if this option were not present.
18231 @item -mfixed-range=@var{register-range}
18232 @opindex mfixed-range
18233 Generate code treating the given register range as fixed registers.
18234 A fixed register is one that the register allocator can not use. This is
18235 useful when compiling kernel code. A register range is specified as
18236 two registers separated by a dash. Multiple register ranges can be
18237 specified separated by a comma.
18239 @item -mindexed-addressing
18240 @opindex mindexed-addressing
18241 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
18242 This is only safe if the hardware and/or OS implement 32-bit wrap-around
18243 semantics for the indexed addressing mode. The architecture allows the
18244 implementation of processors with 64-bit MMU, which the OS could use to
18245 get 32-bit addressing, but since no current hardware implementation supports
18246 this or any other way to make the indexed addressing mode safe to use in
18247 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
18249 @item -mgettrcost=@var{number}
18250 @opindex mgettrcost=@var{number}
18251 Set the cost assumed for the @code{gettr} instruction to @var{number}.
18252 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
18256 Assume @code{pt*} instructions won't trap. This generally generates
18257 better-scheduled code, but is unsafe on current hardware.
18258 The current architecture
18259 definition says that @code{ptabs} and @code{ptrel} trap when the target
18261 This has the unintentional effect of making it unsafe to schedule these
18262 instructions before a branch, or hoist them out of a loop. For example,
18263 @code{__do_global_ctors}, a part of @file{libgcc}
18264 that runs constructors at program
18265 startup, calls functions in a list which is delimited by @minus{}1. With the
18266 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
18267 That means that all the constructors run a bit more quickly, but when
18268 the loop comes to the end of the list, the program crashes because @code{ptabs}
18269 loads @minus{}1 into a target register.
18271 Since this option is unsafe for any
18272 hardware implementing the current architecture specification, the default
18273 is @option{-mno-pt-fixed}. Unless specified explicitly with
18274 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
18275 this deters register allocation from using target registers for storing
18278 @item -minvalid-symbols
18279 @opindex minvalid-symbols
18280 Assume symbols might be invalid. Ordinary function symbols generated by
18281 the compiler are always valid to load with
18282 @code{movi}/@code{shori}/@code{ptabs} or
18283 @code{movi}/@code{shori}/@code{ptrel},
18284 but with assembler and/or linker tricks it is possible
18285 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
18286 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
18287 It prevents cross-basic-block CSE, hoisting and most scheduling
18288 of symbol loads. The default is @option{-mno-invalid-symbols}.
18290 @item -mbranch-cost=@var{num}
18291 @opindex mbranch-cost=@var{num}
18292 Assume @var{num} to be the cost for a branch instruction. Higher numbers
18293 make the compiler try to generate more branch-free code if possible.
18294 If not specified the value is selected depending on the processor type that
18295 is being compiled for.
18298 @opindex mcbranchdi
18299 Enable the @code{cbranchdi4} instruction pattern.
18303 Emit the @code{cmpeqdi_t} instruction pattern even when @option{-mcbranchdi}
18307 @itemx -mno-fused-madd
18308 @opindex mfused-madd
18309 @opindex mno-fused-madd
18310 If the processor type supports it, setting @code{-mfused-madd} will allow the
18311 usage of the @code{fmac} instruction (floating-point multiply-accumulate) for
18312 regular calculations. Enabling this option might generate faster code but also
18313 produce different numeric floating-point results compared to strict IEEE 754
18314 arithmetic. @code{-mfused-madd} is enabled by default by option
18315 @option{-funsafe-math-optimizations}. Setting @code{-mno-fused-madd} will
18316 disallow the usage of the @code{fmac} instruction for regular calculations
18317 even if @option{-funsafe-math-optimizations} is in effect.
18323 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
18324 and cosine approximations. The option @code{-mfsca} must be used in
18325 combination with @code{-funsafe-math-optimizations}. It is enabled by default
18326 when generating code for SH4A. Using @code{-mno-fsca} disables sine and cosine
18327 approximations even if @code{-funsafe-math-optimizations} is in effect.
18333 Allow or disallow the compiler to emit the @code{fsrra} instruction for
18334 reciprocal square root approximations. The option @code{-mfsrra} must be used
18335 in combination with @code{-funsafe-math-optimizations} and
18336 @code{-ffinite-math-only}. It is enabled by default when generating code for
18337 SH4A. Using @code{-mno-fsrra} disables reciprocal square root approximations
18338 even if @code{-funsafe-math-optimizations} and @code{-ffinite-math-only} are
18341 @item -mpretend-cmove
18342 @opindex mpretend-cmove
18343 Prefer zero-displacement conditional branches for conditional move instruction
18344 patterns. This can result in faster code on the SH4 processor.
18348 @node Solaris 2 Options
18349 @subsection Solaris 2 Options
18350 @cindex Solaris 2 options
18352 These @samp{-m} options are supported on Solaris 2:
18355 @item -mimpure-text
18356 @opindex mimpure-text
18357 @option{-mimpure-text}, used in addition to @option{-shared}, tells
18358 the compiler to not pass @option{-z text} to the linker when linking a
18359 shared object. Using this option, you can link position-dependent
18360 code into a shared object.
18362 @option{-mimpure-text} suppresses the ``relocations remain against
18363 allocatable but non-writable sections'' linker error message.
18364 However, the necessary relocations trigger copy-on-write, and the
18365 shared object is not actually shared across processes. Instead of
18366 using @option{-mimpure-text}, you should compile all source code with
18367 @option{-fpic} or @option{-fPIC}.
18371 These switches are supported in addition to the above on Solaris 2:
18376 Add support for multithreading using the POSIX threads library. This
18377 option sets flags for both the preprocessor and linker. This option does
18378 not affect the thread safety of object code produced by the compiler or
18379 that of libraries supplied with it.
18383 This is a synonym for @option{-pthreads}.
18386 @node SPARC Options
18387 @subsection SPARC Options
18388 @cindex SPARC options
18390 These @samp{-m} options are supported on the SPARC:
18393 @item -mno-app-regs
18395 @opindex mno-app-regs
18397 Specify @option{-mapp-regs} to generate output using the global registers
18398 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
18401 To be fully SVR4 ABI compliant at the cost of some performance loss,
18402 specify @option{-mno-app-regs}. You should compile libraries and system
18403 software with this option.
18409 With @option{-mflat}, the compiler does not generate save/restore instructions
18410 and uses a ``flat'' or single register window model. This model is compatible
18411 with the regular register window model. The local registers and the input
18412 registers (0--5) are still treated as ``call-saved'' registers and are
18413 saved on the stack as needed.
18415 With @option{-mno-flat} (the default), the compiler generates save/restore
18416 instructions (except for leaf functions). This is the normal operating mode.
18419 @itemx -mhard-float
18421 @opindex mhard-float
18422 Generate output containing floating-point instructions. This is the
18426 @itemx -msoft-float
18428 @opindex msoft-float
18429 Generate output containing library calls for floating point.
18430 @strong{Warning:} the requisite libraries are not available for all SPARC
18431 targets. Normally the facilities of the machine's usual C compiler are
18432 used, but this cannot be done directly in cross-compilation. You must make
18433 your own arrangements to provide suitable library functions for
18434 cross-compilation. The embedded targets @samp{sparc-*-aout} and
18435 @samp{sparclite-*-*} do provide software floating-point support.
18437 @option{-msoft-float} changes the calling convention in the output file;
18438 therefore, it is only useful if you compile @emph{all} of a program with
18439 this option. In particular, you need to compile @file{libgcc.a}, the
18440 library that comes with GCC, with @option{-msoft-float} in order for
18443 @item -mhard-quad-float
18444 @opindex mhard-quad-float
18445 Generate output containing quad-word (long double) floating-point
18448 @item -msoft-quad-float
18449 @opindex msoft-quad-float
18450 Generate output containing library calls for quad-word (long double)
18451 floating-point instructions. The functions called are those specified
18452 in the SPARC ABI@. This is the default.
18454 As of this writing, there are no SPARC implementations that have hardware
18455 support for the quad-word floating-point instructions. They all invoke
18456 a trap handler for one of these instructions, and then the trap handler
18457 emulates the effect of the instruction. Because of the trap handler overhead,
18458 this is much slower than calling the ABI library routines. Thus the
18459 @option{-msoft-quad-float} option is the default.
18461 @item -mno-unaligned-doubles
18462 @itemx -munaligned-doubles
18463 @opindex mno-unaligned-doubles
18464 @opindex munaligned-doubles
18465 Assume that doubles have 8-byte alignment. This is the default.
18467 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
18468 alignment only if they are contained in another type, or if they have an
18469 absolute address. Otherwise, it assumes they have 4-byte alignment.
18470 Specifying this option avoids some rare compatibility problems with code
18471 generated by other compilers. It is not the default because it results
18472 in a performance loss, especially for floating-point code.
18474 @item -mno-faster-structs
18475 @itemx -mfaster-structs
18476 @opindex mno-faster-structs
18477 @opindex mfaster-structs
18478 With @option{-mfaster-structs}, the compiler assumes that structures
18479 should have 8-byte alignment. This enables the use of pairs of
18480 @code{ldd} and @code{std} instructions for copies in structure
18481 assignment, in place of twice as many @code{ld} and @code{st} pairs.
18482 However, the use of this changed alignment directly violates the SPARC
18483 ABI@. Thus, it's intended only for use on targets where the developer
18484 acknowledges that their resulting code is not directly in line with
18485 the rules of the ABI@.
18487 @item -mcpu=@var{cpu_type}
18489 Set the instruction set, register set, and instruction scheduling parameters
18490 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
18491 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
18492 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
18493 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
18494 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
18495 and @samp{niagara4}.
18497 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
18498 which selects the best architecture option for the host processor.
18499 @option{-mcpu=native} has no effect if GCC does not recognize
18502 Default instruction scheduling parameters are used for values that select
18503 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
18504 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
18506 Here is a list of each supported architecture and their supported
18514 supersparc, hypersparc, leon
18517 f930, f934, sparclite86x
18523 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
18526 By default (unless configured otherwise), GCC generates code for the V7
18527 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
18528 additionally optimizes it for the Cypress CY7C602 chip, as used in the
18529 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
18530 SPARCStation 1, 2, IPX etc.
18532 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
18533 architecture. The only difference from V7 code is that the compiler emits
18534 the integer multiply and integer divide instructions which exist in SPARC-V8
18535 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
18536 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
18539 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
18540 the SPARC architecture. This adds the integer multiply, integer divide step
18541 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
18542 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
18543 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
18544 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
18545 MB86934 chip, which is the more recent SPARClite with FPU@.
18547 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
18548 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
18549 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
18550 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
18551 optimizes it for the TEMIC SPARClet chip.
18553 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
18554 architecture. This adds 64-bit integer and floating-point move instructions,
18555 3 additional floating-point condition code registers and conditional move
18556 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
18557 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
18558 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
18559 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
18560 @option{-mcpu=niagara}, the compiler additionally optimizes it for
18561 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
18562 additionally optimizes it for Sun UltraSPARC T2 chips. With
18563 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
18564 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
18565 additionally optimizes it for Sun UltraSPARC T4 chips.
18567 @item -mtune=@var{cpu_type}
18569 Set the instruction scheduling parameters for machine type
18570 @var{cpu_type}, but do not set the instruction set or register set that the
18571 option @option{-mcpu=@var{cpu_type}} does.
18573 The same values for @option{-mcpu=@var{cpu_type}} can be used for
18574 @option{-mtune=@var{cpu_type}}, but the only useful values are those
18575 that select a particular CPU implementation. Those are @samp{cypress},
18576 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
18577 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
18578 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
18579 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
18584 @opindex mno-v8plus
18585 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
18586 difference from the V8 ABI is that the global and out registers are
18587 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
18588 mode for all SPARC-V9 processors.
18594 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
18595 Visual Instruction Set extensions. The default is @option{-mno-vis}.
18601 With @option{-mvis2}, GCC generates code that takes advantage of
18602 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
18603 default is @option{-mvis2} when targeting a cpu that supports such
18604 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
18605 also sets @option{-mvis}.
18611 With @option{-mvis3}, GCC generates code that takes advantage of
18612 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
18613 default is @option{-mvis3} when targeting a cpu that supports such
18614 instructions, such as niagara-3 and later. Setting @option{-mvis3}
18615 also sets @option{-mvis2} and @option{-mvis}.
18621 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
18622 population count instruction. The default is @option{-mpopc}
18623 when targeting a cpu that supports such instructions, such as Niagara-2 and
18630 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
18631 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
18632 when targeting a cpu that supports such instructions, such as Niagara-3 and
18636 @opindex mfix-at697f
18637 Enable the documented workaround for the single erratum of the Atmel AT697F
18638 processor (which corresponds to erratum #13 of the AT697E processor).
18641 These @samp{-m} options are supported in addition to the above
18642 on SPARC-V9 processors in 64-bit environments:
18645 @item -mlittle-endian
18646 @opindex mlittle-endian
18647 Generate code for a processor running in little-endian mode. It is only
18648 available for a few configurations and most notably not on Solaris and Linux.
18654 Generate code for a 32-bit or 64-bit environment.
18655 The 32-bit environment sets int, long and pointer to 32 bits.
18656 The 64-bit environment sets int to 32 bits and long and pointer
18659 @item -mcmodel=@var{which}
18661 Set the code model to one of
18665 The Medium/Low code model: 64-bit addresses, programs
18666 must be linked in the low 32 bits of memory. Programs can be statically
18667 or dynamically linked.
18670 The Medium/Middle code model: 64-bit addresses, programs
18671 must be linked in the low 44 bits of memory, the text and data segments must
18672 be less than 2GB in size and the data segment must be located within 2GB of
18676 The Medium/Anywhere code model: 64-bit addresses, programs
18677 may be linked anywhere in memory, the text and data segments must be less
18678 than 2GB in size and the data segment must be located within 2GB of the
18682 The Medium/Anywhere code model for embedded systems:
18683 64-bit addresses, the text and data segments must be less than 2GB in
18684 size, both starting anywhere in memory (determined at link time). The
18685 global register %g4 points to the base of the data segment. Programs
18686 are statically linked and PIC is not supported.
18689 @item -mmemory-model=@var{mem-model}
18690 @opindex mmemory-model
18691 Set the memory model in force on the processor to one of
18695 The default memory model for the processor and operating system.
18698 Relaxed Memory Order
18701 Partial Store Order
18707 Sequential Consistency
18710 These memory models are formally defined in Appendix D of the Sparc V9
18711 architecture manual, as set in the processor's @code{PSTATE.MM} field.
18714 @itemx -mno-stack-bias
18715 @opindex mstack-bias
18716 @opindex mno-stack-bias
18717 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
18718 frame pointer if present, are offset by @minus{}2047 which must be added back
18719 when making stack frame references. This is the default in 64-bit mode.
18720 Otherwise, assume no such offset is present.
18724 @subsection SPU Options
18725 @cindex SPU options
18727 These @samp{-m} options are supported on the SPU:
18731 @itemx -merror-reloc
18732 @opindex mwarn-reloc
18733 @opindex merror-reloc
18735 The loader for SPU does not handle dynamic relocations. By default, GCC
18736 gives an error when it generates code that requires a dynamic
18737 relocation. @option{-mno-error-reloc} disables the error,
18738 @option{-mwarn-reloc} generates a warning instead.
18741 @itemx -munsafe-dma
18743 @opindex munsafe-dma
18745 Instructions that initiate or test completion of DMA must not be
18746 reordered with respect to loads and stores of the memory that is being
18748 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
18749 memory accesses, but that can lead to inefficient code in places where the
18750 memory is known to not change. Rather than mark the memory as volatile,
18751 you can use @option{-msafe-dma} to tell the compiler to treat
18752 the DMA instructions as potentially affecting all memory.
18754 @item -mbranch-hints
18755 @opindex mbranch-hints
18757 By default, GCC generates a branch hint instruction to avoid
18758 pipeline stalls for always-taken or probably-taken branches. A hint
18759 is not generated closer than 8 instructions away from its branch.
18760 There is little reason to disable them, except for debugging purposes,
18761 or to make an object a little bit smaller.
18765 @opindex msmall-mem
18766 @opindex mlarge-mem
18768 By default, GCC generates code assuming that addresses are never larger
18769 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
18770 a full 32-bit address.
18775 By default, GCC links against startup code that assumes the SPU-style
18776 main function interface (which has an unconventional parameter list).
18777 With @option{-mstdmain}, GCC links your program against startup
18778 code that assumes a C99-style interface to @code{main}, including a
18779 local copy of @code{argv} strings.
18781 @item -mfixed-range=@var{register-range}
18782 @opindex mfixed-range
18783 Generate code treating the given register range as fixed registers.
18784 A fixed register is one that the register allocator can not use. This is
18785 useful when compiling kernel code. A register range is specified as
18786 two registers separated by a dash. Multiple register ranges can be
18787 specified separated by a comma.
18793 Compile code assuming that pointers to the PPU address space accessed
18794 via the @code{__ea} named address space qualifier are either 32 or 64
18795 bits wide. The default is 32 bits. As this is an ABI changing option,
18796 all object code in an executable must be compiled with the same setting.
18798 @item -maddress-space-conversion
18799 @itemx -mno-address-space-conversion
18800 @opindex maddress-space-conversion
18801 @opindex mno-address-space-conversion
18802 Allow/disallow treating the @code{__ea} address space as superset
18803 of the generic address space. This enables explicit type casts
18804 between @code{__ea} and generic pointer as well as implicit
18805 conversions of generic pointers to @code{__ea} pointers. The
18806 default is to allow address space pointer conversions.
18808 @item -mcache-size=@var{cache-size}
18809 @opindex mcache-size
18810 This option controls the version of libgcc that the compiler links to an
18811 executable and selects a software-managed cache for accessing variables
18812 in the @code{__ea} address space with a particular cache size. Possible
18813 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
18814 and @samp{128}. The default cache size is 64KB.
18816 @item -matomic-updates
18817 @itemx -mno-atomic-updates
18818 @opindex matomic-updates
18819 @opindex mno-atomic-updates
18820 This option controls the version of libgcc that the compiler links to an
18821 executable and selects whether atomic updates to the software-managed
18822 cache of PPU-side variables are used. If you use atomic updates, changes
18823 to a PPU variable from SPU code using the @code{__ea} named address space
18824 qualifier do not interfere with changes to other PPU variables residing
18825 in the same cache line from PPU code. If you do not use atomic updates,
18826 such interference may occur; however, writing back cache lines is
18827 more efficient. The default behavior is to use atomic updates.
18830 @itemx -mdual-nops=@var{n}
18831 @opindex mdual-nops
18832 By default, GCC inserts nops to increase dual issue when it expects
18833 it to increase performance. @var{n} can be a value from 0 to 10. A
18834 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
18835 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
18837 @item -mhint-max-nops=@var{n}
18838 @opindex mhint-max-nops
18839 Maximum number of nops to insert for a branch hint. A branch hint must
18840 be at least 8 instructions away from the branch it is affecting. GCC
18841 inserts up to @var{n} nops to enforce this, otherwise it does not
18842 generate the branch hint.
18844 @item -mhint-max-distance=@var{n}
18845 @opindex mhint-max-distance
18846 The encoding of the branch hint instruction limits the hint to be within
18847 256 instructions of the branch it is affecting. By default, GCC makes
18848 sure it is within 125.
18851 @opindex msafe-hints
18852 Work around a hardware bug that causes the SPU to stall indefinitely.
18853 By default, GCC inserts the @code{hbrp} instruction to make sure
18854 this stall won't happen.
18858 @node System V Options
18859 @subsection Options for System V
18861 These additional options are available on System V Release 4 for
18862 compatibility with other compilers on those systems:
18867 Create a shared object.
18868 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
18872 Identify the versions of each tool used by the compiler, in a
18873 @code{.ident} assembler directive in the output.
18877 Refrain from adding @code{.ident} directives to the output file (this is
18880 @item -YP,@var{dirs}
18882 Search the directories @var{dirs}, and no others, for libraries
18883 specified with @option{-l}.
18885 @item -Ym,@var{dir}
18887 Look in the directory @var{dir} to find the M4 preprocessor.
18888 The assembler uses this option.
18889 @c This is supposed to go with a -Yd for predefined M4 macro files, but
18890 @c the generic assembler that comes with Solaris takes just -Ym.
18893 @node TILE-Gx Options
18894 @subsection TILE-Gx Options
18895 @cindex TILE-Gx options
18897 These @samp{-m} options are supported on the TILE-Gx:
18900 @item -mcpu=@var{name}
18902 Selects the type of CPU to be targeted. Currently the only supported
18903 type is @samp{tilegx}.
18909 Generate code for a 32-bit or 64-bit environment. The 32-bit
18910 environment sets int, long, and pointer to 32 bits. The 64-bit
18911 environment sets int to 32 bits and long and pointer to 64 bits.
18914 @node TILEPro Options
18915 @subsection TILEPro Options
18916 @cindex TILEPro options
18918 These @samp{-m} options are supported on the TILEPro:
18921 @item -mcpu=@var{name}
18923 Selects the type of CPU to be targeted. Currently the only supported
18924 type is @samp{tilepro}.
18928 Generate code for a 32-bit environment, which sets int, long, and
18929 pointer to 32 bits. This is the only supported behavior so the flag
18930 is essentially ignored.
18934 @subsection V850 Options
18935 @cindex V850 Options
18937 These @samp{-m} options are defined for V850 implementations:
18941 @itemx -mno-long-calls
18942 @opindex mlong-calls
18943 @opindex mno-long-calls
18944 Treat all calls as being far away (near). If calls are assumed to be
18945 far away, the compiler always loads the function's address into a
18946 register, and calls indirect through the pointer.
18952 Do not optimize (do optimize) basic blocks that use the same index
18953 pointer 4 or more times to copy pointer into the @code{ep} register, and
18954 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
18955 option is on by default if you optimize.
18957 @item -mno-prolog-function
18958 @itemx -mprolog-function
18959 @opindex mno-prolog-function
18960 @opindex mprolog-function
18961 Do not use (do use) external functions to save and restore registers
18962 at the prologue and epilogue of a function. The external functions
18963 are slower, but use less code space if more than one function saves
18964 the same number of registers. The @option{-mprolog-function} option
18965 is on by default if you optimize.
18969 Try to make the code as small as possible. At present, this just turns
18970 on the @option{-mep} and @option{-mprolog-function} options.
18972 @item -mtda=@var{n}
18974 Put static or global variables whose size is @var{n} bytes or less into
18975 the tiny data area that register @code{ep} points to. The tiny data
18976 area can hold up to 256 bytes in total (128 bytes for byte references).
18978 @item -msda=@var{n}
18980 Put static or global variables whose size is @var{n} bytes or less into
18981 the small data area that register @code{gp} points to. The small data
18982 area can hold up to 64 kilobytes.
18984 @item -mzda=@var{n}
18986 Put static or global variables whose size is @var{n} bytes or less into
18987 the first 32 kilobytes of memory.
18991 Specify that the target processor is the V850.
18994 @opindex mbig-switch
18995 Generate code suitable for big switch tables. Use this option only if
18996 the assembler/linker complain about out of range branches within a switch
19001 This option causes r2 and r5 to be used in the code generated by
19002 the compiler. This setting is the default.
19004 @item -mno-app-regs
19005 @opindex mno-app-regs
19006 This option causes r2 and r5 to be treated as fixed registers.
19010 Specify that the target processor is the V850E2V3. The preprocessor
19011 constant @samp{__v850e2v3__} is defined if
19012 this option is used.
19016 Specify that the target processor is the V850E2. The preprocessor
19017 constant @samp{__v850e2__} is defined if this option is used.
19021 Specify that the target processor is the V850E1. The preprocessor
19022 constants @samp{__v850e1__} and @samp{__v850e__} are defined if
19023 this option is used.
19027 Specify that the target processor is the V850ES. This is an alias for
19028 the @option{-mv850e1} option.
19032 Specify that the target processor is the V850E@. The preprocessor
19033 constant @samp{__v850e__} is defined if this option is used.
19035 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
19036 nor @option{-mv850e2} nor @option{-mv850e2v3}
19037 are defined then a default target processor is chosen and the
19038 relevant @samp{__v850*__} preprocessor constant is defined.
19040 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
19041 defined, regardless of which processor variant is the target.
19043 @item -mdisable-callt
19044 @opindex mdisable-callt
19045 This option suppresses generation of the @code{CALLT} instruction for the
19046 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
19047 @option{-mno-disable-callt} which allows the @code{CALLT} instruction to be used.
19052 @subsection VAX Options
19053 @cindex VAX options
19055 These @samp{-m} options are defined for the VAX:
19060 Do not output certain jump instructions (@code{aobleq} and so on)
19061 that the Unix assembler for the VAX cannot handle across long
19066 Do output those jump instructions, on the assumption that the
19067 GNU assembler is being used.
19071 Output code for G-format floating-point numbers instead of D-format.
19075 @subsection VMS Options
19077 These @samp{-m} options are defined for the VMS implementations:
19080 @item -mvms-return-codes
19081 @opindex mvms-return-codes
19082 Return VMS condition codes from @code{main}. The default is to return POSIX-style
19083 condition (e.g.@ error) codes.
19085 @item -mdebug-main=@var{prefix}
19086 @opindex mdebug-main=@var{prefix}
19087 Flag the first routine whose name starts with @var{prefix} as the main
19088 routine for the debugger.
19092 Default to 64-bit memory allocation routines.
19094 @item -mpointer-size=@var{size}
19095 @opindex -mpointer-size=@var{size}
19096 Set the default size of pointers. Possible options for @var{size} are
19097 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
19098 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
19099 The later option disables @code{pragma pointer_size}.
19102 @node VxWorks Options
19103 @subsection VxWorks Options
19104 @cindex VxWorks Options
19106 The options in this section are defined for all VxWorks targets.
19107 Options specific to the target hardware are listed with the other
19108 options for that target.
19113 GCC can generate code for both VxWorks kernels and real time processes
19114 (RTPs). This option switches from the former to the latter. It also
19115 defines the preprocessor macro @code{__RTP__}.
19118 @opindex non-static
19119 Link an RTP executable against shared libraries rather than static
19120 libraries. The options @option{-static} and @option{-shared} can
19121 also be used for RTPs (@pxref{Link Options}); @option{-static}
19128 These options are passed down to the linker. They are defined for
19129 compatibility with Diab.
19132 @opindex Xbind-lazy
19133 Enable lazy binding of function calls. This option is equivalent to
19134 @option{-Wl,-z,now} and is defined for compatibility with Diab.
19138 Disable lazy binding of function calls. This option is the default and
19139 is defined for compatibility with Diab.
19142 @node x86-64 Options
19143 @subsection x86-64 Options
19144 @cindex x86-64 options
19146 These are listed under @xref{i386 and x86-64 Options}.
19148 @node Xstormy16 Options
19149 @subsection Xstormy16 Options
19150 @cindex Xstormy16 Options
19152 These options are defined for Xstormy16:
19157 Choose startup files and linker script suitable for the simulator.
19160 @node Xtensa Options
19161 @subsection Xtensa Options
19162 @cindex Xtensa Options
19164 These options are supported for Xtensa targets:
19168 @itemx -mno-const16
19170 @opindex mno-const16
19171 Enable or disable use of @code{CONST16} instructions for loading
19172 constant values. The @code{CONST16} instruction is currently not a
19173 standard option from Tensilica. When enabled, @code{CONST16}
19174 instructions are always used in place of the standard @code{L32R}
19175 instructions. The use of @code{CONST16} is enabled by default only if
19176 the @code{L32R} instruction is not available.
19179 @itemx -mno-fused-madd
19180 @opindex mfused-madd
19181 @opindex mno-fused-madd
19182 Enable or disable use of fused multiply/add and multiply/subtract
19183 instructions in the floating-point option. This has no effect if the
19184 floating-point option is not also enabled. Disabling fused multiply/add
19185 and multiply/subtract instructions forces the compiler to use separate
19186 instructions for the multiply and add/subtract operations. This may be
19187 desirable in some cases where strict IEEE 754-compliant results are
19188 required: the fused multiply add/subtract instructions do not round the
19189 intermediate result, thereby producing results with @emph{more} bits of
19190 precision than specified by the IEEE standard. Disabling fused multiply
19191 add/subtract instructions also ensures that the program output is not
19192 sensitive to the compiler's ability to combine multiply and add/subtract
19195 @item -mserialize-volatile
19196 @itemx -mno-serialize-volatile
19197 @opindex mserialize-volatile
19198 @opindex mno-serialize-volatile
19199 When this option is enabled, GCC inserts @code{MEMW} instructions before
19200 @code{volatile} memory references to guarantee sequential consistency.
19201 The default is @option{-mserialize-volatile}. Use
19202 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
19204 @item -mforce-no-pic
19205 @opindex mforce-no-pic
19206 For targets, like GNU/Linux, where all user-mode Xtensa code must be
19207 position-independent code (PIC), this option disables PIC for compiling
19210 @item -mtext-section-literals
19211 @itemx -mno-text-section-literals
19212 @opindex mtext-section-literals
19213 @opindex mno-text-section-literals
19214 Control the treatment of literal pools. The default is
19215 @option{-mno-text-section-literals}, which places literals in a separate
19216 section in the output file. This allows the literal pool to be placed
19217 in a data RAM/ROM, and it also allows the linker to combine literal
19218 pools from separate object files to remove redundant literals and
19219 improve code size. With @option{-mtext-section-literals}, the literals
19220 are interspersed in the text section in order to keep them as close as
19221 possible to their references. This may be necessary for large assembly
19224 @item -mtarget-align
19225 @itemx -mno-target-align
19226 @opindex mtarget-align
19227 @opindex mno-target-align
19228 When this option is enabled, GCC instructs the assembler to
19229 automatically align instructions to reduce branch penalties at the
19230 expense of some code density. The assembler attempts to widen density
19231 instructions to align branch targets and the instructions following call
19232 instructions. If there are not enough preceding safe density
19233 instructions to align a target, no widening is performed. The
19234 default is @option{-mtarget-align}. These options do not affect the
19235 treatment of auto-aligned instructions like @code{LOOP}, which the
19236 assembler always aligns, either by widening density instructions or
19237 by inserting NOP instructions.
19240 @itemx -mno-longcalls
19241 @opindex mlongcalls
19242 @opindex mno-longcalls
19243 When this option is enabled, GCC instructs the assembler to translate
19244 direct calls to indirect calls unless it can determine that the target
19245 of a direct call is in the range allowed by the call instruction. This
19246 translation typically occurs for calls to functions in other source
19247 files. Specifically, the assembler translates a direct @code{CALL}
19248 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
19249 The default is @option{-mno-longcalls}. This option should be used in
19250 programs where the call target can potentially be out of range. This
19251 option is implemented in the assembler, not the compiler, so the
19252 assembly code generated by GCC still shows direct call
19253 instructions---look at the disassembled object code to see the actual
19254 instructions. Note that the assembler uses an indirect call for
19255 every cross-file call, not just those that really are out of range.
19258 @node zSeries Options
19259 @subsection zSeries Options
19260 @cindex zSeries options
19262 These are listed under @xref{S/390 and zSeries Options}.
19264 @node Code Gen Options
19265 @section Options for Code Generation Conventions
19266 @cindex code generation conventions
19267 @cindex options, code generation
19268 @cindex run-time options
19270 These machine-independent options control the interface conventions
19271 used in code generation.
19273 Most of them have both positive and negative forms; the negative form
19274 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
19275 one of the forms is listed---the one that is not the default. You
19276 can figure out the other form by either removing @samp{no-} or adding
19280 @item -fbounds-check
19281 @opindex fbounds-check
19282 For front ends that support it, generate additional code to check that
19283 indices used to access arrays are within the declared range. This is
19284 currently only supported by the Java and Fortran front ends, where
19285 this option defaults to true and false respectively.
19289 This option generates traps for signed overflow on addition, subtraction,
19290 multiplication operations.
19294 This option instructs the compiler to assume that signed arithmetic
19295 overflow of addition, subtraction and multiplication wraps around
19296 using twos-complement representation. This flag enables some optimizations
19297 and disables others. This option is enabled by default for the Java
19298 front end, as required by the Java language specification.
19301 @opindex fexceptions
19302 Enable exception handling. Generates extra code needed to propagate
19303 exceptions. For some targets, this implies GCC generates frame
19304 unwind information for all functions, which can produce significant data
19305 size overhead, although it does not affect execution. If you do not
19306 specify this option, GCC enables it by default for languages like
19307 C++ that normally require exception handling, and disables it for
19308 languages like C that do not normally require it. However, you may need
19309 to enable this option when compiling C code that needs to interoperate
19310 properly with exception handlers written in C++. You may also wish to
19311 disable this option if you are compiling older C++ programs that don't
19312 use exception handling.
19314 @item -fnon-call-exceptions
19315 @opindex fnon-call-exceptions
19316 Generate code that allows trapping instructions to throw exceptions.
19317 Note that this requires platform-specific runtime support that does
19318 not exist everywhere. Moreover, it only allows @emph{trapping}
19319 instructions to throw exceptions, i.e.@: memory references or floating-point
19320 instructions. It does not allow exceptions to be thrown from
19321 arbitrary signal handlers such as @code{SIGALRM}.
19323 @item -funwind-tables
19324 @opindex funwind-tables
19325 Similar to @option{-fexceptions}, except that it just generates any needed
19326 static data, but does not affect the generated code in any other way.
19327 You normally do not need to enable this option; instead, a language processor
19328 that needs this handling enables it on your behalf.
19330 @item -fasynchronous-unwind-tables
19331 @opindex fasynchronous-unwind-tables
19332 Generate unwind table in DWARF 2 format, if supported by target machine. The
19333 table is exact at each instruction boundary, so it can be used for stack
19334 unwinding from asynchronous events (such as debugger or garbage collector).
19336 @item -fpcc-struct-return
19337 @opindex fpcc-struct-return
19338 Return ``short'' @code{struct} and @code{union} values in memory like
19339 longer ones, rather than in registers. This convention is less
19340 efficient, but it has the advantage of allowing intercallability between
19341 GCC-compiled files and files compiled with other compilers, particularly
19342 the Portable C Compiler (pcc).
19344 The precise convention for returning structures in memory depends
19345 on the target configuration macros.
19347 Short structures and unions are those whose size and alignment match
19348 that of some integer type.
19350 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
19351 switch is not binary compatible with code compiled with the
19352 @option{-freg-struct-return} switch.
19353 Use it to conform to a non-default application binary interface.
19355 @item -freg-struct-return
19356 @opindex freg-struct-return
19357 Return @code{struct} and @code{union} values in registers when possible.
19358 This is more efficient for small structures than
19359 @option{-fpcc-struct-return}.
19361 If you specify neither @option{-fpcc-struct-return} nor
19362 @option{-freg-struct-return}, GCC defaults to whichever convention is
19363 standard for the target. If there is no standard convention, GCC
19364 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
19365 the principal compiler. In those cases, we can choose the standard, and
19366 we chose the more efficient register return alternative.
19368 @strong{Warning:} code compiled with the @option{-freg-struct-return}
19369 switch is not binary compatible with code compiled with the
19370 @option{-fpcc-struct-return} switch.
19371 Use it to conform to a non-default application binary interface.
19373 @item -fshort-enums
19374 @opindex fshort-enums
19375 Allocate to an @code{enum} type only as many bytes as it needs for the
19376 declared range of possible values. Specifically, the @code{enum} type
19377 is equivalent to the smallest integer type that has enough room.
19379 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
19380 code that is not binary compatible with code generated without that switch.
19381 Use it to conform to a non-default application binary interface.
19383 @item -fshort-double
19384 @opindex fshort-double
19385 Use the same size for @code{double} as for @code{float}.
19387 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
19388 code that is not binary compatible with code generated without that switch.
19389 Use it to conform to a non-default application binary interface.
19391 @item -fshort-wchar
19392 @opindex fshort-wchar
19393 Override the underlying type for @samp{wchar_t} to be @samp{short
19394 unsigned int} instead of the default for the target. This option is
19395 useful for building programs to run under WINE@.
19397 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
19398 code that is not binary compatible with code generated without that switch.
19399 Use it to conform to a non-default application binary interface.
19402 @opindex fno-common
19403 In C code, controls the placement of uninitialized global variables.
19404 Unix C compilers have traditionally permitted multiple definitions of
19405 such variables in different compilation units by placing the variables
19407 This is the behavior specified by @option{-fcommon}, and is the default
19408 for GCC on most targets.
19409 On the other hand, this behavior is not required by ISO C, and on some
19410 targets may carry a speed or code size penalty on variable references.
19411 The @option{-fno-common} option specifies that the compiler should place
19412 uninitialized global variables in the data section of the object file,
19413 rather than generating them as common blocks.
19414 This has the effect that if the same variable is declared
19415 (without @code{extern}) in two different compilations,
19416 you get a multiple-definition error when you link them.
19417 In this case, you must compile with @option{-fcommon} instead.
19418 Compiling with @option{-fno-common} is useful on targets for which
19419 it provides better performance, or if you wish to verify that the
19420 program will work on other systems that always treat uninitialized
19421 variable declarations this way.
19425 Ignore the @samp{#ident} directive.
19427 @item -finhibit-size-directive
19428 @opindex finhibit-size-directive
19429 Don't output a @code{.size} assembler directive, or anything else that
19430 would cause trouble if the function is split in the middle, and the
19431 two halves are placed at locations far apart in memory. This option is
19432 used when compiling @file{crtstuff.c}; you should not need to use it
19435 @item -fverbose-asm
19436 @opindex fverbose-asm
19437 Put extra commentary information in the generated assembly code to
19438 make it more readable. This option is generally only of use to those
19439 who actually need to read the generated assembly code (perhaps while
19440 debugging the compiler itself).
19442 @option{-fno-verbose-asm}, the default, causes the
19443 extra information to be omitted and is useful when comparing two assembler
19446 @item -frecord-gcc-switches
19447 @opindex frecord-gcc-switches
19448 This switch causes the command line used to invoke the
19449 compiler to be recorded into the object file that is being created.
19450 This switch is only implemented on some targets and the exact format
19451 of the recording is target and binary file format dependent, but it
19452 usually takes the form of a section containing ASCII text. This
19453 switch is related to the @option{-fverbose-asm} switch, but that
19454 switch only records information in the assembler output file as
19455 comments, so it never reaches the object file.
19456 See also @option{-grecord-gcc-switches} for another
19457 way of storing compiler options into the object file.
19461 @cindex global offset table
19463 Generate position-independent code (PIC) suitable for use in a shared
19464 library, if supported for the target machine. Such code accesses all
19465 constant addresses through a global offset table (GOT)@. The dynamic
19466 loader resolves the GOT entries when the program starts (the dynamic
19467 loader is not part of GCC; it is part of the operating system). If
19468 the GOT size for the linked executable exceeds a machine-specific
19469 maximum size, you get an error message from the linker indicating that
19470 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
19471 instead. (These maximums are 8k on the SPARC and 32k
19472 on the m68k and RS/6000. The 386 has no such limit.)
19474 Position-independent code requires special support, and therefore works
19475 only on certain machines. For the 386, GCC supports PIC for System V
19476 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
19477 position-independent.
19479 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19484 If supported for the target machine, emit position-independent code,
19485 suitable for dynamic linking and avoiding any limit on the size of the
19486 global offset table. This option makes a difference on the m68k,
19487 PowerPC and SPARC@.
19489 Position-independent code requires special support, and therefore works
19490 only on certain machines.
19492 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19499 These options are similar to @option{-fpic} and @option{-fPIC}, but
19500 generated position independent code can be only linked into executables.
19501 Usually these options are used when @option{-pie} GCC option is
19502 used during linking.
19504 @option{-fpie} and @option{-fPIE} both define the macros
19505 @code{__pie__} and @code{__PIE__}. The macros have the value 1
19506 for @option{-fpie} and 2 for @option{-fPIE}.
19508 @item -fno-jump-tables
19509 @opindex fno-jump-tables
19510 Do not use jump tables for switch statements even where it would be
19511 more efficient than other code generation strategies. This option is
19512 of use in conjunction with @option{-fpic} or @option{-fPIC} for
19513 building code that forms part of a dynamic linker and cannot
19514 reference the address of a jump table. On some targets, jump tables
19515 do not require a GOT and this option is not needed.
19517 @item -ffixed-@var{reg}
19519 Treat the register named @var{reg} as a fixed register; generated code
19520 should never refer to it (except perhaps as a stack pointer, frame
19521 pointer or in some other fixed role).
19523 @var{reg} must be the name of a register. The register names accepted
19524 are machine-specific and are defined in the @code{REGISTER_NAMES}
19525 macro in the machine description macro file.
19527 This flag does not have a negative form, because it specifies a
19530 @item -fcall-used-@var{reg}
19531 @opindex fcall-used
19532 Treat the register named @var{reg} as an allocable register that is
19533 clobbered by function calls. It may be allocated for temporaries or
19534 variables that do not live across a call. Functions compiled this way
19535 do not save and restore the register @var{reg}.
19537 It is an error to use this flag with the frame pointer or stack pointer.
19538 Use of this flag for other registers that have fixed pervasive roles in
19539 the machine's execution model produces disastrous results.
19541 This flag does not have a negative form, because it specifies a
19544 @item -fcall-saved-@var{reg}
19545 @opindex fcall-saved
19546 Treat the register named @var{reg} as an allocable register saved by
19547 functions. It may be allocated even for temporaries or variables that
19548 live across a call. Functions compiled this way save and restore
19549 the register @var{reg} if they use it.
19551 It is an error to use this flag with the frame pointer or stack pointer.
19552 Use of this flag for other registers that have fixed pervasive roles in
19553 the machine's execution model produces disastrous results.
19555 A different sort of disaster results from the use of this flag for
19556 a register in which function values may be returned.
19558 This flag does not have a negative form, because it specifies a
19561 @item -fpack-struct[=@var{n}]
19562 @opindex fpack-struct
19563 Without a value specified, pack all structure members together without
19564 holes. When a value is specified (which must be a small power of two), pack
19565 structure members according to this value, representing the maximum
19566 alignment (that is, objects with default alignment requirements larger than
19567 this are output potentially unaligned at the next fitting location.
19569 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
19570 code that is not binary compatible with code generated without that switch.
19571 Additionally, it makes the code suboptimal.
19572 Use it to conform to a non-default application binary interface.
19574 @item -finstrument-functions
19575 @opindex finstrument-functions
19576 Generate instrumentation calls for entry and exit to functions. Just
19577 after function entry and just before function exit, the following
19578 profiling functions are called with the address of the current
19579 function and its call site. (On some platforms,
19580 @code{__builtin_return_address} does not work beyond the current
19581 function, so the call site information may not be available to the
19582 profiling functions otherwise.)
19585 void __cyg_profile_func_enter (void *this_fn,
19587 void __cyg_profile_func_exit (void *this_fn,
19591 The first argument is the address of the start of the current function,
19592 which may be looked up exactly in the symbol table.
19594 This instrumentation is also done for functions expanded inline in other
19595 functions. The profiling calls indicate where, conceptually, the
19596 inline function is entered and exited. This means that addressable
19597 versions of such functions must be available. If all your uses of a
19598 function are expanded inline, this may mean an additional expansion of
19599 code size. If you use @samp{extern inline} in your C code, an
19600 addressable version of such functions must be provided. (This is
19601 normally the case anyway, but if you get lucky and the optimizer always
19602 expands the functions inline, you might have gotten away without
19603 providing static copies.)
19605 A function may be given the attribute @code{no_instrument_function}, in
19606 which case this instrumentation is not done. This can be used, for
19607 example, for the profiling functions listed above, high-priority
19608 interrupt routines, and any functions from which the profiling functions
19609 cannot safely be called (perhaps signal handlers, if the profiling
19610 routines generate output or allocate memory).
19612 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
19613 @opindex finstrument-functions-exclude-file-list
19615 Set the list of functions that are excluded from instrumentation (see
19616 the description of @code{-finstrument-functions}). If the file that
19617 contains a function definition matches with one of @var{file}, then
19618 that function is not instrumented. The match is done on substrings:
19619 if the @var{file} parameter is a substring of the file name, it is
19620 considered to be a match.
19625 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
19629 excludes any inline function defined in files whose pathnames
19630 contain @code{/bits/stl} or @code{include/sys}.
19632 If, for some reason, you want to include letter @code{','} in one of
19633 @var{sym}, write @code{'\,'}. For example,
19634 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
19635 (note the single quote surrounding the option).
19637 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
19638 @opindex finstrument-functions-exclude-function-list
19640 This is similar to @code{-finstrument-functions-exclude-file-list},
19641 but this option sets the list of function names to be excluded from
19642 instrumentation. The function name to be matched is its user-visible
19643 name, such as @code{vector<int> blah(const vector<int> &)}, not the
19644 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
19645 match is done on substrings: if the @var{sym} parameter is a substring
19646 of the function name, it is considered to be a match. For C99 and C++
19647 extended identifiers, the function name must be given in UTF-8, not
19648 using universal character names.
19650 @item -fstack-check
19651 @opindex fstack-check
19652 Generate code to verify that you do not go beyond the boundary of the
19653 stack. You should specify this flag if you are running in an
19654 environment with multiple threads, but only rarely need to specify it in
19655 a single-threaded environment since stack overflow is automatically
19656 detected on nearly all systems if there is only one stack.
19658 Note that this switch does not actually cause checking to be done; the
19659 operating system or the language runtime must do that. The switch causes
19660 generation of code to ensure that they see the stack being extended.
19662 You can additionally specify a string parameter: @code{no} means no
19663 checking, @code{generic} means force the use of old-style checking,
19664 @code{specific} means use the best checking method and is equivalent
19665 to bare @option{-fstack-check}.
19667 Old-style checking is a generic mechanism that requires no specific
19668 target support in the compiler but comes with the following drawbacks:
19672 Modified allocation strategy for large objects: they are always
19673 allocated dynamically if their size exceeds a fixed threshold.
19676 Fixed limit on the size of the static frame of functions: when it is
19677 topped by a particular function, stack checking is not reliable and
19678 a warning is issued by the compiler.
19681 Inefficiency: because of both the modified allocation strategy and the
19682 generic implementation, the performances of the code are hampered.
19685 Note that old-style stack checking is also the fallback method for
19686 @code{specific} if no target support has been added in the compiler.
19688 @item -fstack-limit-register=@var{reg}
19689 @itemx -fstack-limit-symbol=@var{sym}
19690 @itemx -fno-stack-limit
19691 @opindex fstack-limit-register
19692 @opindex fstack-limit-symbol
19693 @opindex fno-stack-limit
19694 Generate code to ensure that the stack does not grow beyond a certain value,
19695 either the value of a register or the address of a symbol. If a larger
19696 stack is required, a signal is raised at run time. For most targets,
19697 the signal is raised before the stack overruns the boundary, so
19698 it is possible to catch the signal without taking special precautions.
19700 For instance, if the stack starts at absolute address @samp{0x80000000}
19701 and grows downwards, you can use the flags
19702 @option{-fstack-limit-symbol=__stack_limit} and
19703 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
19704 of 128KB@. Note that this may only work with the GNU linker.
19706 @item -fsplit-stack
19707 @opindex fsplit-stack
19708 Generate code to automatically split the stack before it overflows.
19709 The resulting program has a discontiguous stack which can only
19710 overflow if the program is unable to allocate any more memory. This
19711 is most useful when running threaded programs, as it is no longer
19712 necessary to calculate a good stack size to use for each thread. This
19713 is currently only implemented for the i386 and x86_64 back ends running
19716 When code compiled with @option{-fsplit-stack} calls code compiled
19717 without @option{-fsplit-stack}, there may not be much stack space
19718 available for the latter code to run. If compiling all code,
19719 including library code, with @option{-fsplit-stack} is not an option,
19720 then the linker can fix up these calls so that the code compiled
19721 without @option{-fsplit-stack} always has a large stack. Support for
19722 this is implemented in the gold linker in GNU binutils release 2.21
19725 @item -fleading-underscore
19726 @opindex fleading-underscore
19727 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
19728 change the way C symbols are represented in the object file. One use
19729 is to help link with legacy assembly code.
19731 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
19732 generate code that is not binary compatible with code generated without that
19733 switch. Use it to conform to a non-default application binary interface.
19734 Not all targets provide complete support for this switch.
19736 @item -ftls-model=@var{model}
19737 @opindex ftls-model
19738 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
19739 The @var{model} argument should be one of @code{global-dynamic},
19740 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
19742 The default without @option{-fpic} is @code{initial-exec}; with
19743 @option{-fpic} the default is @code{global-dynamic}.
19745 @item -fvisibility=@var{default|internal|hidden|protected}
19746 @opindex fvisibility
19747 Set the default ELF image symbol visibility to the specified option---all
19748 symbols are marked with this unless overridden within the code.
19749 Using this feature can very substantially improve linking and
19750 load times of shared object libraries, produce more optimized
19751 code, provide near-perfect API export and prevent symbol clashes.
19752 It is @strong{strongly} recommended that you use this in any shared objects
19755 Despite the nomenclature, @code{default} always means public; i.e.,
19756 available to be linked against from outside the shared object.
19757 @code{protected} and @code{internal} are pretty useless in real-world
19758 usage so the only other commonly used option is @code{hidden}.
19759 The default if @option{-fvisibility} isn't specified is
19760 @code{default}, i.e., make every
19761 symbol public---this causes the same behavior as previous versions of
19764 A good explanation of the benefits offered by ensuring ELF
19765 symbols have the correct visibility is given by ``How To Write
19766 Shared Libraries'' by Ulrich Drepper (which can be found at
19767 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
19768 solution made possible by this option to marking things hidden when
19769 the default is public is to make the default hidden and mark things
19770 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
19771 and @code{__attribute__ ((visibility("default")))} instead of
19772 @code{__declspec(dllexport)} you get almost identical semantics with
19773 identical syntax. This is a great boon to those working with
19774 cross-platform projects.
19776 For those adding visibility support to existing code, you may find
19777 @samp{#pragma GCC visibility} of use. This works by you enclosing
19778 the declarations you wish to set visibility for with (for example)
19779 @samp{#pragma GCC visibility push(hidden)} and
19780 @samp{#pragma GCC visibility pop}.
19781 Bear in mind that symbol visibility should be viewed @strong{as
19782 part of the API interface contract} and thus all new code should
19783 always specify visibility when it is not the default; i.e., declarations
19784 only for use within the local DSO should @strong{always} be marked explicitly
19785 as hidden as so to avoid PLT indirection overheads---making this
19786 abundantly clear also aids readability and self-documentation of the code.
19787 Note that due to ISO C++ specification requirements, @code{operator new} and
19788 @code{operator delete} must always be of default visibility.
19790 Be aware that headers from outside your project, in particular system
19791 headers and headers from any other library you use, may not be
19792 expecting to be compiled with visibility other than the default. You
19793 may need to explicitly say @samp{#pragma GCC visibility push(default)}
19794 before including any such headers.
19796 @samp{extern} declarations are not affected by @option{-fvisibility}, so
19797 a lot of code can be recompiled with @option{-fvisibility=hidden} with
19798 no modifications. However, this means that calls to @code{extern}
19799 functions with no explicit visibility use the PLT, so it is more
19800 effective to use @code{__attribute ((visibility))} and/or
19801 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
19802 declarations should be treated as hidden.
19804 Note that @option{-fvisibility} does affect C++ vague linkage
19805 entities. This means that, for instance, an exception class that is
19806 be thrown between DSOs must be explicitly marked with default
19807 visibility so that the @samp{type_info} nodes are unified between
19810 An overview of these techniques, their benefits and how to use them
19811 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
19813 @item -fstrict-volatile-bitfields
19814 @opindex fstrict-volatile-bitfields
19815 This option should be used if accesses to volatile bit-fields (or other
19816 structure fields, although the compiler usually honors those types
19817 anyway) should use a single access of the width of the
19818 field's type, aligned to a natural alignment if possible. For
19819 example, targets with memory-mapped peripheral registers might require
19820 all such accesses to be 16 bits wide; with this flag the user could
19821 declare all peripheral bit-fields as @code{unsigned short} (assuming short
19822 is 16 bits on these targets) to force GCC to use 16-bit accesses
19823 instead of, perhaps, a more efficient 32-bit access.
19825 If this option is disabled, the compiler uses the most efficient
19826 instruction. In the previous example, that might be a 32-bit load
19827 instruction, even though that accesses bytes that do not contain
19828 any portion of the bit-field, or memory-mapped registers unrelated to
19829 the one being updated.
19831 If the target requires strict alignment, and honoring the field
19832 type would require violating this alignment, a warning is issued.
19833 If the field has @code{packed} attribute, the access is done without
19834 honoring the field type. If the field doesn't have @code{packed}
19835 attribute, the access is done honoring the field type. In both cases,
19836 GCC assumes that the user knows something about the target hardware
19837 that it is unaware of.
19839 The default value of this option is determined by the application binary
19840 interface for the target processor.
19842 @item -fsync-libcalls
19843 @opindex fsync-libcalls
19844 This option controls whether any out-of-line instance of the @code{__sync}
19845 family of functions may be used to implement the C++11 @code{__atomic}
19846 family of functions.
19848 The default value of this option is enabled, thus the only useful form
19849 of the option is @option{-fno-sync-libcalls}. This option is used in
19850 the implementation of the @file{libatomic} runtime library.
19856 @node Environment Variables
19857 @section Environment Variables Affecting GCC
19858 @cindex environment variables
19860 @c man begin ENVIRONMENT
19861 This section describes several environment variables that affect how GCC
19862 operates. Some of them work by specifying directories or prefixes to use
19863 when searching for various kinds of files. Some are used to specify other
19864 aspects of the compilation environment.
19866 Note that you can also specify places to search using options such as
19867 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
19868 take precedence over places specified using environment variables, which
19869 in turn take precedence over those specified by the configuration of GCC@.
19870 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
19871 GNU Compiler Collection (GCC) Internals}.
19876 @c @itemx LC_COLLATE
19878 @c @itemx LC_MONETARY
19879 @c @itemx LC_NUMERIC
19884 @c @findex LC_COLLATE
19885 @findex LC_MESSAGES
19886 @c @findex LC_MONETARY
19887 @c @findex LC_NUMERIC
19891 These environment variables control the way that GCC uses
19892 localization information which allows GCC to work with different
19893 national conventions. GCC inspects the locale categories
19894 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
19895 so. These locale categories can be set to any value supported by your
19896 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
19897 Kingdom encoded in UTF-8.
19899 The @env{LC_CTYPE} environment variable specifies character
19900 classification. GCC uses it to determine the character boundaries in
19901 a string; this is needed for some multibyte encodings that contain quote
19902 and escape characters that are otherwise interpreted as a string
19905 The @env{LC_MESSAGES} environment variable specifies the language to
19906 use in diagnostic messages.
19908 If the @env{LC_ALL} environment variable is set, it overrides the value
19909 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
19910 and @env{LC_MESSAGES} default to the value of the @env{LANG}
19911 environment variable. If none of these variables are set, GCC
19912 defaults to traditional C English behavior.
19916 If @env{TMPDIR} is set, it specifies the directory to use for temporary
19917 files. GCC uses temporary files to hold the output of one stage of
19918 compilation which is to be used as input to the next stage: for example,
19919 the output of the preprocessor, which is the input to the compiler
19922 @item GCC_COMPARE_DEBUG
19923 @findex GCC_COMPARE_DEBUG
19924 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
19925 @option{-fcompare-debug} to the compiler driver. See the documentation
19926 of this option for more details.
19928 @item GCC_EXEC_PREFIX
19929 @findex GCC_EXEC_PREFIX
19930 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
19931 names of the subprograms executed by the compiler. No slash is added
19932 when this prefix is combined with the name of a subprogram, but you can
19933 specify a prefix that ends with a slash if you wish.
19935 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
19936 an appropriate prefix to use based on the pathname it is invoked with.
19938 If GCC cannot find the subprogram using the specified prefix, it
19939 tries looking in the usual places for the subprogram.
19941 The default value of @env{GCC_EXEC_PREFIX} is
19942 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
19943 the installed compiler. In many cases @var{prefix} is the value
19944 of @code{prefix} when you ran the @file{configure} script.
19946 Other prefixes specified with @option{-B} take precedence over this prefix.
19948 This prefix is also used for finding files such as @file{crt0.o} that are
19951 In addition, the prefix is used in an unusual way in finding the
19952 directories to search for header files. For each of the standard
19953 directories whose name normally begins with @samp{/usr/local/lib/gcc}
19954 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
19955 replacing that beginning with the specified prefix to produce an
19956 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
19957 @file{foo/bar} just before it searches the standard directory
19958 @file{/usr/local/lib/bar}.
19959 If a standard directory begins with the configured
19960 @var{prefix} then the value of @var{prefix} is replaced by
19961 @env{GCC_EXEC_PREFIX} when looking for header files.
19963 @item COMPILER_PATH
19964 @findex COMPILER_PATH
19965 The value of @env{COMPILER_PATH} is a colon-separated list of
19966 directories, much like @env{PATH}. GCC tries the directories thus
19967 specified when searching for subprograms, if it can't find the
19968 subprograms using @env{GCC_EXEC_PREFIX}.
19971 @findex LIBRARY_PATH
19972 The value of @env{LIBRARY_PATH} is a colon-separated list of
19973 directories, much like @env{PATH}. When configured as a native compiler,
19974 GCC tries the directories thus specified when searching for special
19975 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
19976 using GCC also uses these directories when searching for ordinary
19977 libraries for the @option{-l} option (but directories specified with
19978 @option{-L} come first).
19982 @cindex locale definition
19983 This variable is used to pass locale information to the compiler. One way in
19984 which this information is used is to determine the character set to be used
19985 when character literals, string literals and comments are parsed in C and C++.
19986 When the compiler is configured to allow multibyte characters,
19987 the following values for @env{LANG} are recognized:
19991 Recognize JIS characters.
19993 Recognize SJIS characters.
19995 Recognize EUCJP characters.
19998 If @env{LANG} is not defined, or if it has some other value, then the
19999 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
20000 recognize and translate multibyte characters.
20004 Some additional environment variables affect the behavior of the
20007 @include cppenv.texi
20011 @node Precompiled Headers
20012 @section Using Precompiled Headers
20013 @cindex precompiled headers
20014 @cindex speed of compilation
20016 Often large projects have many header files that are included in every
20017 source file. The time the compiler takes to process these header files
20018 over and over again can account for nearly all of the time required to
20019 build the project. To make builds faster, GCC allows you to
20020 @dfn{precompile} a header file.
20022 To create a precompiled header file, simply compile it as you would any
20023 other file, if necessary using the @option{-x} option to make the driver
20024 treat it as a C or C++ header file. You may want to use a
20025 tool like @command{make} to keep the precompiled header up-to-date when
20026 the headers it contains change.
20028 A precompiled header file is searched for when @code{#include} is
20029 seen in the compilation. As it searches for the included file
20030 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
20031 compiler looks for a precompiled header in each directory just before it
20032 looks for the include file in that directory. The name searched for is
20033 the name specified in the @code{#include} with @samp{.gch} appended. If
20034 the precompiled header file can't be used, it is ignored.
20036 For instance, if you have @code{#include "all.h"}, and you have
20037 @file{all.h.gch} in the same directory as @file{all.h}, then the
20038 precompiled header file is used if possible, and the original
20039 header is used otherwise.
20041 Alternatively, you might decide to put the precompiled header file in a
20042 directory and use @option{-I} to ensure that directory is searched
20043 before (or instead of) the directory containing the original header.
20044 Then, if you want to check that the precompiled header file is always
20045 used, you can put a file of the same name as the original header in this
20046 directory containing an @code{#error} command.
20048 This also works with @option{-include}. So yet another way to use
20049 precompiled headers, good for projects not designed with precompiled
20050 header files in mind, is to simply take most of the header files used by
20051 a project, include them from another header file, precompile that header
20052 file, and @option{-include} the precompiled header. If the header files
20053 have guards against multiple inclusion, they are skipped because
20054 they've already been included (in the precompiled header).
20056 If you need to precompile the same header file for different
20057 languages, targets, or compiler options, you can instead make a
20058 @emph{directory} named like @file{all.h.gch}, and put each precompiled
20059 header in the directory, perhaps using @option{-o}. It doesn't matter
20060 what you call the files in the directory; every precompiled header in
20061 the directory is considered. The first precompiled header
20062 encountered in the directory that is valid for this compilation is
20063 used; they're searched in no particular order.
20065 There are many other possibilities, limited only by your imagination,
20066 good sense, and the constraints of your build system.
20068 A precompiled header file can be used only when these conditions apply:
20072 Only one precompiled header can be used in a particular compilation.
20075 A precompiled header can't be used once the first C token is seen. You
20076 can have preprocessor directives before a precompiled header; you cannot
20077 include a precompiled header from inside another header.
20080 The precompiled header file must be produced for the same language as
20081 the current compilation. You can't use a C precompiled header for a C++
20085 The precompiled header file must have been produced by the same compiler
20086 binary as the current compilation is using.
20089 Any macros defined before the precompiled header is included must
20090 either be defined in the same way as when the precompiled header was
20091 generated, or must not affect the precompiled header, which usually
20092 means that they don't appear in the precompiled header at all.
20094 The @option{-D} option is one way to define a macro before a
20095 precompiled header is included; using a @code{#define} can also do it.
20096 There are also some options that define macros implicitly, like
20097 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
20100 @item If debugging information is output when using the precompiled
20101 header, using @option{-g} or similar, the same kind of debugging information
20102 must have been output when building the precompiled header. However,
20103 a precompiled header built using @option{-g} can be used in a compilation
20104 when no debugging information is being output.
20106 @item The same @option{-m} options must generally be used when building
20107 and using the precompiled header. @xref{Submodel Options},
20108 for any cases where this rule is relaxed.
20110 @item Each of the following options must be the same when building and using
20111 the precompiled header:
20113 @gccoptlist{-fexceptions}
20116 Some other command-line options starting with @option{-f},
20117 @option{-p}, or @option{-O} must be defined in the same way as when
20118 the precompiled header was generated. At present, it's not clear
20119 which options are safe to change and which are not; the safest choice
20120 is to use exactly the same options when generating and using the
20121 precompiled header. The following are known to be safe:
20123 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
20124 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
20125 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
20130 For all of these except the last, the compiler automatically
20131 ignores the precompiled header if the conditions aren't met. If you
20132 find an option combination that doesn't work and doesn't cause the
20133 precompiled header to be ignored, please consider filing a bug report,
20136 If you do use differing options when generating and using the
20137 precompiled header, the actual behavior is a mixture of the
20138 behavior for the options. For instance, if you use @option{-g} to
20139 generate the precompiled header but not when using it, you may or may
20140 not get debugging information for routines in the precompiled header.