1 @c Copyright (C) 1988-2014 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1988-2014 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
76 Other options are passed on to one stage of processing. Some options
77 control the preprocessor and others the compiler itself. Yet other
78 options control the assembler and linker; most of these are not
79 documented here, since you rarely need to use any of them.
81 @cindex C compilation options
82 Most of the command-line options that you can use with GCC are useful
83 for C programs; when an option is only useful with another language
84 (usually C++), the explanation says so explicitly. If the description
85 for a particular option does not mention a source language, you can use
86 that option with all supported languages.
88 @cindex C++ compilation options
89 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
90 options for compiling C++ programs.
92 @cindex grouping options
93 @cindex options, grouping
94 The @command{gcc} program accepts options and file names as operands. Many
95 options have multi-letter names; therefore multiple single-letter options
96 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
99 @cindex order of options
100 @cindex options, order
101 You can mix options and other arguments. For the most part, the order
102 you use doesn't matter. Order does matter when you use several
103 options of the same kind; for example, if you specify @option{-L} more
104 than once, the directories are searched in the order specified. Also,
105 the placement of the @option{-l} option is significant.
107 Many options have long names starting with @samp{-f} or with
108 @samp{-W}---for example,
109 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
110 these have both positive and negative forms; the negative form of
111 @option{-ffoo} is @option{-fno-foo}. This manual documents
112 only one of these two forms, whichever one is not the default.
116 @xref{Option Index}, for an index to GCC's options.
119 * Option Summary:: Brief list of all options, without explanations.
120 * Overall Options:: Controlling the kind of output:
121 an executable, object files, assembler files,
122 or preprocessed source.
123 * Invoking G++:: Compiling C++ programs.
124 * C Dialect Options:: Controlling the variant of C language compiled.
125 * C++ Dialect Options:: Variations on C++.
126 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
128 * Language Independent Options:: Controlling how diagnostics should be
130 * Warning Options:: How picky should the compiler be?
131 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
132 * Optimize Options:: How much optimization?
133 * Preprocessor Options:: Controlling header files and macro definitions.
134 Also, getting dependency information for Make.
135 * Assembler Options:: Passing options to the assembler.
136 * Link Options:: Specifying libraries and so on.
137 * Directory Options:: Where to find header files and libraries.
138 Where to find the compiler executable files.
139 * Spec Files:: How to pass switches to sub-processes.
140 * Target Options:: Running a cross-compiler, or an old version of GCC.
141 * Submodel Options:: Specifying minor hardware or convention variations,
142 such as 68010 vs 68020.
143 * Code Gen Options:: Specifying conventions for function calls, data layout
145 * Environment Variables:: Env vars that affect GCC.
146 * Precompiled Headers:: Compiling a header once, and using it many times.
152 @section Option Summary
154 Here is a summary of all the options, grouped by type. Explanations are
155 in the following sections.
158 @item Overall Options
159 @xref{Overall Options,,Options Controlling the Kind of Output}.
160 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
161 -pipe -pass-exit-codes @gol
162 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
163 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
164 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
166 @item C Language Options
167 @xref{C Dialect Options,,Options Controlling C Dialect}.
168 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
169 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
170 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
171 -fhosted -ffreestanding -fopenmp -fopenmp-simd -fms-extensions @gol
172 -fplan9-extensions -trigraphs -traditional -traditional-cpp @gol
173 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
174 -fsigned-bitfields -fsigned-char @gol
175 -funsigned-bitfields -funsigned-char}
177 @item C++ Language Options
178 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
179 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
180 -fconstexpr-depth=@var{n} -ffriend-injection @gol
181 -fno-elide-constructors @gol
182 -fno-enforce-eh-specs @gol
183 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
184 -fno-implicit-templates @gol
185 -fno-implicit-inline-templates @gol
186 -fno-implement-inlines -fms-extensions @gol
187 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
188 -fno-optional-diags -fpermissive @gol
189 -fno-pretty-templates @gol
190 -frepo -fno-rtti -fstats -ftemplate-backtrace-limit=@var{n} @gol
191 -ftemplate-depth=@var{n} @gol
192 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
193 -fvisibility-inlines-hidden @gol
194 -fvtable-verify=@var{std|preinit|none} @gol
195 -fvtv-counts -fvtv-debug @gol
196 -fvisibility-ms-compat @gol
197 -fext-numeric-literals @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wnarrowing @gol
200 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
201 -Weffc++ -Wstrict-null-sentinel @gol
202 -Wno-non-template-friend -Wold-style-cast @gol
203 -Woverloaded-virtual -Wno-pmf-conversions @gol
206 @item Objective-C and Objective-C++ Language Options
207 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
208 Objective-C and Objective-C++ Dialects}.
209 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
210 -fgnu-runtime -fnext-runtime @gol
211 -fno-nil-receivers @gol
212 -fobjc-abi-version=@var{n} @gol
213 -fobjc-call-cxx-cdtors @gol
214 -fobjc-direct-dispatch @gol
215 -fobjc-exceptions @gol
218 -fobjc-std=objc1 @gol
219 -freplace-objc-classes @gol
222 -Wassign-intercept @gol
223 -Wno-protocol -Wselector @gol
224 -Wstrict-selector-match @gol
225 -Wundeclared-selector}
227 @item Language Independent Options
228 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
229 @gccoptlist{-fmessage-length=@var{n} @gol
230 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
231 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
232 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
234 @item Warning Options
235 @xref{Warning Options,,Options to Request or Suppress Warnings}.
236 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
237 -pedantic-errors @gol
238 -w -Wextra -Wall -Waddress -Waggregate-return @gol
239 -Waggressive-loop-optimizations -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 -Wconditionally-supported @gol
243 -Wconversion -Wcoverage-mismatch -Wdate-time -Wdelete-incomplete -Wno-cpp @gol
244 -Wno-deprecated -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
259 -Wno-multichar -Wnonnull -Wno-overflow -Wopenmp-simd @gol
260 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
261 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
262 -Wpointer-arith -Wno-pointer-to-int-cast @gol
263 -Wredundant-decls -Wno-return-local-addr @gol
264 -Wreturn-type -Wsequence-point -Wshadow @gol
265 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
266 -Wsizeof-pointer-memaccess @gol
267 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
268 -Wstrict-aliasing=n @gol -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 -fsanitize=@var{style} @gol
292 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
293 -fdisable-ipa-@var{pass_name} @gol
294 -fdisable-rtl-@var{pass_name} @gol
295 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
296 -fdisable-tree-@var{pass_name} @gol
297 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
298 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
299 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
300 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
301 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
303 -fdump-statistics @gol
305 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-cfg -fdump-tree-alias @gol
309 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-gimple@r{[}-raw@r{]} @gol
312 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
318 -fdump-tree-nrv -fdump-tree-vect @gol
319 -fdump-tree-sink @gol
320 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
321 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
322 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
323 -fdump-tree-vtable-verify @gol
324 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
325 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
326 -fdump-final-insns=@var{file} @gol
327 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
328 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
329 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
330 -fenable-@var{kind}-@var{pass} @gol
331 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
332 -fdebug-types-section -fmem-report-wpa @gol
333 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
335 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
336 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
337 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
338 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
339 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
340 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
341 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
342 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
343 -gvms -gxcoff -gxcoff+ @gol
344 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
345 -fdebug-prefix-map=@var{old}=@var{new} @gol
346 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
347 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
348 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
349 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
350 -print-prog-name=@var{program} -print-search-dirs -Q @gol
351 -print-sysroot -print-sysroot-headers-suffix @gol
352 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
354 @item Optimization Options
355 @xref{Optimize Options,,Options that Control Optimization}.
356 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
357 -falign-jumps[=@var{n}] @gol
358 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
359 -fassociative-math -fauto-inc-dec -fbranch-probabilities @gol
360 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
361 -fbtr-bb-exclusive -fcaller-saves @gol
362 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
363 -fcompare-elim -fcprop-registers -fcrossjumping @gol
364 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
365 -fcx-limited-range @gol
366 -fdata-sections -fdce -fdelayed-branch @gol
367 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively -fdse @gol
368 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
369 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
370 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
371 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
372 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
373 -fif-conversion2 -findirect-inlining @gol
374 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
375 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
376 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
377 -fira-algorithm=@var{algorithm} @gol
378 -fira-region=@var{region} -fira-hoist-pressure @gol
379 -fira-loop-pressure -fno-ira-share-save-slots @gol
380 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
381 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute
382 -fivopts -fkeep-inline-functions -fkeep-static-consts -flive-range-shrinkage @gol
383 -floop-block -floop-interchange -floop-strip-mine -floop-nest-optimize @gol
384 -floop-parallelize-all -flto -flto-compression-level @gol
385 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
386 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
387 -fmove-loop-invariants -fno-branch-count-reg @gol
388 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
389 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
390 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
391 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
392 -fomit-frame-pointer -foptimize-sibling-calls @gol
393 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
394 -fprefetch-loop-arrays -fprofile-report @gol
395 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
396 -fprofile-generate=@var{path} @gol
397 -fprofile-use -fprofile-use=@var{path} -fprofile-values -fprofile-reorder-functions @gol
398 -freciprocal-math -free -frename-registers -freorder-blocks @gol
399 -freorder-blocks-and-partition -freorder-functions @gol
400 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
401 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
402 -fsched-spec-load -fsched-spec-load-dangerous @gol
403 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
404 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
405 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
406 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
407 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
408 -fselective-scheduling -fselective-scheduling2 @gol
409 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
410 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
411 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
412 -fstack-protector-all -fstack-protector-strong -fstrict-aliasing @gol
413 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
414 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
415 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
416 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
417 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
418 -ftree-loop-if-convert-stores -ftree-loop-im @gol
419 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
420 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
421 -ftree-loop-vectorize @gol
422 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
423 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
424 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
425 -ftree-vectorize -ftree-vrp @gol
426 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
427 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
428 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
429 -fwhole-program -fwpa -fuse-ld=@var{linker} -fuse-linker-plugin @gol
430 --param @var{name}=@var{value}
431 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
433 @item Preprocessor Options
434 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
435 @gccoptlist{-A@var{question}=@var{answer} @gol
436 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
437 -C -dD -dI -dM -dN @gol
438 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
439 -idirafter @var{dir} @gol
440 -include @var{file} -imacros @var{file} @gol
441 -iprefix @var{file} -iwithprefix @var{dir} @gol
442 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
443 -imultilib @var{dir} -isysroot @var{dir} @gol
444 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
445 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
446 -remap -trigraphs -undef -U@var{macro} @gol
447 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
449 @item Assembler Option
450 @xref{Assembler Options,,Passing Options to the Assembler}.
451 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
454 @xref{Link Options,,Options for Linking}.
455 @gccoptlist{@var{object-file-name} -l@var{library} @gol
456 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
457 -s -static -static-libgcc -static-libstdc++ @gol
458 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
459 -shared -shared-libgcc -symbolic @gol
460 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
463 @item Directory Options
464 @xref{Directory Options,,Options for Directory Search}.
465 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
466 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
467 --sysroot=@var{dir} --no-sysroot-suffix}
469 @item Machine Dependent Options
470 @xref{Submodel Options,,Hardware Models and Configurations}.
471 @c This list is ordered alphanumerically by subsection name.
472 @c Try and put the significant identifier (CPU or system) first,
473 @c so users have a clue at guessing where the ones they want will be.
475 @emph{AArch64 Options}
476 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
477 -mgeneral-regs-only @gol
478 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
480 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
481 -mtls-dialect=desc -mtls-dialect=traditional @gol
482 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
484 @emph{Adapteva Epiphany Options}
485 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
486 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
487 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
488 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
489 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
490 -msplit-vecmove-early -m1reg-@var{reg}}
493 @gccoptlist{-mbarrel-shifter @gol
494 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
495 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
496 -mea -mno-mpy -mmul32x16 -mmul64 @gol
497 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
498 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
499 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
500 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
501 -mucb-mcount -mvolatile-cache @gol
502 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
503 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
504 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
505 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
506 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
507 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
510 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
511 -mabi=@var{name} @gol
512 -mapcs-stack-check -mno-apcs-stack-check @gol
513 -mapcs-float -mno-apcs-float @gol
514 -mapcs-reentrant -mno-apcs-reentrant @gol
515 -msched-prolog -mno-sched-prolog @gol
516 -mlittle-endian -mbig-endian -mwords-little-endian @gol
517 -mfloat-abi=@var{name} @gol
518 -mfp16-format=@var{name}
519 -mthumb-interwork -mno-thumb-interwork @gol
520 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
521 -mstructure-size-boundary=@var{n} @gol
522 -mabort-on-noreturn @gol
523 -mlong-calls -mno-long-calls @gol
524 -msingle-pic-base -mno-single-pic-base @gol
525 -mpic-register=@var{reg} @gol
526 -mnop-fun-dllimport @gol
527 -mpoke-function-name @gol
529 -mtpcs-frame -mtpcs-leaf-frame @gol
530 -mcaller-super-interworking -mcallee-super-interworking @gol
531 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
532 -mword-relocations @gol
533 -mfix-cortex-m3-ldrd @gol
534 -munaligned-access @gol
535 -mneon-for-64bits @gol
536 -mslow-flash-data @gol
540 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
541 -mcall-prologues -mint8 -mno-interrupts -mrelax @gol
542 -mstrict-X -mtiny-stack -Waddr-space-convert}
544 @emph{Blackfin Options}
545 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
546 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
547 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
548 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
549 -mno-id-shared-library -mshared-library-id=@var{n} @gol
550 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
551 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
552 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
556 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
557 -msim -msdata=@var{sdata-type}}
560 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
561 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
562 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
563 -mstack-align -mdata-align -mconst-align @gol
564 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
565 -melf -maout -melinux -mlinux -sim -sim2 @gol
566 -mmul-bug-workaround -mno-mul-bug-workaround}
569 @gccoptlist{-mmac @gol
570 -mcr16cplus -mcr16c @gol
571 -msim -mint32 -mbit-ops
572 -mdata-model=@var{model}}
574 @emph{Darwin Options}
575 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
576 -arch_only -bind_at_load -bundle -bundle_loader @gol
577 -client_name -compatibility_version -current_version @gol
579 -dependency-file -dylib_file -dylinker_install_name @gol
580 -dynamic -dynamiclib -exported_symbols_list @gol
581 -filelist -flat_namespace -force_cpusubtype_ALL @gol
582 -force_flat_namespace -headerpad_max_install_names @gol
584 -image_base -init -install_name -keep_private_externs @gol
585 -multi_module -multiply_defined -multiply_defined_unused @gol
586 -noall_load -no_dead_strip_inits_and_terms @gol
587 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
588 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
589 -private_bundle -read_only_relocs -sectalign @gol
590 -sectobjectsymbols -whyload -seg1addr @gol
591 -sectcreate -sectobjectsymbols -sectorder @gol
592 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
593 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
594 -segprot -segs_read_only_addr -segs_read_write_addr @gol
595 -single_module -static -sub_library -sub_umbrella @gol
596 -twolevel_namespace -umbrella -undefined @gol
597 -unexported_symbols_list -weak_reference_mismatches @gol
598 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
599 -mkernel -mone-byte-bool}
601 @emph{DEC Alpha Options}
602 @gccoptlist{-mno-fp-regs -msoft-float @gol
603 -mieee -mieee-with-inexact -mieee-conformant @gol
604 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
605 -mtrap-precision=@var{mode} -mbuild-constants @gol
606 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
607 -mbwx -mmax -mfix -mcix @gol
608 -mfloat-vax -mfloat-ieee @gol
609 -mexplicit-relocs -msmall-data -mlarge-data @gol
610 -msmall-text -mlarge-text @gol
611 -mmemory-latency=@var{time}}
614 @gccoptlist{-msmall-model -mno-lsim}
617 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
618 -mhard-float -msoft-float @gol
619 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
620 -mdouble -mno-double @gol
621 -mmedia -mno-media -mmuladd -mno-muladd @gol
622 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
623 -mlinked-fp -mlong-calls -malign-labels @gol
624 -mlibrary-pic -macc-4 -macc-8 @gol
625 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
626 -moptimize-membar -mno-optimize-membar @gol
627 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
628 -mvliw-branch -mno-vliw-branch @gol
629 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
630 -mno-nested-cond-exec -mtomcat-stats @gol
634 @emph{GNU/Linux Options}
635 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
636 -tno-android-cc -tno-android-ld}
638 @emph{H8/300 Options}
639 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
642 @gccoptlist{-march=@var{architecture-type} @gol
643 -mdisable-fpregs -mdisable-indexing @gol
644 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
645 -mfixed-range=@var{register-range} @gol
646 -mjump-in-delay -mlinker-opt -mlong-calls @gol
647 -mlong-load-store -mno-disable-fpregs @gol
648 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
649 -mno-jump-in-delay -mno-long-load-store @gol
650 -mno-portable-runtime -mno-soft-float @gol
651 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
652 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
653 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
654 -munix=@var{unix-std} -nolibdld -static -threads}
656 @emph{i386 and x86-64 Options}
657 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
658 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
659 -mfpmath=@var{unit} @gol
660 -masm=@var{dialect} -mno-fancy-math-387 @gol
661 -mno-fp-ret-in-387 -msoft-float @gol
662 -mno-wide-multiply -mrtd -malign-double @gol
663 -mpreferred-stack-boundary=@var{num} @gol
664 -mincoming-stack-boundary=@var{num} @gol
665 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
666 -mrecip -mrecip=@var{opt} @gol
667 -mvzeroupper -mprefer-avx128 @gol
668 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
669 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
670 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
671 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
672 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mthreads @gol
673 -mno-align-stringops -minline-all-stringops @gol
674 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
675 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy}
676 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
677 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
678 -mregparm=@var{num} -msseregparm @gol
679 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
680 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
681 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
682 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
683 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
684 -msse2avx -mfentry -m8bit-idiv @gol
685 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
686 -mstack-protector-guard=@var{guard}}
688 @emph{i386 and x86-64 Windows Options}
689 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
690 -mnop-fun-dllimport -mthread @gol
691 -municode -mwin32 -mwindows -fno-set-stack-executable}
694 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
695 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
696 -mconstant-gp -mauto-pic -mfused-madd @gol
697 -minline-float-divide-min-latency @gol
698 -minline-float-divide-max-throughput @gol
699 -mno-inline-float-divide @gol
700 -minline-int-divide-min-latency @gol
701 -minline-int-divide-max-throughput @gol
702 -mno-inline-int-divide @gol
703 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
704 -mno-inline-sqrt @gol
705 -mdwarf2-asm -mearly-stop-bits @gol
706 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
707 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
708 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
709 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
710 -msched-spec-ldc -msched-spec-control-ldc @gol
711 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
712 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
713 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
714 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
717 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
718 -msign-extend-enabled -muser-enabled}
720 @emph{M32R/D Options}
721 @gccoptlist{-m32r2 -m32rx -m32r @gol
723 -malign-loops -mno-align-loops @gol
724 -missue-rate=@var{number} @gol
725 -mbranch-cost=@var{number} @gol
726 -mmodel=@var{code-size-model-type} @gol
727 -msdata=@var{sdata-type} @gol
728 -mno-flush-func -mflush-func=@var{name} @gol
729 -mno-flush-trap -mflush-trap=@var{number} @gol
733 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
735 @emph{M680x0 Options}
736 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
737 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
738 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
739 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
740 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
741 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
742 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
743 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
747 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
748 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
749 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
750 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
751 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
754 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
755 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
756 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
757 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
760 @emph{MicroBlaze Options}
761 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
762 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
763 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
764 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
765 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
768 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
769 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
770 -mips64 -mips64r2 @gol
771 -mips16 -mno-mips16 -mflip-mips16 @gol
772 -minterlink-compressed -mno-interlink-compressed @gol
773 -minterlink-mips16 -mno-interlink-mips16 @gol
774 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
775 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
776 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
777 -mno-float -msingle-float -mdouble-float @gol
778 -mabs=@var{mode} -mnan=@var{encoding} @gol
779 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
782 -mvirt -mno-virt @gol
783 -mmicromips -mno-micromips @gol
784 -mfpu=@var{fpu-type} @gol
785 -msmartmips -mno-smartmips @gol
786 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
787 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
788 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
789 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
790 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
791 -membedded-data -mno-embedded-data @gol
792 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
793 -mcode-readable=@var{setting} @gol
794 -msplit-addresses -mno-split-addresses @gol
795 -mexplicit-relocs -mno-explicit-relocs @gol
796 -mcheck-zero-division -mno-check-zero-division @gol
797 -mdivide-traps -mdivide-breaks @gol
798 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
799 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
800 -mfix-24k -mno-fix-24k @gol
801 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
802 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
803 -mfix-vr4120 -mno-fix-vr4120 @gol
804 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
805 -mflush-func=@var{func} -mno-flush-func @gol
806 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
807 -mfp-exceptions -mno-fp-exceptions @gol
808 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
809 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
812 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
813 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
814 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
815 -mno-base-addresses -msingle-exit -mno-single-exit}
817 @emph{MN10300 Options}
818 @gccoptlist{-mmult-bug -mno-mult-bug @gol
819 -mno-am33 -mam33 -mam33-2 -mam34 @gol
820 -mtune=@var{cpu-type} @gol
821 -mreturn-pointer-on-d0 @gol
822 -mno-crt0 -mrelax -mliw -msetlb}
825 @gccoptlist{-meb -mel -mno-crt0}
827 @emph{MSP430 Options}
828 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax}
831 @gccoptlist{-mbig-endian -mlittle-endian @gol
832 -mreduced-regs -mfull-regs @gol
833 -mcmov -mno-cmov @gol
834 -mperf-ext -mno-perf-ext @gol
835 -mv3push -mno-v3push @gol
836 -m16bit -mno-16bit @gol
837 -mgp-direct -mno-gp-direct @gol
838 -misr-vector-size=@var{num} @gol
839 -mcache-block-size=@var{num} @gol
840 -march=@var{arch} @gol
841 -mforce-fp-as-gp -mforbid-fp-as-gp @gol
842 -mex9 -mctor-dtor -mrelax}
844 @emph{Nios II Options}
845 @gccoptlist{-G @var{num} -mgpopt -mno-gpopt -mel -meb @gol
846 -mno-bypass-cache -mbypass-cache @gol
847 -mno-cache-volatile -mcache-volatile @gol
848 -mno-fast-sw-div -mfast-sw-div @gol
849 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
850 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
851 -mcustom-fpu-cfg=@var{name} @gol
852 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
854 @emph{PDP-11 Options}
855 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
856 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
857 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
858 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
859 -mbranch-expensive -mbranch-cheap @gol
860 -munix-asm -mdec-asm}
862 @emph{picoChip Options}
863 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
864 -msymbol-as-address -mno-inefficient-warnings}
866 @emph{PowerPC Options}
867 See RS/6000 and PowerPC Options.
870 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78}
872 @emph{RS/6000 and PowerPC Options}
873 @gccoptlist{-mcpu=@var{cpu-type} @gol
874 -mtune=@var{cpu-type} @gol
875 -mcmodel=@var{code-model} @gol
877 -maltivec -mno-altivec @gol
878 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
879 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
880 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
881 -mfprnd -mno-fprnd @gol
882 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
883 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
884 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
885 -malign-power -malign-natural @gol
886 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
887 -msingle-float -mdouble-float -msimple-fpu @gol
888 -mstring -mno-string -mupdate -mno-update @gol
889 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
890 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
891 -mstrict-align -mno-strict-align -mrelocatable @gol
892 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
893 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
894 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
895 -mprioritize-restricted-insns=@var{priority} @gol
896 -msched-costly-dep=@var{dependence_type} @gol
897 -minsert-sched-nops=@var{scheme} @gol
898 -mcall-sysv -mcall-netbsd @gol
899 -maix-struct-return -msvr4-struct-return @gol
900 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
901 -mblock-move-inline-limit=@var{num} @gol
902 -misel -mno-isel @gol
903 -misel=yes -misel=no @gol
905 -mspe=yes -mspe=no @gol
907 -mgen-cell-microcode -mwarn-cell-microcode @gol
908 -mvrsave -mno-vrsave @gol
909 -mmulhw -mno-mulhw @gol
910 -mdlmzb -mno-dlmzb @gol
911 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
912 -mprototype -mno-prototype @gol
913 -msim -mmvme -mads -myellowknife -memb -msdata @gol
914 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
915 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
916 -mno-recip-precision @gol
917 -mveclibabi=@var{type} -mfriz -mno-friz @gol
918 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
919 -msave-toc-indirect -mno-save-toc-indirect @gol
920 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
921 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
922 -mquad-memory -mno-quad-memory @gol
923 -mquad-memory-atomic -mno-quad-memory-atomic @gol
924 -mcompat-align-parm -mno-compat-align-parm}
927 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
929 -mbig-endian-data -mlittle-endian-data @gol
932 -mas100-syntax -mno-as100-syntax@gol
934 -mmax-constant-size=@gol
937 -mno-warn-multiple-fast-interrupts@gol
938 -msave-acc-in-interrupts}
940 @emph{S/390 and zSeries Options}
941 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
942 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
943 -mlong-double-64 -mlong-double-128 @gol
944 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
945 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
946 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
947 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
948 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
949 -mhotpatch[=@var{halfwords}] -mno-hotpatch}
952 @gccoptlist{-meb -mel @gol
956 -mscore5 -mscore5u -mscore7 -mscore7d}
959 @gccoptlist{-m1 -m2 -m2e @gol
960 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
962 -m4-nofpu -m4-single-only -m4-single -m4 @gol
963 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
964 -m5-64media -m5-64media-nofpu @gol
965 -m5-32media -m5-32media-nofpu @gol
966 -m5-compact -m5-compact-nofpu @gol
967 -mb -ml -mdalign -mrelax @gol
968 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
969 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
970 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
971 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
972 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
973 -maccumulate-outgoing-args -minvalid-symbols @gol
974 -matomic-model=@var{atomic-model} @gol
975 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
976 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
977 -mpretend-cmove -mtas}
979 @emph{Solaris 2 Options}
980 @gccoptlist{-mimpure-text -mno-impure-text @gol
984 @gccoptlist{-mcpu=@var{cpu-type} @gol
985 -mtune=@var{cpu-type} @gol
986 -mcmodel=@var{code-model} @gol
987 -mmemory-model=@var{mem-model} @gol
988 -m32 -m64 -mapp-regs -mno-app-regs @gol
989 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
990 -mfpu -mno-fpu -mhard-float -msoft-float @gol
991 -mhard-quad-float -msoft-quad-float @gol
992 -mstack-bias -mno-stack-bias @gol
993 -munaligned-doubles -mno-unaligned-doubles @gol
994 -mv8plus -mno-v8plus -mvis -mno-vis @gol
995 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
996 -mcbcond -mno-cbcond @gol
997 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
998 -mfix-at697f -mfix-ut699}
1001 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1002 -msafe-dma -munsafe-dma @gol
1004 -msmall-mem -mlarge-mem -mstdmain @gol
1005 -mfixed-range=@var{register-range} @gol
1007 -maddress-space-conversion -mno-address-space-conversion @gol
1008 -mcache-size=@var{cache-size} @gol
1009 -matomic-updates -mno-atomic-updates}
1011 @emph{System V Options}
1012 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1014 @emph{TILE-Gx Options}
1015 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1016 -mcmodel=@var{code-model}}
1018 @emph{TILEPro Options}
1019 @gccoptlist{-mcpu=@var{cpu} -m32}
1022 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1023 -mprolog-function -mno-prolog-function -mspace @gol
1024 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1025 -mapp-regs -mno-app-regs @gol
1026 -mdisable-callt -mno-disable-callt @gol
1027 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1028 -mv850e -mv850 -mv850e3v5 @gol
1039 @gccoptlist{-mg -mgnu -munix}
1042 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1043 -mpointer-size=@var{size}}
1045 @emph{VxWorks Options}
1046 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1047 -Xbind-lazy -Xbind-now}
1049 @emph{x86-64 Options}
1050 See i386 and x86-64 Options.
1052 @emph{Xstormy16 Options}
1055 @emph{Xtensa Options}
1056 @gccoptlist{-mconst16 -mno-const16 @gol
1057 -mfused-madd -mno-fused-madd @gol
1059 -mserialize-volatile -mno-serialize-volatile @gol
1060 -mtext-section-literals -mno-text-section-literals @gol
1061 -mtarget-align -mno-target-align @gol
1062 -mlongcalls -mno-longcalls}
1064 @emph{zSeries Options}
1065 See S/390 and zSeries Options.
1067 @item Code Generation Options
1068 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1069 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1070 -ffixed-@var{reg} -fexceptions @gol
1071 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1072 -fasynchronous-unwind-tables @gol
1073 -finhibit-size-directive -finstrument-functions @gol
1074 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1075 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1076 -fno-common -fno-ident @gol
1077 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1078 -fno-jump-tables @gol
1079 -frecord-gcc-switches @gol
1080 -freg-struct-return -fshort-enums @gol
1081 -fshort-double -fshort-wchar @gol
1082 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1083 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1084 -fno-stack-limit -fsplit-stack @gol
1085 -fleading-underscore -ftls-model=@var{model} @gol
1086 -fstack-reuse=@var{reuse_level} @gol
1087 -ftrapv -fwrapv -fbounds-check @gol
1088 -fvisibility -fstrict-volatile-bitfields -fsync-libcalls}
1092 * Overall Options:: Controlling the kind of output:
1093 an executable, object files, assembler files,
1094 or preprocessed source.
1095 * C Dialect Options:: Controlling the variant of C language compiled.
1096 * C++ Dialect Options:: Variations on C++.
1097 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
1099 * Language Independent Options:: Controlling how diagnostics should be
1101 * Warning Options:: How picky should the compiler be?
1102 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
1103 * Optimize Options:: How much optimization?
1104 * Preprocessor Options:: Controlling header files and macro definitions.
1105 Also, getting dependency information for Make.
1106 * Assembler Options:: Passing options to the assembler.
1107 * Link Options:: Specifying libraries and so on.
1108 * Directory Options:: Where to find header files and libraries.
1109 Where to find the compiler executable files.
1110 * Spec Files:: How to pass switches to sub-processes.
1111 * Target Options:: Running a cross-compiler, or an old version of GCC.
1114 @node Overall Options
1115 @section Options Controlling the Kind of Output
1117 Compilation can involve up to four stages: preprocessing, compilation
1118 proper, assembly and linking, always in that order. GCC is capable of
1119 preprocessing and compiling several files either into several
1120 assembler input files, or into one assembler input file; then each
1121 assembler input file produces an object file, and linking combines all
1122 the object files (those newly compiled, and those specified as input)
1123 into an executable file.
1125 @cindex file name suffix
1126 For any given input file, the file name suffix determines what kind of
1127 compilation is done:
1131 C source code that must be preprocessed.
1134 C source code that should not be preprocessed.
1137 C++ source code that should not be preprocessed.
1140 Objective-C source code. Note that you must link with the @file{libobjc}
1141 library to make an Objective-C program work.
1144 Objective-C source code that should not be preprocessed.
1148 Objective-C++ source code. Note that you must link with the @file{libobjc}
1149 library to make an Objective-C++ program work. Note that @samp{.M} refers
1150 to a literal capital M@.
1152 @item @var{file}.mii
1153 Objective-C++ source code that should not be preprocessed.
1156 C, C++, Objective-C or Objective-C++ header file to be turned into a
1157 precompiled header (default), or C, C++ header file to be turned into an
1158 Ada spec (via the @option{-fdump-ada-spec} switch).
1161 @itemx @var{file}.cp
1162 @itemx @var{file}.cxx
1163 @itemx @var{file}.cpp
1164 @itemx @var{file}.CPP
1165 @itemx @var{file}.c++
1167 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1168 the last two letters must both be literally @samp{x}. Likewise,
1169 @samp{.C} refers to a literal capital C@.
1173 Objective-C++ source code that must be preprocessed.
1175 @item @var{file}.mii
1176 Objective-C++ source code that should not be preprocessed.
1180 @itemx @var{file}.hp
1181 @itemx @var{file}.hxx
1182 @itemx @var{file}.hpp
1183 @itemx @var{file}.HPP
1184 @itemx @var{file}.h++
1185 @itemx @var{file}.tcc
1186 C++ header file to be turned into a precompiled header or Ada spec.
1189 @itemx @var{file}.for
1190 @itemx @var{file}.ftn
1191 Fixed form Fortran source code that should not be preprocessed.
1194 @itemx @var{file}.FOR
1195 @itemx @var{file}.fpp
1196 @itemx @var{file}.FPP
1197 @itemx @var{file}.FTN
1198 Fixed form Fortran source code that must be preprocessed (with the traditional
1201 @item @var{file}.f90
1202 @itemx @var{file}.f95
1203 @itemx @var{file}.f03
1204 @itemx @var{file}.f08
1205 Free form Fortran source code that should not be preprocessed.
1207 @item @var{file}.F90
1208 @itemx @var{file}.F95
1209 @itemx @var{file}.F03
1210 @itemx @var{file}.F08
1211 Free form Fortran source code that must be preprocessed (with the
1212 traditional preprocessor).
1217 @c FIXME: Descriptions of Java file types.
1223 @item @var{file}.ads
1224 Ada source code file that contains a library unit declaration (a
1225 declaration of a package, subprogram, or generic, or a generic
1226 instantiation), or a library unit renaming declaration (a package,
1227 generic, or subprogram renaming declaration). Such files are also
1230 @item @var{file}.adb
1231 Ada source code file containing a library unit body (a subprogram or
1232 package body). Such files are also called @dfn{bodies}.
1234 @c GCC also knows about some suffixes for languages not yet included:
1245 @itemx @var{file}.sx
1246 Assembler code that must be preprocessed.
1249 An object file to be fed straight into linking.
1250 Any file name with no recognized suffix is treated this way.
1254 You can specify the input language explicitly with the @option{-x} option:
1257 @item -x @var{language}
1258 Specify explicitly the @var{language} for the following input files
1259 (rather than letting the compiler choose a default based on the file
1260 name suffix). This option applies to all following input files until
1261 the next @option{-x} option. Possible values for @var{language} are:
1263 c c-header cpp-output
1264 c++ c++-header c++-cpp-output
1265 objective-c objective-c-header objective-c-cpp-output
1266 objective-c++ objective-c++-header objective-c++-cpp-output
1267 assembler assembler-with-cpp
1269 f77 f77-cpp-input f95 f95-cpp-input
1275 Turn off any specification of a language, so that subsequent files are
1276 handled according to their file name suffixes (as they are if @option{-x}
1277 has not been used at all).
1279 @item -pass-exit-codes
1280 @opindex pass-exit-codes
1281 Normally the @command{gcc} program exits with the code of 1 if any
1282 phase of the compiler returns a non-success return code. If you specify
1283 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1284 the numerically highest error produced by any phase returning an error
1285 indication. The C, C++, and Fortran front ends return 4 if an internal
1286 compiler error is encountered.
1289 If you only want some of the stages of compilation, you can use
1290 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1291 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1292 @command{gcc} is to stop. Note that some combinations (for example,
1293 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1298 Compile or assemble the source files, but do not link. The linking
1299 stage simply is not done. The ultimate output is in the form of an
1300 object file for each source file.
1302 By default, the object file name for a source file is made by replacing
1303 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1305 Unrecognized input files, not requiring compilation or assembly, are
1310 Stop after the stage of compilation proper; do not assemble. The output
1311 is in the form of an assembler code file for each non-assembler input
1314 By default, the assembler file name for a source file is made by
1315 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1317 Input files that don't require compilation are ignored.
1321 Stop after the preprocessing stage; do not run the compiler proper. The
1322 output is in the form of preprocessed source code, which is sent to the
1325 Input files that don't require preprocessing are ignored.
1327 @cindex output file option
1330 Place output in file @var{file}. This applies to whatever
1331 sort of output is being produced, whether it be an executable file,
1332 an object file, an assembler file or preprocessed C code.
1334 If @option{-o} is not specified, the default is to put an executable
1335 file in @file{a.out}, the object file for
1336 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1337 assembler file in @file{@var{source}.s}, a precompiled header file in
1338 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1343 Print (on standard error output) the commands executed to run the stages
1344 of compilation. Also print the version number of the compiler driver
1345 program and of the preprocessor and the compiler proper.
1349 Like @option{-v} except the commands are not executed and arguments
1350 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1351 This is useful for shell scripts to capture the driver-generated command lines.
1355 Use pipes rather than temporary files for communication between the
1356 various stages of compilation. This fails to work on some systems where
1357 the assembler is unable to read from a pipe; but the GNU assembler has
1362 Print (on the standard output) a description of the command-line options
1363 understood by @command{gcc}. If the @option{-v} option is also specified
1364 then @option{--help} is also passed on to the various processes
1365 invoked by @command{gcc}, so that they can display the command-line options
1366 they accept. If the @option{-Wextra} option has also been specified
1367 (prior to the @option{--help} option), then command-line options that
1368 have no documentation associated with them are also displayed.
1371 @opindex target-help
1372 Print (on the standard output) a description of target-specific command-line
1373 options for each tool. For some targets extra target-specific
1374 information may also be printed.
1376 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1377 Print (on the standard output) a description of the command-line
1378 options understood by the compiler that fit into all specified classes
1379 and qualifiers. These are the supported classes:
1382 @item @samp{optimizers}
1383 Display all of the optimization options supported by the
1386 @item @samp{warnings}
1387 Display all of the options controlling warning messages
1388 produced by the compiler.
1391 Display target-specific options. Unlike the
1392 @option{--target-help} option however, target-specific options of the
1393 linker and assembler are not displayed. This is because those
1394 tools do not currently support the extended @option{--help=} syntax.
1397 Display the values recognized by the @option{--param}
1400 @item @var{language}
1401 Display the options supported for @var{language}, where
1402 @var{language} is the name of one of the languages supported in this
1406 Display the options that are common to all languages.
1409 These are the supported qualifiers:
1412 @item @samp{undocumented}
1413 Display only those options that are undocumented.
1416 Display options taking an argument that appears after an equal
1417 sign in the same continuous piece of text, such as:
1418 @samp{--help=target}.
1420 @item @samp{separate}
1421 Display options taking an argument that appears as a separate word
1422 following the original option, such as: @samp{-o output-file}.
1425 Thus for example to display all the undocumented target-specific
1426 switches supported by the compiler, use:
1429 --help=target,undocumented
1432 The sense of a qualifier can be inverted by prefixing it with the
1433 @samp{^} character, so for example to display all binary warning
1434 options (i.e., ones that are either on or off and that do not take an
1435 argument) that have a description, use:
1438 --help=warnings,^joined,^undocumented
1441 The argument to @option{--help=} should not consist solely of inverted
1444 Combining several classes is possible, although this usually
1445 restricts the output so much that there is nothing to display. One
1446 case where it does work, however, is when one of the classes is
1447 @var{target}. For example, to display all the target-specific
1448 optimization options, use:
1451 --help=target,optimizers
1454 The @option{--help=} option can be repeated on the command line. Each
1455 successive use displays its requested class of options, skipping
1456 those that have already been displayed.
1458 If the @option{-Q} option appears on the command line before the
1459 @option{--help=} option, then the descriptive text displayed by
1460 @option{--help=} is changed. Instead of describing the displayed
1461 options, an indication is given as to whether the option is enabled,
1462 disabled or set to a specific value (assuming that the compiler
1463 knows this at the point where the @option{--help=} option is used).
1465 Here is a truncated example from the ARM port of @command{gcc}:
1468 % gcc -Q -mabi=2 --help=target -c
1469 The following options are target specific:
1471 -mabort-on-noreturn [disabled]
1475 The output is sensitive to the effects of previous command-line
1476 options, so for example it is possible to find out which optimizations
1477 are enabled at @option{-O2} by using:
1480 -Q -O2 --help=optimizers
1483 Alternatively you can discover which binary optimizations are enabled
1484 by @option{-O3} by using:
1487 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1488 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1489 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1492 @item -no-canonical-prefixes
1493 @opindex no-canonical-prefixes
1494 Do not expand any symbolic links, resolve references to @samp{/../}
1495 or @samp{/./}, or make the path absolute when generating a relative
1500 Display the version number and copyrights of the invoked GCC@.
1504 Invoke all subcommands under a wrapper program. The name of the
1505 wrapper program and its parameters are passed as a comma separated
1509 gcc -c t.c -wrapper gdb,--args
1513 This invokes all subprograms of @command{gcc} under
1514 @samp{gdb --args}, thus the invocation of @command{cc1} is
1515 @samp{gdb --args cc1 @dots{}}.
1517 @item -fplugin=@var{name}.so
1519 Load the plugin code in file @var{name}.so, assumed to be a
1520 shared object to be dlopen'd by the compiler. The base name of
1521 the shared object file is used to identify the plugin for the
1522 purposes of argument parsing (See
1523 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1524 Each plugin should define the callback functions specified in the
1527 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1528 @opindex fplugin-arg
1529 Define an argument called @var{key} with a value of @var{value}
1530 for the plugin called @var{name}.
1532 @item -fdump-ada-spec@r{[}-slim@r{]}
1533 @opindex fdump-ada-spec
1534 For C and C++ source and include files, generate corresponding Ada specs.
1535 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1536 GNAT User's Guide}, which provides detailed documentation on this feature.
1538 @item -fada-spec-parent=@var{unit}
1539 @opindex fada-spec-parent
1540 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1541 Ada specs as child units of parent @var{unit}.
1543 @item -fdump-go-spec=@var{file}
1544 @opindex fdump-go-spec
1545 For input files in any language, generate corresponding Go
1546 declarations in @var{file}. This generates Go @code{const},
1547 @code{type}, @code{var}, and @code{func} declarations which may be a
1548 useful way to start writing a Go interface to code written in some
1551 @include @value{srcdir}/../libiberty/at-file.texi
1555 @section Compiling C++ Programs
1557 @cindex suffixes for C++ source
1558 @cindex C++ source file suffixes
1559 C++ source files conventionally use one of the suffixes @samp{.C},
1560 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1561 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1562 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1563 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1564 files with these names and compiles them as C++ programs even if you
1565 call the compiler the same way as for compiling C programs (usually
1566 with the name @command{gcc}).
1570 However, the use of @command{gcc} does not add the C++ library.
1571 @command{g++} is a program that calls GCC and automatically specifies linking
1572 against the C++ library. It treats @samp{.c},
1573 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1574 files unless @option{-x} is used. This program is also useful when
1575 precompiling a C header file with a @samp{.h} extension for use in C++
1576 compilations. On many systems, @command{g++} is also installed with
1577 the name @command{c++}.
1579 @cindex invoking @command{g++}
1580 When you compile C++ programs, you may specify many of the same
1581 command-line options that you use for compiling programs in any
1582 language; or command-line options meaningful for C and related
1583 languages; or options that are meaningful only for C++ programs.
1584 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1585 explanations of options for languages related to C@.
1586 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1587 explanations of options that are meaningful only for C++ programs.
1589 @node C Dialect Options
1590 @section Options Controlling C Dialect
1591 @cindex dialect options
1592 @cindex language dialect options
1593 @cindex options, dialect
1595 The following options control the dialect of C (or languages derived
1596 from C, such as C++, Objective-C and Objective-C++) that the compiler
1600 @cindex ANSI support
1604 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1605 equivalent to @option{-std=c++98}.
1607 This turns off certain features of GCC that are incompatible with ISO
1608 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1609 such as the @code{asm} and @code{typeof} keywords, and
1610 predefined macros such as @code{unix} and @code{vax} that identify the
1611 type of system you are using. It also enables the undesirable and
1612 rarely used ISO trigraph feature. For the C compiler,
1613 it disables recognition of C++ style @samp{//} comments as well as
1614 the @code{inline} keyword.
1616 The alternate keywords @code{__asm__}, @code{__extension__},
1617 @code{__inline__} and @code{__typeof__} continue to work despite
1618 @option{-ansi}. You would not want to use them in an ISO C program, of
1619 course, but it is useful to put them in header files that might be included
1620 in compilations done with @option{-ansi}. Alternate predefined macros
1621 such as @code{__unix__} and @code{__vax__} are also available, with or
1622 without @option{-ansi}.
1624 The @option{-ansi} option does not cause non-ISO programs to be
1625 rejected gratuitously. For that, @option{-Wpedantic} is required in
1626 addition to @option{-ansi}. @xref{Warning Options}.
1628 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1629 option is used. Some header files may notice this macro and refrain
1630 from declaring certain functions or defining certain macros that the
1631 ISO standard doesn't call for; this is to avoid interfering with any
1632 programs that might use these names for other things.
1634 Functions that are normally built in but do not have semantics
1635 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1636 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1637 built-in functions provided by GCC}, for details of the functions
1642 Determine the language standard. @xref{Standards,,Language Standards
1643 Supported by GCC}, for details of these standard versions. This option
1644 is currently only supported when compiling C or C++.
1646 The compiler can accept several base standards, such as @samp{c90} or
1647 @samp{c++98}, and GNU dialects of those standards, such as
1648 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1649 compiler accepts all programs following that standard plus those
1650 using GNU extensions that do not contradict it. For example,
1651 @option{-std=c90} turns off certain features of GCC that are
1652 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1653 keywords, but not other GNU extensions that do not have a meaning in
1654 ISO C90, such as omitting the middle term of a @code{?:}
1655 expression. On the other hand, when a GNU dialect of a standard is
1656 specified, all features supported by the compiler are enabled, even when
1657 those features change the meaning of the base standard. As a result, some
1658 strict-conforming programs may be rejected. The particular standard
1659 is used by @option{-Wpedantic} to identify which features are GNU
1660 extensions given that version of the standard. For example
1661 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1662 comments, while @option{-std=gnu99 -Wpedantic} does not.
1664 A value for this option must be provided; possible values are
1670 Support all ISO C90 programs (certain GNU extensions that conflict
1671 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1673 @item iso9899:199409
1674 ISO C90 as modified in amendment 1.
1680 ISO C99. This standard is substantially completely supported, modulo
1681 bugs, extended identifiers (supported except for corner cases when
1682 @option{-fextended-identifiers} is used) and floating-point issues
1683 (mainly but not entirely relating to optional C99 features from
1684 Annexes F and G). See
1685 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1686 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1691 ISO C11, the 2011 revision of the ISO C standard. This standard is
1692 substantially completely supported, modulo bugs, extended identifiers
1693 (supported except for corner cases when
1694 @option{-fextended-identifiers} is used), floating-point issues
1695 (mainly but not entirely relating to optional C11 features from
1696 Annexes F and G) and the optional Annexes K (Bounds-checking
1697 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1701 GNU dialect of ISO C90 (including some C99 features). This
1702 is the default for C code.
1706 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1710 GNU dialect of ISO C11. This is intended to become the default in a
1711 future release of GCC. The name @samp{gnu1x} is deprecated.
1715 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1716 additional defect reports. Same as @option{-ansi} for C++ code.
1720 GNU dialect of @option{-std=c++98}. This is the default for
1725 The 2011 ISO C++ standard plus amendments.
1726 The name @samp{c++0x} is deprecated.
1730 GNU dialect of @option{-std=c++11}.
1731 The name @samp{gnu++0x} is deprecated.
1734 The next revision of the ISO C++ standard, tentatively planned for
1735 2014. Support is highly experimental, and will almost certainly
1736 change in incompatible ways in future releases.
1739 GNU dialect of @option{-std=c++1y}. Support is highly experimental,
1740 and will almost certainly change in incompatible ways in future
1744 @item -fgnu89-inline
1745 @opindex fgnu89-inline
1746 The option @option{-fgnu89-inline} tells GCC to use the traditional
1747 GNU semantics for @code{inline} functions when in C99 mode.
1748 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1749 is accepted and ignored by GCC versions 4.1.3 up to but not including
1750 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1751 C99 mode. Using this option is roughly equivalent to adding the
1752 @code{gnu_inline} function attribute to all inline functions
1753 (@pxref{Function Attributes}).
1755 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1756 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1757 specifies the default behavior). This option was first supported in
1758 GCC 4.3. This option is not supported in @option{-std=c90} or
1759 @option{-std=gnu90} mode.
1761 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1762 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1763 in effect for @code{inline} functions. @xref{Common Predefined
1764 Macros,,,cpp,The C Preprocessor}.
1766 @item -aux-info @var{filename}
1768 Output to the given filename prototyped declarations for all functions
1769 declared and/or defined in a translation unit, including those in header
1770 files. This option is silently ignored in any language other than C@.
1772 Besides declarations, the file indicates, in comments, the origin of
1773 each declaration (source file and line), whether the declaration was
1774 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1775 @samp{O} for old, respectively, in the first character after the line
1776 number and the colon), and whether it came from a declaration or a
1777 definition (@samp{C} or @samp{F}, respectively, in the following
1778 character). In the case of function definitions, a K&R-style list of
1779 arguments followed by their declarations is also provided, inside
1780 comments, after the declaration.
1782 @item -fallow-parameterless-variadic-functions
1783 @opindex fallow-parameterless-variadic-functions
1784 Accept variadic functions without named parameters.
1786 Although it is possible to define such a function, this is not very
1787 useful as it is not possible to read the arguments. This is only
1788 supported for C as this construct is allowed by C++.
1792 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1793 keyword, so that code can use these words as identifiers. You can use
1794 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1795 instead. @option{-ansi} implies @option{-fno-asm}.
1797 In C++, this switch only affects the @code{typeof} keyword, since
1798 @code{asm} and @code{inline} are standard keywords. You may want to
1799 use the @option{-fno-gnu-keywords} flag instead, which has the same
1800 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1801 switch only affects the @code{asm} and @code{typeof} keywords, since
1802 @code{inline} is a standard keyword in ISO C99.
1805 @itemx -fno-builtin-@var{function}
1806 @opindex fno-builtin
1807 @cindex built-in functions
1808 Don't recognize built-in functions that do not begin with
1809 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1810 functions provided by GCC}, for details of the functions affected,
1811 including those which are not built-in functions when @option{-ansi} or
1812 @option{-std} options for strict ISO C conformance are used because they
1813 do not have an ISO standard meaning.
1815 GCC normally generates special code to handle certain built-in functions
1816 more efficiently; for instance, calls to @code{alloca} may become single
1817 instructions which adjust the stack directly, and calls to @code{memcpy}
1818 may become inline copy loops. The resulting code is often both smaller
1819 and faster, but since the function calls no longer appear as such, you
1820 cannot set a breakpoint on those calls, nor can you change the behavior
1821 of the functions by linking with a different library. In addition,
1822 when a function is recognized as a built-in function, GCC may use
1823 information about that function to warn about problems with calls to
1824 that function, or to generate more efficient code, even if the
1825 resulting code still contains calls to that function. For example,
1826 warnings are given with @option{-Wformat} for bad calls to
1827 @code{printf} when @code{printf} is built in and @code{strlen} is
1828 known not to modify global memory.
1830 With the @option{-fno-builtin-@var{function}} option
1831 only the built-in function @var{function} is
1832 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1833 function is named that is not built-in in this version of GCC, this
1834 option is ignored. There is no corresponding
1835 @option{-fbuiltin-@var{function}} option; if you wish to enable
1836 built-in functions selectively when using @option{-fno-builtin} or
1837 @option{-ffreestanding}, you may define macros such as:
1840 #define abs(n) __builtin_abs ((n))
1841 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1846 @cindex hosted environment
1848 Assert that compilation targets a hosted environment. This implies
1849 @option{-fbuiltin}. A hosted environment is one in which the
1850 entire standard library is available, and in which @code{main} has a return
1851 type of @code{int}. Examples are nearly everything except a kernel.
1852 This is equivalent to @option{-fno-freestanding}.
1854 @item -ffreestanding
1855 @opindex ffreestanding
1856 @cindex hosted environment
1858 Assert that compilation targets a freestanding environment. This
1859 implies @option{-fno-builtin}. A freestanding environment
1860 is one in which the standard library may not exist, and program startup may
1861 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1862 This is equivalent to @option{-fno-hosted}.
1864 @xref{Standards,,Language Standards Supported by GCC}, for details of
1865 freestanding and hosted environments.
1869 @cindex OpenMP parallel
1870 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1871 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1872 compiler generates parallel code according to the OpenMP Application
1873 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1874 implies @option{-pthread}, and thus is only supported on targets that
1875 have support for @option{-pthread}. @option{-fopenmp} implies
1876 @option{-fopenmp-simd}.
1879 @opindex fopenmp-simd
1882 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1883 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1888 @cindex Enable Cilk Plus
1889 Enable the usage of Cilk Plus language extension features for C/C++.
1890 When the option @option{-fcilkplus} is specified, enable the usage of
1891 the Cilk Plus Language extension features for C/C++. The present
1892 implementation follows ABI version 1.2. This is an experimental
1893 feature that is only partially complete, and whose interface may
1894 change in future versions of GCC as the official specification
1895 changes. Currently, all features but @code{_Cilk_for} have been
1900 When the option @option{-fgnu-tm} is specified, the compiler
1901 generates code for the Linux variant of Intel's current Transactional
1902 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1903 an experimental feature whose interface may change in future versions
1904 of GCC, as the official specification changes. Please note that not
1905 all architectures are supported for this feature.
1907 For more information on GCC's support for transactional memory,
1908 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1909 Transactional Memory Library}.
1911 Note that the transactional memory feature is not supported with
1912 non-call exceptions (@option{-fnon-call-exceptions}).
1914 @item -fms-extensions
1915 @opindex fms-extensions
1916 Accept some non-standard constructs used in Microsoft header files.
1918 In C++ code, this allows member names in structures to be similar
1919 to previous types declarations.
1928 Some cases of unnamed fields in structures and unions are only
1929 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1930 fields within structs/unions}, for details.
1932 Note that this option is off for all targets but i?86 and x86_64
1933 targets using ms-abi.
1934 @item -fplan9-extensions
1935 Accept some non-standard constructs used in Plan 9 code.
1937 This enables @option{-fms-extensions}, permits passing pointers to
1938 structures with anonymous fields to functions that expect pointers to
1939 elements of the type of the field, and permits referring to anonymous
1940 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1941 struct/union fields within structs/unions}, for details. This is only
1942 supported for C, not C++.
1946 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1947 options for strict ISO C conformance) implies @option{-trigraphs}.
1949 @cindex traditional C language
1950 @cindex C language, traditional
1952 @itemx -traditional-cpp
1953 @opindex traditional-cpp
1954 @opindex traditional
1955 Formerly, these options caused GCC to attempt to emulate a pre-standard
1956 C compiler. They are now only supported with the @option{-E} switch.
1957 The preprocessor continues to support a pre-standard mode. See the GNU
1958 CPP manual for details.
1960 @item -fcond-mismatch
1961 @opindex fcond-mismatch
1962 Allow conditional expressions with mismatched types in the second and
1963 third arguments. The value of such an expression is void. This option
1964 is not supported for C++.
1966 @item -flax-vector-conversions
1967 @opindex flax-vector-conversions
1968 Allow implicit conversions between vectors with differing numbers of
1969 elements and/or incompatible element types. This option should not be
1972 @item -funsigned-char
1973 @opindex funsigned-char
1974 Let the type @code{char} be unsigned, like @code{unsigned char}.
1976 Each kind of machine has a default for what @code{char} should
1977 be. It is either like @code{unsigned char} by default or like
1978 @code{signed char} by default.
1980 Ideally, a portable program should always use @code{signed char} or
1981 @code{unsigned char} when it depends on the signedness of an object.
1982 But many programs have been written to use plain @code{char} and
1983 expect it to be signed, or expect it to be unsigned, depending on the
1984 machines they were written for. This option, and its inverse, let you
1985 make such a program work with the opposite default.
1987 The type @code{char} is always a distinct type from each of
1988 @code{signed char} or @code{unsigned char}, even though its behavior
1989 is always just like one of those two.
1992 @opindex fsigned-char
1993 Let the type @code{char} be signed, like @code{signed char}.
1995 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1996 the negative form of @option{-funsigned-char}. Likewise, the option
1997 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1999 @item -fsigned-bitfields
2000 @itemx -funsigned-bitfields
2001 @itemx -fno-signed-bitfields
2002 @itemx -fno-unsigned-bitfields
2003 @opindex fsigned-bitfields
2004 @opindex funsigned-bitfields
2005 @opindex fno-signed-bitfields
2006 @opindex fno-unsigned-bitfields
2007 These options control whether a bit-field is signed or unsigned, when the
2008 declaration does not use either @code{signed} or @code{unsigned}. By
2009 default, such a bit-field is signed, because this is consistent: the
2010 basic integer types such as @code{int} are signed types.
2013 @node C++ Dialect Options
2014 @section Options Controlling C++ Dialect
2016 @cindex compiler options, C++
2017 @cindex C++ options, command-line
2018 @cindex options, C++
2019 This section describes the command-line options that are only meaningful
2020 for C++ programs. You can also use most of the GNU compiler options
2021 regardless of what language your program is in. For example, you
2022 might compile a file @code{firstClass.C} like this:
2025 g++ -g -frepo -O -c firstClass.C
2029 In this example, only @option{-frepo} is an option meant
2030 only for C++ programs; you can use the other options with any
2031 language supported by GCC@.
2033 Here is a list of options that are @emph{only} for compiling C++ programs:
2037 @item -fabi-version=@var{n}
2038 @opindex fabi-version
2039 Use version @var{n} of the C++ ABI@. The default is version 2.
2041 Version 0 refers to the version conforming most closely to
2042 the C++ ABI specification. Therefore, the ABI obtained using version 0
2043 will change in different versions of G++ as ABI bugs are fixed.
2045 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2047 Version 2 is the version of the C++ ABI that first appeared in G++ 3.4.
2049 Version 3 corrects an error in mangling a constant address as a
2052 Version 4, which first appeared in G++ 4.5, implements a standard
2053 mangling for vector types.
2055 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2056 attribute const/volatile on function pointer types, decltype of a
2057 plain decl, and use of a function parameter in the declaration of
2060 Version 6, which first appeared in G++ 4.7, corrects the promotion
2061 behavior of C++11 scoped enums and the mangling of template argument
2062 packs, const/static_cast, prefix ++ and --, and a class scope function
2063 used as a template argument.
2065 See also @option{-Wabi}.
2067 @item -fno-access-control
2068 @opindex fno-access-control
2069 Turn off all access checking. This switch is mainly useful for working
2070 around bugs in the access control code.
2074 Check that the pointer returned by @code{operator new} is non-null
2075 before attempting to modify the storage allocated. This check is
2076 normally unnecessary because the C++ standard specifies that
2077 @code{operator new} only returns @code{0} if it is declared
2078 @samp{throw()}, in which case the compiler always checks the
2079 return value even without this option. In all other cases, when
2080 @code{operator new} has a non-empty exception specification, memory
2081 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2082 @samp{new (nothrow)}.
2084 @item -fconstexpr-depth=@var{n}
2085 @opindex fconstexpr-depth
2086 Set the maximum nested evaluation depth for C++11 constexpr functions
2087 to @var{n}. A limit is needed to detect endless recursion during
2088 constant expression evaluation. The minimum specified by the standard
2091 @item -fdeduce-init-list
2092 @opindex fdeduce-init-list
2093 Enable deduction of a template type parameter as
2094 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2097 template <class T> auto forward(T t) -> decltype (realfn (t))
2104 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2108 This deduction was implemented as a possible extension to the
2109 originally proposed semantics for the C++11 standard, but was not part
2110 of the final standard, so it is disabled by default. This option is
2111 deprecated, and may be removed in a future version of G++.
2113 @item -ffriend-injection
2114 @opindex ffriend-injection
2115 Inject friend functions into the enclosing namespace, so that they are
2116 visible outside the scope of the class in which they are declared.
2117 Friend functions were documented to work this way in the old Annotated
2118 C++ Reference Manual, and versions of G++ before 4.1 always worked
2119 that way. However, in ISO C++ a friend function that is not declared
2120 in an enclosing scope can only be found using argument dependent
2121 lookup. This option causes friends to be injected as they were in
2124 This option is for compatibility, and may be removed in a future
2127 @item -fno-elide-constructors
2128 @opindex fno-elide-constructors
2129 The C++ standard allows an implementation to omit creating a temporary
2130 that is only used to initialize another object of the same type.
2131 Specifying this option disables that optimization, and forces G++ to
2132 call the copy constructor in all cases.
2134 @item -fno-enforce-eh-specs
2135 @opindex fno-enforce-eh-specs
2136 Don't generate code to check for violation of exception specifications
2137 at run time. This option violates the C++ standard, but may be useful
2138 for reducing code size in production builds, much like defining
2139 @samp{NDEBUG}. This does not give user code permission to throw
2140 exceptions in violation of the exception specifications; the compiler
2141 still optimizes based on the specifications, so throwing an
2142 unexpected exception results in undefined behavior at run time.
2144 @item -fextern-tls-init
2145 @itemx -fno-extern-tls-init
2146 @opindex fextern-tls-init
2147 @opindex fno-extern-tls-init
2148 The C++11 and OpenMP standards allow @samp{thread_local} and
2149 @samp{threadprivate} variables to have dynamic (runtime)
2150 initialization. To support this, any use of such a variable goes
2151 through a wrapper function that performs any necessary initialization.
2152 When the use and definition of the variable are in the same
2153 translation unit, this overhead can be optimized away, but when the
2154 use is in a different translation unit there is significant overhead
2155 even if the variable doesn't actually need dynamic initialization. If
2156 the programmer can be sure that no use of the variable in a
2157 non-defining TU needs to trigger dynamic initialization (either
2158 because the variable is statically initialized, or a use of the
2159 variable in the defining TU will be executed before any uses in
2160 another TU), they can avoid this overhead with the
2161 @option{-fno-extern-tls-init} option.
2163 On targets that support symbol aliases, the default is
2164 @option{-fextern-tls-init}. On targets that do not support symbol
2165 aliases, the default is @option{-fno-extern-tls-init}.
2168 @itemx -fno-for-scope
2170 @opindex fno-for-scope
2171 If @option{-ffor-scope} is specified, the scope of variables declared in
2172 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2173 as specified by the C++ standard.
2174 If @option{-fno-for-scope} is specified, the scope of variables declared in
2175 a @i{for-init-statement} extends to the end of the enclosing scope,
2176 as was the case in old versions of G++, and other (traditional)
2177 implementations of C++.
2179 If neither flag is given, the default is to follow the standard,
2180 but to allow and give a warning for old-style code that would
2181 otherwise be invalid, or have different behavior.
2183 @item -fno-gnu-keywords
2184 @opindex fno-gnu-keywords
2185 Do not recognize @code{typeof} as a keyword, so that code can use this
2186 word as an identifier. You can use the keyword @code{__typeof__} instead.
2187 @option{-ansi} implies @option{-fno-gnu-keywords}.
2189 @item -fno-implicit-templates
2190 @opindex fno-implicit-templates
2191 Never emit code for non-inline templates that are instantiated
2192 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2193 @xref{Template Instantiation}, for more information.
2195 @item -fno-implicit-inline-templates
2196 @opindex fno-implicit-inline-templates
2197 Don't emit code for implicit instantiations of inline templates, either.
2198 The default is to handle inlines differently so that compiles with and
2199 without optimization need the same set of explicit instantiations.
2201 @item -fno-implement-inlines
2202 @opindex fno-implement-inlines
2203 To save space, do not emit out-of-line copies of inline functions
2204 controlled by @samp{#pragma implementation}. This causes linker
2205 errors if these functions are not inlined everywhere they are called.
2207 @item -fms-extensions
2208 @opindex fms-extensions
2209 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2210 int and getting a pointer to member function via non-standard syntax.
2212 @item -fno-nonansi-builtins
2213 @opindex fno-nonansi-builtins
2214 Disable built-in declarations of functions that are not mandated by
2215 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2216 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2219 @opindex fnothrow-opt
2220 Treat a @code{throw()} exception specification as if it were a
2221 @code{noexcept} specification to reduce or eliminate the text size
2222 overhead relative to a function with no exception specification. If
2223 the function has local variables of types with non-trivial
2224 destructors, the exception specification actually makes the
2225 function smaller because the EH cleanups for those variables can be
2226 optimized away. The semantic effect is that an exception thrown out of
2227 a function with such an exception specification results in a call
2228 to @code{terminate} rather than @code{unexpected}.
2230 @item -fno-operator-names
2231 @opindex fno-operator-names
2232 Do not treat the operator name keywords @code{and}, @code{bitand},
2233 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2234 synonyms as keywords.
2236 @item -fno-optional-diags
2237 @opindex fno-optional-diags
2238 Disable diagnostics that the standard says a compiler does not need to
2239 issue. Currently, the only such diagnostic issued by G++ is the one for
2240 a name having multiple meanings within a class.
2243 @opindex fpermissive
2244 Downgrade some diagnostics about nonconformant code from errors to
2245 warnings. Thus, using @option{-fpermissive} allows some
2246 nonconforming code to compile.
2248 @item -fno-pretty-templates
2249 @opindex fno-pretty-templates
2250 When an error message refers to a specialization of a function
2251 template, the compiler normally prints the signature of the
2252 template followed by the template arguments and any typedefs or
2253 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2254 rather than @code{void f(int)}) so that it's clear which template is
2255 involved. When an error message refers to a specialization of a class
2256 template, the compiler omits any template arguments that match
2257 the default template arguments for that template. If either of these
2258 behaviors make it harder to understand the error message rather than
2259 easier, you can use @option{-fno-pretty-templates} to disable them.
2263 Enable automatic template instantiation at link time. This option also
2264 implies @option{-fno-implicit-templates}. @xref{Template
2265 Instantiation}, for more information.
2269 Disable generation of information about every class with virtual
2270 functions for use by the C++ run-time type identification features
2271 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2272 of the language, you can save some space by using this flag. Note that
2273 exception handling uses the same information, but G++ generates it as
2274 needed. The @samp{dynamic_cast} operator can still be used for casts that
2275 do not require run-time type information, i.e.@: casts to @code{void *} or to
2276 unambiguous base classes.
2280 Emit statistics about front-end processing at the end of the compilation.
2281 This information is generally only useful to the G++ development team.
2283 @item -fstrict-enums
2284 @opindex fstrict-enums
2285 Allow the compiler to optimize using the assumption that a value of
2286 enumerated type can only be one of the values of the enumeration (as
2287 defined in the C++ standard; basically, a value that can be
2288 represented in the minimum number of bits needed to represent all the
2289 enumerators). This assumption may not be valid if the program uses a
2290 cast to convert an arbitrary integer value to the enumerated type.
2292 @item -ftemplate-backtrace-limit=@var{n}
2293 @opindex ftemplate-backtrace-limit
2294 Set the maximum number of template instantiation notes for a single
2295 warning or error to @var{n}. The default value is 10.
2297 @item -ftemplate-depth=@var{n}
2298 @opindex ftemplate-depth
2299 Set the maximum instantiation depth for template classes to @var{n}.
2300 A limit on the template instantiation depth is needed to detect
2301 endless recursions during template class instantiation. ANSI/ISO C++
2302 conforming programs must not rely on a maximum depth greater than 17
2303 (changed to 1024 in C++11). The default value is 900, as the compiler
2304 can run out of stack space before hitting 1024 in some situations.
2306 @item -fno-threadsafe-statics
2307 @opindex fno-threadsafe-statics
2308 Do not emit the extra code to use the routines specified in the C++
2309 ABI for thread-safe initialization of local statics. You can use this
2310 option to reduce code size slightly in code that doesn't need to be
2313 @item -fuse-cxa-atexit
2314 @opindex fuse-cxa-atexit
2315 Register destructors for objects with static storage duration with the
2316 @code{__cxa_atexit} function rather than the @code{atexit} function.
2317 This option is required for fully standards-compliant handling of static
2318 destructors, but only works if your C library supports
2319 @code{__cxa_atexit}.
2321 @item -fno-use-cxa-get-exception-ptr
2322 @opindex fno-use-cxa-get-exception-ptr
2323 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2324 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2325 if the runtime routine is not available.
2327 @item -fvisibility-inlines-hidden
2328 @opindex fvisibility-inlines-hidden
2329 This switch declares that the user does not attempt to compare
2330 pointers to inline functions or methods where the addresses of the two functions
2331 are taken in different shared objects.
2333 The effect of this is that GCC may, effectively, mark inline methods with
2334 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2335 appear in the export table of a DSO and do not require a PLT indirection
2336 when used within the DSO@. Enabling this option can have a dramatic effect
2337 on load and link times of a DSO as it massively reduces the size of the
2338 dynamic export table when the library makes heavy use of templates.
2340 The behavior of this switch is not quite the same as marking the
2341 methods as hidden directly, because it does not affect static variables
2342 local to the function or cause the compiler to deduce that
2343 the function is defined in only one shared object.
2345 You may mark a method as having a visibility explicitly to negate the
2346 effect of the switch for that method. For example, if you do want to
2347 compare pointers to a particular inline method, you might mark it as
2348 having default visibility. Marking the enclosing class with explicit
2349 visibility has no effect.
2351 Explicitly instantiated inline methods are unaffected by this option
2352 as their linkage might otherwise cross a shared library boundary.
2353 @xref{Template Instantiation}.
2355 @item -fvisibility-ms-compat
2356 @opindex fvisibility-ms-compat
2357 This flag attempts to use visibility settings to make GCC's C++
2358 linkage model compatible with that of Microsoft Visual Studio.
2360 The flag makes these changes to GCC's linkage model:
2364 It sets the default visibility to @code{hidden}, like
2365 @option{-fvisibility=hidden}.
2368 Types, but not their members, are not hidden by default.
2371 The One Definition Rule is relaxed for types without explicit
2372 visibility specifications that are defined in more than one
2373 shared object: those declarations are permitted if they are
2374 permitted when this option is not used.
2377 In new code it is better to use @option{-fvisibility=hidden} and
2378 export those classes that are intended to be externally visible.
2379 Unfortunately it is possible for code to rely, perhaps accidentally,
2380 on the Visual Studio behavior.
2382 Among the consequences of these changes are that static data members
2383 of the same type with the same name but defined in different shared
2384 objects are different, so changing one does not change the other;
2385 and that pointers to function members defined in different shared
2386 objects may not compare equal. When this flag is given, it is a
2387 violation of the ODR to define types with the same name differently.
2389 @item -fvtable-verify=@var{std|preinit|none}
2390 @opindex fvtable-verify
2391 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2392 feature that verifies at runtime, for every virtual call that is made, that
2393 the vtable pointer through which the call is made is valid for the type of
2394 the object, and has not been corrupted or overwritten. If an invalid vtable
2395 pointer is detected (at runtime), an error is reported and execution of the
2396 program is immediately halted.
2398 This option causes runtime data structures to be built, at program start up,
2399 for verifying the vtable pointers. The options @code{std} and @code{preinit}
2400 control the timing of when these data structures are built. In both cases the
2401 data structures are built before execution reaches 'main'. The
2402 @option{-fvtable-verify=std} causes these data structure to be built after the
2403 shared libraries have been loaded and initialized.
2404 @option{-fvtable-verify=preinit} causes them to be built before the shared
2405 libraries have been loaded and initialized.
2407 If this option appears multiple times in the compiler line, with different
2408 values specified, 'none' will take highest priority over both 'std' and
2409 'preinit'; 'preinit' will take priority over 'std'.
2412 @opindex (fvtv-debug)
2413 Causes debug versions of the runtime functions for the vtable verification
2414 feature to be called. This assumes the @option{-fvtable-verify=std} or
2415 @option{-fvtable-verify=preinit} has been used. This flag will also cause the
2416 compiler to keep track of which vtable pointers it found for each class, and
2417 record that information in the file ``vtv_set_ptr_data.log'', in the dump
2418 file directory on the user's machine.
2420 Note: This feature APPENDS data to the log file. If you want a fresh log
2421 file, be sure to delete any existing one.
2424 @opindex fvtv-counts
2425 This is a debugging flag. When used in conjunction with
2426 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2427 causes the compiler to keep track of the total number of virtual calls
2428 it encountered and the number of verifications it inserted. It also
2429 counts the number of calls to certain runtime library functions
2430 that it inserts. This information, for each compilation unit, is written
2431 to a file named ``vtv_count_data.log'', in the dump_file directory on
2432 the user's machine. It also counts the size of the vtable pointer sets
2433 for each class, and writes this information to ``vtv_class_set_sizes.log''
2434 in the same directory.
2436 Note: This feature APPENDS data to the log files. To get a fresh log
2437 files, be sure to delete any existing ones.
2441 Do not use weak symbol support, even if it is provided by the linker.
2442 By default, G++ uses weak symbols if they are available. This
2443 option exists only for testing, and should not be used by end-users;
2444 it results in inferior code and has no benefits. This option may
2445 be removed in a future release of G++.
2449 Do not search for header files in the standard directories specific to
2450 C++, but do still search the other standard directories. (This option
2451 is used when building the C++ library.)
2454 In addition, these optimization, warning, and code generation options
2455 have meanings only for C++ programs:
2458 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2461 Warn when G++ generates code that is probably not compatible with the
2462 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2463 all such cases, there are probably some cases that are not warned about,
2464 even though G++ is generating incompatible code. There may also be
2465 cases where warnings are emitted even though the code that is generated
2468 You should rewrite your code to avoid these warnings if you are
2469 concerned about the fact that code generated by G++ may not be binary
2470 compatible with code generated by other compilers.
2472 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2477 A template with a non-type template parameter of reference type is
2478 mangled incorrectly:
2481 template <int &> struct S @{@};
2485 This is fixed in @option{-fabi-version=3}.
2488 SIMD vector types declared using @code{__attribute ((vector_size))} are
2489 mangled in a non-standard way that does not allow for overloading of
2490 functions taking vectors of different sizes.
2492 The mangling is changed in @option{-fabi-version=4}.
2495 The known incompatibilities in @option{-fabi-version=1} include:
2500 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2501 pack data into the same byte as a base class. For example:
2504 struct A @{ virtual void f(); int f1 : 1; @};
2505 struct B : public A @{ int f2 : 1; @};
2509 In this case, G++ places @code{B::f2} into the same byte
2510 as @code{A::f1}; other compilers do not. You can avoid this problem
2511 by explicitly padding @code{A} so that its size is a multiple of the
2512 byte size on your platform; that causes G++ and other compilers to
2513 lay out @code{B} identically.
2516 Incorrect handling of tail-padding for virtual bases. G++ does not use
2517 tail padding when laying out virtual bases. For example:
2520 struct A @{ virtual void f(); char c1; @};
2521 struct B @{ B(); char c2; @};
2522 struct C : public A, public virtual B @{@};
2526 In this case, G++ does not place @code{B} into the tail-padding for
2527 @code{A}; other compilers do. You can avoid this problem by
2528 explicitly padding @code{A} so that its size is a multiple of its
2529 alignment (ignoring virtual base classes); that causes G++ and other
2530 compilers to lay out @code{C} identically.
2533 Incorrect handling of bit-fields with declared widths greater than that
2534 of their underlying types, when the bit-fields appear in a union. For
2538 union U @{ int i : 4096; @};
2542 Assuming that an @code{int} does not have 4096 bits, G++ makes the
2543 union too small by the number of bits in an @code{int}.
2546 Empty classes can be placed at incorrect offsets. For example:
2556 struct C : public B, public A @{@};
2560 G++ places the @code{A} base class of @code{C} at a nonzero offset;
2561 it should be placed at offset zero. G++ mistakenly believes that the
2562 @code{A} data member of @code{B} is already at offset zero.
2565 Names of template functions whose types involve @code{typename} or
2566 template template parameters can be mangled incorrectly.
2569 template <typename Q>
2570 void f(typename Q::X) @{@}
2572 template <template <typename> class Q>
2573 void f(typename Q<int>::X) @{@}
2577 Instantiations of these templates may be mangled incorrectly.
2581 It also warns about psABI-related changes. The known psABI changes at this
2587 For SysV/x86-64, unions with @code{long double} members are
2588 passed in memory as specified in psABI. For example:
2598 @code{union U} is always passed in memory.
2602 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2603 @opindex Wctor-dtor-privacy
2604 @opindex Wno-ctor-dtor-privacy
2605 Warn when a class seems unusable because all the constructors or
2606 destructors in that class are private, and it has neither friends nor
2607 public static member functions. Also warn if there are no non-private
2608 methods, and there's at least one private member function that isn't
2609 a constructor or destructor.
2611 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2612 @opindex Wdelete-non-virtual-dtor
2613 @opindex Wno-delete-non-virtual-dtor
2614 Warn when @samp{delete} is used to destroy an instance of a class that
2615 has virtual functions and non-virtual destructor. It is unsafe to delete
2616 an instance of a derived class through a pointer to a base class if the
2617 base class does not have a virtual destructor. This warning is enabled
2620 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2621 @opindex Wliteral-suffix
2622 @opindex Wno-literal-suffix
2623 Warn when a string or character literal is followed by a ud-suffix which does
2624 not begin with an underscore. As a conforming extension, GCC treats such
2625 suffixes as separate preprocessing tokens in order to maintain backwards
2626 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2630 #define __STDC_FORMAT_MACROS
2631 #include <inttypes.h>
2636 printf("My int64: %"PRId64"\n", i64);
2640 In this case, @code{PRId64} is treated as a separate preprocessing token.
2642 This warning is enabled by default.
2644 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2646 @opindex Wno-narrowing
2647 Warn when a narrowing conversion prohibited by C++11 occurs within
2651 int i = @{ 2.2 @}; // error: narrowing from double to int
2654 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2656 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2657 required by the standard. Note that this does not affect the meaning
2658 of well-formed code; narrowing conversions are still considered
2659 ill-formed in SFINAE context.
2661 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2663 @opindex Wno-noexcept
2664 Warn when a noexcept-expression evaluates to false because of a call
2665 to a function that does not have a non-throwing exception
2666 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2667 the compiler to never throw an exception.
2669 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2670 @opindex Wnon-virtual-dtor
2671 @opindex Wno-non-virtual-dtor
2672 Warn when a class has virtual functions and an accessible non-virtual
2673 destructor itself or in a base class, or has in which case it is
2674 possible but unsafe to delete an instance of a derived class through a
2675 pointer to the base class. This warning is automatically enabled if
2676 @option{-Weffc++} is specified.
2678 @item -Wreorder @r{(C++ and Objective-C++ only)}
2680 @opindex Wno-reorder
2681 @cindex reordering, warning
2682 @cindex warning for reordering of member initializers
2683 Warn when the order of member initializers given in the code does not
2684 match the order in which they must be executed. For instance:
2690 A(): j (0), i (1) @{ @}
2695 The compiler rearranges the member initializers for @samp{i}
2696 and @samp{j} to match the declaration order of the members, emitting
2697 a warning to that effect. This warning is enabled by @option{-Wall}.
2699 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2700 @opindex fext-numeric-literals
2701 @opindex fno-ext-numeric-literals
2702 Accept imaginary, fixed-point, or machine-defined
2703 literal number suffixes as GNU extensions.
2704 When this option is turned off these suffixes are treated
2705 as C++11 user-defined literal numeric suffixes.
2706 This is on by default for all pre-C++11 dialects and all GNU dialects:
2707 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2708 @option{-std=gnu++1y}.
2709 This option is off by default
2710 for ISO C++11 onwards (@option{-std=c++11}, ...).
2713 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2716 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2719 Warn about violations of the following style guidelines from Scott Meyers'
2720 @cite{Effective C++} series of books:
2724 Define a copy constructor and an assignment operator for classes
2725 with dynamically-allocated memory.
2728 Prefer initialization to assignment in constructors.
2731 Have @code{operator=} return a reference to @code{*this}.
2734 Don't try to return a reference when you must return an object.
2737 Distinguish between prefix and postfix forms of increment and
2738 decrement operators.
2741 Never overload @code{&&}, @code{||}, or @code{,}.
2745 This option also enables @option{-Wnon-virtual-dtor}, which is also
2746 one of the effective C++ recommendations.
2748 When selecting this option, be aware that the standard library
2749 headers do not obey all of these guidelines; use @samp{grep -v}
2750 to filter out those warnings.
2752 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2753 @opindex Wstrict-null-sentinel
2754 @opindex Wno-strict-null-sentinel
2755 Warn about the use of an uncasted @code{NULL} as sentinel. When
2756 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2757 to @code{__null}. Although it is a null pointer constant rather than a
2758 null pointer, it is guaranteed to be of the same size as a pointer.
2759 But this use is not portable across different compilers.
2761 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2762 @opindex Wno-non-template-friend
2763 @opindex Wnon-template-friend
2764 Disable warnings when non-templatized friend functions are declared
2765 within a template. Since the advent of explicit template specification
2766 support in G++, if the name of the friend is an unqualified-id (i.e.,
2767 @samp{friend foo(int)}), the C++ language specification demands that the
2768 friend declare or define an ordinary, nontemplate function. (Section
2769 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2770 could be interpreted as a particular specialization of a templatized
2771 function. Because this non-conforming behavior is no longer the default
2772 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2773 check existing code for potential trouble spots and is on by default.
2774 This new compiler behavior can be turned off with
2775 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2776 but disables the helpful warning.
2778 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2779 @opindex Wold-style-cast
2780 @opindex Wno-old-style-cast
2781 Warn if an old-style (C-style) cast to a non-void type is used within
2782 a C++ program. The new-style casts (@samp{dynamic_cast},
2783 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2784 less vulnerable to unintended effects and much easier to search for.
2786 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2787 @opindex Woverloaded-virtual
2788 @opindex Wno-overloaded-virtual
2789 @cindex overloaded virtual function, warning
2790 @cindex warning for overloaded virtual function
2791 Warn when a function declaration hides virtual functions from a
2792 base class. For example, in:
2799 struct B: public A @{
2804 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2815 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2816 @opindex Wno-pmf-conversions
2817 @opindex Wpmf-conversions
2818 Disable the diagnostic for converting a bound pointer to member function
2821 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2822 @opindex Wsign-promo
2823 @opindex Wno-sign-promo
2824 Warn when overload resolution chooses a promotion from unsigned or
2825 enumerated type to a signed type, over a conversion to an unsigned type of
2826 the same size. Previous versions of G++ tried to preserve
2827 unsignedness, but the standard mandates the current behavior.
2830 @node Objective-C and Objective-C++ Dialect Options
2831 @section Options Controlling Objective-C and Objective-C++ Dialects
2833 @cindex compiler options, Objective-C and Objective-C++
2834 @cindex Objective-C and Objective-C++ options, command-line
2835 @cindex options, Objective-C and Objective-C++
2836 (NOTE: This manual does not describe the Objective-C and Objective-C++
2837 languages themselves. @xref{Standards,,Language Standards
2838 Supported by GCC}, for references.)
2840 This section describes the command-line options that are only meaningful
2841 for Objective-C and Objective-C++ programs. You can also use most of
2842 the language-independent GNU compiler options.
2843 For example, you might compile a file @code{some_class.m} like this:
2846 gcc -g -fgnu-runtime -O -c some_class.m
2850 In this example, @option{-fgnu-runtime} is an option meant only for
2851 Objective-C and Objective-C++ programs; you can use the other options with
2852 any language supported by GCC@.
2854 Note that since Objective-C is an extension of the C language, Objective-C
2855 compilations may also use options specific to the C front-end (e.g.,
2856 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2857 C++-specific options (e.g., @option{-Wabi}).
2859 Here is a list of options that are @emph{only} for compiling Objective-C
2860 and Objective-C++ programs:
2863 @item -fconstant-string-class=@var{class-name}
2864 @opindex fconstant-string-class
2865 Use @var{class-name} as the name of the class to instantiate for each
2866 literal string specified with the syntax @code{@@"@dots{}"}. The default
2867 class name is @code{NXConstantString} if the GNU runtime is being used, and
2868 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2869 @option{-fconstant-cfstrings} option, if also present, overrides the
2870 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2871 to be laid out as constant CoreFoundation strings.
2874 @opindex fgnu-runtime
2875 Generate object code compatible with the standard GNU Objective-C
2876 runtime. This is the default for most types of systems.
2878 @item -fnext-runtime
2879 @opindex fnext-runtime
2880 Generate output compatible with the NeXT runtime. This is the default
2881 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2882 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2885 @item -fno-nil-receivers
2886 @opindex fno-nil-receivers
2887 Assume that all Objective-C message dispatches (@code{[receiver
2888 message:arg]}) in this translation unit ensure that the receiver is
2889 not @code{nil}. This allows for more efficient entry points in the
2890 runtime to be used. This option is only available in conjunction with
2891 the NeXT runtime and ABI version 0 or 1.
2893 @item -fobjc-abi-version=@var{n}
2894 @opindex fobjc-abi-version
2895 Use version @var{n} of the Objective-C ABI for the selected runtime.
2896 This option is currently supported only for the NeXT runtime. In that
2897 case, Version 0 is the traditional (32-bit) ABI without support for
2898 properties and other Objective-C 2.0 additions. Version 1 is the
2899 traditional (32-bit) ABI with support for properties and other
2900 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2901 nothing is specified, the default is Version 0 on 32-bit target
2902 machines, and Version 2 on 64-bit target machines.
2904 @item -fobjc-call-cxx-cdtors
2905 @opindex fobjc-call-cxx-cdtors
2906 For each Objective-C class, check if any of its instance variables is a
2907 C++ object with a non-trivial default constructor. If so, synthesize a
2908 special @code{- (id) .cxx_construct} instance method which runs
2909 non-trivial default constructors on any such instance variables, in order,
2910 and then return @code{self}. Similarly, check if any instance variable
2911 is a C++ object with a non-trivial destructor, and if so, synthesize a
2912 special @code{- (void) .cxx_destruct} method which runs
2913 all such default destructors, in reverse order.
2915 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2916 methods thusly generated only operate on instance variables
2917 declared in the current Objective-C class, and not those inherited
2918 from superclasses. It is the responsibility of the Objective-C
2919 runtime to invoke all such methods in an object's inheritance
2920 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2921 by the runtime immediately after a new object instance is allocated;
2922 the @code{- (void) .cxx_destruct} methods are invoked immediately
2923 before the runtime deallocates an object instance.
2925 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2926 support for invoking the @code{- (id) .cxx_construct} and
2927 @code{- (void) .cxx_destruct} methods.
2929 @item -fobjc-direct-dispatch
2930 @opindex fobjc-direct-dispatch
2931 Allow fast jumps to the message dispatcher. On Darwin this is
2932 accomplished via the comm page.
2934 @item -fobjc-exceptions
2935 @opindex fobjc-exceptions
2936 Enable syntactic support for structured exception handling in
2937 Objective-C, similar to what is offered by C++ and Java. This option
2938 is required to use the Objective-C keywords @code{@@try},
2939 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2940 @code{@@synchronized}. This option is available with both the GNU
2941 runtime and the NeXT runtime (but not available in conjunction with
2942 the NeXT runtime on Mac OS X 10.2 and earlier).
2946 Enable garbage collection (GC) in Objective-C and Objective-C++
2947 programs. This option is only available with the NeXT runtime; the
2948 GNU runtime has a different garbage collection implementation that
2949 does not require special compiler flags.
2951 @item -fobjc-nilcheck
2952 @opindex fobjc-nilcheck
2953 For the NeXT runtime with version 2 of the ABI, check for a nil
2954 receiver in method invocations before doing the actual method call.
2955 This is the default and can be disabled using
2956 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2957 checked for nil in this way no matter what this flag is set to.
2958 Currently this flag does nothing when the GNU runtime, or an older
2959 version of the NeXT runtime ABI, is used.
2961 @item -fobjc-std=objc1
2963 Conform to the language syntax of Objective-C 1.0, the language
2964 recognized by GCC 4.0. This only affects the Objective-C additions to
2965 the C/C++ language; it does not affect conformance to C/C++ standards,
2966 which is controlled by the separate C/C++ dialect option flags. When
2967 this option is used with the Objective-C or Objective-C++ compiler,
2968 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2969 This is useful if you need to make sure that your Objective-C code can
2970 be compiled with older versions of GCC@.
2972 @item -freplace-objc-classes
2973 @opindex freplace-objc-classes
2974 Emit a special marker instructing @command{ld(1)} not to statically link in
2975 the resulting object file, and allow @command{dyld(1)} to load it in at
2976 run time instead. This is used in conjunction with the Fix-and-Continue
2977 debugging mode, where the object file in question may be recompiled and
2978 dynamically reloaded in the course of program execution, without the need
2979 to restart the program itself. Currently, Fix-and-Continue functionality
2980 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2985 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2986 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2987 compile time) with static class references that get initialized at load time,
2988 which improves run-time performance. Specifying the @option{-fzero-link} flag
2989 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2990 to be retained. This is useful in Zero-Link debugging mode, since it allows
2991 for individual class implementations to be modified during program execution.
2992 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2993 regardless of command-line options.
2997 Dump interface declarations for all classes seen in the source file to a
2998 file named @file{@var{sourcename}.decl}.
3000 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3001 @opindex Wassign-intercept
3002 @opindex Wno-assign-intercept
3003 Warn whenever an Objective-C assignment is being intercepted by the
3006 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3007 @opindex Wno-protocol
3009 If a class is declared to implement a protocol, a warning is issued for
3010 every method in the protocol that is not implemented by the class. The
3011 default behavior is to issue a warning for every method not explicitly
3012 implemented in the class, even if a method implementation is inherited
3013 from the superclass. If you use the @option{-Wno-protocol} option, then
3014 methods inherited from the superclass are considered to be implemented,
3015 and no warning is issued for them.
3017 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3019 @opindex Wno-selector
3020 Warn if multiple methods of different types for the same selector are
3021 found during compilation. The check is performed on the list of methods
3022 in the final stage of compilation. Additionally, a check is performed
3023 for each selector appearing in a @code{@@selector(@dots{})}
3024 expression, and a corresponding method for that selector has been found
3025 during compilation. Because these checks scan the method table only at
3026 the end of compilation, these warnings are not produced if the final
3027 stage of compilation is not reached, for example because an error is
3028 found during compilation, or because the @option{-fsyntax-only} option is
3031 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3032 @opindex Wstrict-selector-match
3033 @opindex Wno-strict-selector-match
3034 Warn if multiple methods with differing argument and/or return types are
3035 found for a given selector when attempting to send a message using this
3036 selector to a receiver of type @code{id} or @code{Class}. When this flag
3037 is off (which is the default behavior), the compiler omits such warnings
3038 if any differences found are confined to types that share the same size
3041 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3042 @opindex Wundeclared-selector
3043 @opindex Wno-undeclared-selector
3044 Warn if a @code{@@selector(@dots{})} expression referring to an
3045 undeclared selector is found. A selector is considered undeclared if no
3046 method with that name has been declared before the
3047 @code{@@selector(@dots{})} expression, either explicitly in an
3048 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3049 an @code{@@implementation} section. This option always performs its
3050 checks as soon as a @code{@@selector(@dots{})} expression is found,
3051 while @option{-Wselector} only performs its checks in the final stage of
3052 compilation. This also enforces the coding style convention
3053 that methods and selectors must be declared before being used.
3055 @item -print-objc-runtime-info
3056 @opindex print-objc-runtime-info
3057 Generate C header describing the largest structure that is passed by
3062 @node Language Independent Options
3063 @section Options to Control Diagnostic Messages Formatting
3064 @cindex options to control diagnostics formatting
3065 @cindex diagnostic messages
3066 @cindex message formatting
3068 Traditionally, diagnostic messages have been formatted irrespective of
3069 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3070 options described below
3071 to control the formatting algorithm for diagnostic messages,
3072 e.g.@: how many characters per line, how often source location
3073 information should be reported. Note that some language front ends may not
3074 honor these options.
3077 @item -fmessage-length=@var{n}
3078 @opindex fmessage-length
3079 Try to format error messages so that they fit on lines of about @var{n}
3080 characters. The default is 72 characters for @command{g++} and 0 for the rest of
3081 the front ends supported by GCC@. If @var{n} is zero, then no
3082 line-wrapping is done; each error message appears on a single
3085 @item -fdiagnostics-show-location=once
3086 @opindex fdiagnostics-show-location
3087 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3088 reporter to emit source location information @emph{once}; that is, in
3089 case the message is too long to fit on a single physical line and has to
3090 be wrapped, the source location won't be emitted (as prefix) again,
3091 over and over, in subsequent continuation lines. This is the default
3094 @item -fdiagnostics-show-location=every-line
3095 Only meaningful in line-wrapping mode. Instructs the diagnostic
3096 messages reporter to emit the same source location information (as
3097 prefix) for physical lines that result from the process of breaking
3098 a message which is too long to fit on a single line.
3100 @item -fdiagnostics-color[=@var{WHEN}]
3101 @itemx -fno-diagnostics-color
3102 @opindex fdiagnostics-color
3103 @cindex highlight, color, colour
3104 @vindex GCC_COLORS @r{environment variable}
3105 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3106 or @samp{auto}. The default is @samp{never} if @env{GCC_COLORS} environment
3107 variable isn't present in the environment, and @samp{auto} otherwise.
3108 @samp{auto} means to use color only when the standard error is a terminal.
3109 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3110 aliases for @option{-fdiagnostics-color=always} and
3111 @option{-fdiagnostics-color=never}, respectively.
3113 The colors are defined by the environment variable @env{GCC_COLORS}.
3114 Its value is a colon-separated list of capabilities and Select Graphic
3115 Rendition (SGR) substrings. SGR commands are interpreted by the
3116 terminal or terminal emulator. (See the section in the documentation
3117 of your text terminal for permitted values and their meanings as
3118 character attributes.) These substring values are integers in decimal
3119 representation and can be concatenated with semicolons.
3120 Common values to concatenate include
3122 @samp{4} for underline,
3124 @samp{7} for inverse,
3125 @samp{39} for default foreground color,
3126 @samp{30} to @samp{37} for foreground colors,
3127 @samp{90} to @samp{97} for 16-color mode foreground colors,
3128 @samp{38;5;0} to @samp{38;5;255}
3129 for 88-color and 256-color modes foreground colors,
3130 @samp{49} for default background color,
3131 @samp{40} to @samp{47} for background colors,
3132 @samp{100} to @samp{107} for 16-color mode background colors,
3133 and @samp{48;5;0} to @samp{48;5;255}
3134 for 88-color and 256-color modes background colors.
3136 The default @env{GCC_COLORS} is
3137 @samp{error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01}
3138 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3139 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3140 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3141 string disables colors.
3142 Supported capabilities are as follows.
3146 @vindex error GCC_COLORS @r{capability}
3147 SGR substring for error: markers.
3150 @vindex warning GCC_COLORS @r{capability}
3151 SGR substring for warning: markers.
3154 @vindex note GCC_COLORS @r{capability}
3155 SGR substring for note: markers.
3158 @vindex caret GCC_COLORS @r{capability}
3159 SGR substring for caret line.
3162 @vindex locus GCC_COLORS @r{capability}
3163 SGR substring for location information, @samp{file:line} or
3164 @samp{file:line:column} etc.
3167 @vindex quote GCC_COLORS @r{capability}
3168 SGR substring for information printed within quotes.
3171 @item -fno-diagnostics-show-option
3172 @opindex fno-diagnostics-show-option
3173 @opindex fdiagnostics-show-option
3174 By default, each diagnostic emitted includes text indicating the
3175 command-line option that directly controls the diagnostic (if such an
3176 option is known to the diagnostic machinery). Specifying the
3177 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3179 @item -fno-diagnostics-show-caret
3180 @opindex fno-diagnostics-show-caret
3181 @opindex fdiagnostics-show-caret
3182 By default, each diagnostic emitted includes the original source line
3183 and a caret '^' indicating the column. This option suppresses this
3188 @node Warning Options
3189 @section Options to Request or Suppress Warnings
3190 @cindex options to control warnings
3191 @cindex warning messages
3192 @cindex messages, warning
3193 @cindex suppressing warnings
3195 Warnings are diagnostic messages that report constructions that
3196 are not inherently erroneous but that are risky or suggest there
3197 may have been an error.
3199 The following language-independent options do not enable specific
3200 warnings but control the kinds of diagnostics produced by GCC@.
3203 @cindex syntax checking
3205 @opindex fsyntax-only
3206 Check the code for syntax errors, but don't do anything beyond that.
3208 @item -fmax-errors=@var{n}
3209 @opindex fmax-errors
3210 Limits the maximum number of error messages to @var{n}, at which point
3211 GCC bails out rather than attempting to continue processing the source
3212 code. If @var{n} is 0 (the default), there is no limit on the number
3213 of error messages produced. If @option{-Wfatal-errors} is also
3214 specified, then @option{-Wfatal-errors} takes precedence over this
3219 Inhibit all warning messages.
3224 Make all warnings into errors.
3229 Make the specified warning into an error. The specifier for a warning
3230 is appended; for example @option{-Werror=switch} turns the warnings
3231 controlled by @option{-Wswitch} into errors. This switch takes a
3232 negative form, to be used to negate @option{-Werror} for specific
3233 warnings; for example @option{-Wno-error=switch} makes
3234 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3237 The warning message for each controllable warning includes the
3238 option that controls the warning. That option can then be used with
3239 @option{-Werror=} and @option{-Wno-error=} as described above.
3240 (Printing of the option in the warning message can be disabled using the
3241 @option{-fno-diagnostics-show-option} flag.)
3243 Note that specifying @option{-Werror=}@var{foo} automatically implies
3244 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3247 @item -Wfatal-errors
3248 @opindex Wfatal-errors
3249 @opindex Wno-fatal-errors
3250 This option causes the compiler to abort compilation on the first error
3251 occurred rather than trying to keep going and printing further error
3256 You can request many specific warnings with options beginning with
3257 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3258 implicit declarations. Each of these specific warning options also
3259 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3260 example, @option{-Wno-implicit}. This manual lists only one of the
3261 two forms, whichever is not the default. For further
3262 language-specific options also refer to @ref{C++ Dialect Options} and
3263 @ref{Objective-C and Objective-C++ Dialect Options}.
3265 When an unrecognized warning option is requested (e.g.,
3266 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3267 that the option is not recognized. However, if the @option{-Wno-} form
3268 is used, the behavior is slightly different: no diagnostic is
3269 produced for @option{-Wno-unknown-warning} unless other diagnostics
3270 are being produced. This allows the use of new @option{-Wno-} options
3271 with old compilers, but if something goes wrong, the compiler
3272 warns that an unrecognized option is present.
3279 Issue all the warnings demanded by strict ISO C and ISO C++;
3280 reject all programs that use forbidden extensions, and some other
3281 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3282 version of the ISO C standard specified by any @option{-std} option used.
3284 Valid ISO C and ISO C++ programs should compile properly with or without
3285 this option (though a rare few require @option{-ansi} or a
3286 @option{-std} option specifying the required version of ISO C)@. However,
3287 without this option, certain GNU extensions and traditional C and C++
3288 features are supported as well. With this option, they are rejected.
3290 @option{-Wpedantic} does not cause warning messages for use of the
3291 alternate keywords whose names begin and end with @samp{__}. Pedantic
3292 warnings are also disabled in the expression that follows
3293 @code{__extension__}. However, only system header files should use
3294 these escape routes; application programs should avoid them.
3295 @xref{Alternate Keywords}.
3297 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3298 C conformance. They soon find that it does not do quite what they want:
3299 it finds some non-ISO practices, but not all---only those for which
3300 ISO C @emph{requires} a diagnostic, and some others for which
3301 diagnostics have been added.
3303 A feature to report any failure to conform to ISO C might be useful in
3304 some instances, but would require considerable additional work and would
3305 be quite different from @option{-Wpedantic}. We don't have plans to
3306 support such a feature in the near future.
3308 Where the standard specified with @option{-std} represents a GNU
3309 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3310 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3311 extended dialect is based. Warnings from @option{-Wpedantic} are given
3312 where they are required by the base standard. (It does not make sense
3313 for such warnings to be given only for features not in the specified GNU
3314 C dialect, since by definition the GNU dialects of C include all
3315 features the compiler supports with the given option, and there would be
3316 nothing to warn about.)
3318 @item -pedantic-errors
3319 @opindex pedantic-errors
3320 Like @option{-Wpedantic}, except that errors are produced rather than
3326 This enables all the warnings about constructions that some users
3327 consider questionable, and that are easy to avoid (or modify to
3328 prevent the warning), even in conjunction with macros. This also
3329 enables some language-specific warnings described in @ref{C++ Dialect
3330 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3332 @option{-Wall} turns on the following warning flags:
3334 @gccoptlist{-Waddress @gol
3335 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3337 -Wchar-subscripts @gol
3338 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3339 -Wimplicit-int @r{(C and Objective-C only)} @gol
3340 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3343 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3344 -Wmaybe-uninitialized @gol
3345 -Wmissing-braces @r{(only for C/ObjC)} @gol
3352 -Wsequence-point @gol
3353 -Wsign-compare @r{(only in C++)} @gol
3354 -Wstrict-aliasing @gol
3355 -Wstrict-overflow=1 @gol
3358 -Wuninitialized @gol
3359 -Wunknown-pragmas @gol
3360 -Wunused-function @gol
3363 -Wunused-variable @gol
3364 -Wvolatile-register-var @gol
3367 Note that some warning flags are not implied by @option{-Wall}. Some of
3368 them warn about constructions that users generally do not consider
3369 questionable, but which occasionally you might wish to check for;
3370 others warn about constructions that are necessary or hard to avoid in
3371 some cases, and there is no simple way to modify the code to suppress
3372 the warning. Some of them are enabled by @option{-Wextra} but many of
3373 them must be enabled individually.
3379 This enables some extra warning flags that are not enabled by
3380 @option{-Wall}. (This option used to be called @option{-W}. The older
3381 name is still supported, but the newer name is more descriptive.)
3383 @gccoptlist{-Wclobbered @gol
3385 -Wignored-qualifiers @gol
3386 -Wmissing-field-initializers @gol
3387 -Wmissing-parameter-type @r{(C only)} @gol
3388 -Wold-style-declaration @r{(C only)} @gol
3389 -Woverride-init @gol
3392 -Wuninitialized @gol
3393 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3394 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3397 The option @option{-Wextra} also prints warning messages for the
3403 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3404 @samp{>}, or @samp{>=}.
3407 (C++ only) An enumerator and a non-enumerator both appear in a
3408 conditional expression.
3411 (C++ only) Ambiguous virtual bases.
3414 (C++ only) Subscripting an array that has been declared @samp{register}.
3417 (C++ only) Taking the address of a variable that has been declared
3421 (C++ only) A base class is not initialized in a derived class's copy
3426 @item -Wchar-subscripts
3427 @opindex Wchar-subscripts
3428 @opindex Wno-char-subscripts
3429 Warn if an array subscript has type @code{char}. This is a common cause
3430 of error, as programmers often forget that this type is signed on some
3432 This warning is enabled by @option{-Wall}.
3436 @opindex Wno-comment
3437 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3438 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3439 This warning is enabled by @option{-Wall}.
3441 @item -Wno-coverage-mismatch
3442 @opindex Wno-coverage-mismatch
3443 Warn if feedback profiles do not match when using the
3444 @option{-fprofile-use} option.
3445 If a source file is changed between compiling with @option{-fprofile-gen} and
3446 with @option{-fprofile-use}, the files with the profile feedback can fail
3447 to match the source file and GCC cannot use the profile feedback
3448 information. By default, this warning is enabled and is treated as an
3449 error. @option{-Wno-coverage-mismatch} can be used to disable the
3450 warning or @option{-Wno-error=coverage-mismatch} can be used to
3451 disable the error. Disabling the error for this warning can result in
3452 poorly optimized code and is useful only in the
3453 case of very minor changes such as bug fixes to an existing code-base.
3454 Completely disabling the warning is not recommended.
3457 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3459 Suppress warning messages emitted by @code{#warning} directives.
3461 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3462 @opindex Wdouble-promotion
3463 @opindex Wno-double-promotion
3464 Give a warning when a value of type @code{float} is implicitly
3465 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3466 floating-point unit implement @code{float} in hardware, but emulate
3467 @code{double} in software. On such a machine, doing computations
3468 using @code{double} values is much more expensive because of the
3469 overhead required for software emulation.
3471 It is easy to accidentally do computations with @code{double} because
3472 floating-point literals are implicitly of type @code{double}. For
3476 float area(float radius)
3478 return 3.14159 * radius * radius;
3482 the compiler performs the entire computation with @code{double}
3483 because the floating-point literal is a @code{double}.
3486 @itemx -Wformat=@var{n}
3489 @opindex ffreestanding
3490 @opindex fno-builtin
3492 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3493 the arguments supplied have types appropriate to the format string
3494 specified, and that the conversions specified in the format string make
3495 sense. This includes standard functions, and others specified by format
3496 attributes (@pxref{Function Attributes}), in the @code{printf},
3497 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3498 not in the C standard) families (or other target-specific families).
3499 Which functions are checked without format attributes having been
3500 specified depends on the standard version selected, and such checks of
3501 functions without the attribute specified are disabled by
3502 @option{-ffreestanding} or @option{-fno-builtin}.
3504 The formats are checked against the format features supported by GNU
3505 libc version 2.2. These include all ISO C90 and C99 features, as well
3506 as features from the Single Unix Specification and some BSD and GNU
3507 extensions. Other library implementations may not support all these
3508 features; GCC does not support warning about features that go beyond a
3509 particular library's limitations. However, if @option{-Wpedantic} is used
3510 with @option{-Wformat}, warnings are given about format features not
3511 in the selected standard version (but not for @code{strfmon} formats,
3512 since those are not in any version of the C standard). @xref{C Dialect
3513 Options,,Options Controlling C Dialect}.
3520 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3521 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3522 @option{-Wformat} also checks for null format arguments for several
3523 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3524 aspects of this level of format checking can be disabled by the
3525 options: @option{-Wno-format-contains-nul},
3526 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3527 @option{-Wformat} is enabled by @option{-Wall}.
3529 @item -Wno-format-contains-nul
3530 @opindex Wno-format-contains-nul
3531 @opindex Wformat-contains-nul
3532 If @option{-Wformat} is specified, do not warn about format strings that
3535 @item -Wno-format-extra-args
3536 @opindex Wno-format-extra-args
3537 @opindex Wformat-extra-args
3538 If @option{-Wformat} is specified, do not warn about excess arguments to a
3539 @code{printf} or @code{scanf} format function. The C standard specifies
3540 that such arguments are ignored.
3542 Where the unused arguments lie between used arguments that are
3543 specified with @samp{$} operand number specifications, normally
3544 warnings are still given, since the implementation could not know what
3545 type to pass to @code{va_arg} to skip the unused arguments. However,
3546 in the case of @code{scanf} formats, this option suppresses the
3547 warning if the unused arguments are all pointers, since the Single
3548 Unix Specification says that such unused arguments are allowed.
3550 @item -Wno-format-zero-length
3551 @opindex Wno-format-zero-length
3552 @opindex Wformat-zero-length
3553 If @option{-Wformat} is specified, do not warn about zero-length formats.
3554 The C standard specifies that zero-length formats are allowed.
3559 Enable @option{-Wformat} plus additional format checks. Currently
3560 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3563 @item -Wformat-nonliteral
3564 @opindex Wformat-nonliteral
3565 @opindex Wno-format-nonliteral
3566 If @option{-Wformat} is specified, also warn if the format string is not a
3567 string literal and so cannot be checked, unless the format function
3568 takes its format arguments as a @code{va_list}.
3570 @item -Wformat-security
3571 @opindex Wformat-security
3572 @opindex Wno-format-security
3573 If @option{-Wformat} is specified, also warn about uses of format
3574 functions that represent possible security problems. At present, this
3575 warns about calls to @code{printf} and @code{scanf} functions where the
3576 format string is not a string literal and there are no format arguments,
3577 as in @code{printf (foo);}. This may be a security hole if the format
3578 string came from untrusted input and contains @samp{%n}. (This is
3579 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3580 in future warnings may be added to @option{-Wformat-security} that are not
3581 included in @option{-Wformat-nonliteral}.)
3584 @opindex Wformat-y2k
3585 @opindex Wno-format-y2k
3586 If @option{-Wformat} is specified, also warn about @code{strftime}
3587 formats that may yield only a two-digit year.
3592 @opindex Wno-nonnull
3593 Warn about passing a null pointer for arguments marked as
3594 requiring a non-null value by the @code{nonnull} function attribute.
3596 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3597 can be disabled with the @option{-Wno-nonnull} option.
3599 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3601 @opindex Wno-init-self
3602 Warn about uninitialized variables that are initialized with themselves.
3603 Note this option can only be used with the @option{-Wuninitialized} option.
3605 For example, GCC warns about @code{i} being uninitialized in the
3606 following snippet only when @option{-Winit-self} has been specified:
3617 This warning is enabled by @option{-Wall} in C++.
3619 @item -Wimplicit-int @r{(C and Objective-C only)}
3620 @opindex Wimplicit-int
3621 @opindex Wno-implicit-int
3622 Warn when a declaration does not specify a type.
3623 This warning is enabled by @option{-Wall}.
3625 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3626 @opindex Wimplicit-function-declaration
3627 @opindex Wno-implicit-function-declaration
3628 Give a warning whenever a function is used before being declared. In
3629 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3630 enabled by default and it is made into an error by
3631 @option{-pedantic-errors}. This warning is also enabled by
3634 @item -Wimplicit @r{(C and Objective-C only)}
3636 @opindex Wno-implicit
3637 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3638 This warning is enabled by @option{-Wall}.
3640 @item -Wignored-qualifiers @r{(C and C++ only)}
3641 @opindex Wignored-qualifiers
3642 @opindex Wno-ignored-qualifiers
3643 Warn if the return type of a function has a type qualifier
3644 such as @code{const}. For ISO C such a type qualifier has no effect,
3645 since the value returned by a function is not an lvalue.
3646 For C++, the warning is only emitted for scalar types or @code{void}.
3647 ISO C prohibits qualified @code{void} return types on function
3648 definitions, so such return types always receive a warning
3649 even without this option.
3651 This warning is also enabled by @option{-Wextra}.
3656 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3657 a function with external linkage, returning int, taking either zero
3658 arguments, two, or three arguments of appropriate types. This warning
3659 is enabled by default in C++ and is enabled by either @option{-Wall}
3660 or @option{-Wpedantic}.
3662 @item -Wmissing-braces
3663 @opindex Wmissing-braces
3664 @opindex Wno-missing-braces
3665 Warn if an aggregate or union initializer is not fully bracketed. In
3666 the following example, the initializer for @samp{a} is not fully
3667 bracketed, but that for @samp{b} is fully bracketed. This warning is
3668 enabled by @option{-Wall} in C.
3671 int a[2][2] = @{ 0, 1, 2, 3 @};
3672 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3675 This warning is enabled by @option{-Wall}.
3677 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3678 @opindex Wmissing-include-dirs
3679 @opindex Wno-missing-include-dirs
3680 Warn if a user-supplied include directory does not exist.
3683 @opindex Wparentheses
3684 @opindex Wno-parentheses
3685 Warn if parentheses are omitted in certain contexts, such
3686 as when there is an assignment in a context where a truth value
3687 is expected, or when operators are nested whose precedence people
3688 often get confused about.
3690 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3691 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3692 interpretation from that of ordinary mathematical notation.
3694 Also warn about constructions where there may be confusion to which
3695 @code{if} statement an @code{else} branch belongs. Here is an example of
3710 In C/C++, every @code{else} branch belongs to the innermost possible
3711 @code{if} statement, which in this example is @code{if (b)}. This is
3712 often not what the programmer expected, as illustrated in the above
3713 example by indentation the programmer chose. When there is the
3714 potential for this confusion, GCC issues a warning when this flag
3715 is specified. To eliminate the warning, add explicit braces around
3716 the innermost @code{if} statement so there is no way the @code{else}
3717 can belong to the enclosing @code{if}. The resulting code
3734 Also warn for dangerous uses of the GNU extension to
3735 @code{?:} with omitted middle operand. When the condition
3736 in the @code{?}: operator is a boolean expression, the omitted value is
3737 always 1. Often programmers expect it to be a value computed
3738 inside the conditional expression instead.
3740 This warning is enabled by @option{-Wall}.
3742 @item -Wsequence-point
3743 @opindex Wsequence-point
3744 @opindex Wno-sequence-point
3745 Warn about code that may have undefined semantics because of violations
3746 of sequence point rules in the C and C++ standards.
3748 The C and C++ standards define the order in which expressions in a C/C++
3749 program are evaluated in terms of @dfn{sequence points}, which represent
3750 a partial ordering between the execution of parts of the program: those
3751 executed before the sequence point, and those executed after it. These
3752 occur after the evaluation of a full expression (one which is not part
3753 of a larger expression), after the evaluation of the first operand of a
3754 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3755 function is called (but after the evaluation of its arguments and the
3756 expression denoting the called function), and in certain other places.
3757 Other than as expressed by the sequence point rules, the order of
3758 evaluation of subexpressions of an expression is not specified. All
3759 these rules describe only a partial order rather than a total order,
3760 since, for example, if two functions are called within one expression
3761 with no sequence point between them, the order in which the functions
3762 are called is not specified. However, the standards committee have
3763 ruled that function calls do not overlap.
3765 It is not specified when between sequence points modifications to the
3766 values of objects take effect. Programs whose behavior depends on this
3767 have undefined behavior; the C and C++ standards specify that ``Between
3768 the previous and next sequence point an object shall have its stored
3769 value modified at most once by the evaluation of an expression.
3770 Furthermore, the prior value shall be read only to determine the value
3771 to be stored.''. If a program breaks these rules, the results on any
3772 particular implementation are entirely unpredictable.
3774 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3775 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3776 diagnosed by this option, and it may give an occasional false positive
3777 result, but in general it has been found fairly effective at detecting
3778 this sort of problem in programs.
3780 The standard is worded confusingly, therefore there is some debate
3781 over the precise meaning of the sequence point rules in subtle cases.
3782 Links to discussions of the problem, including proposed formal
3783 definitions, may be found on the GCC readings page, at
3784 @uref{http://gcc.gnu.org/@/readings.html}.
3786 This warning is enabled by @option{-Wall} for C and C++.
3788 @item -Wno-return-local-addr
3789 @opindex Wno-return-local-addr
3790 @opindex Wreturn-local-addr
3791 Do not warn about returning a pointer (or in C++, a reference) to a
3792 variable that goes out of scope after the function returns.
3795 @opindex Wreturn-type
3796 @opindex Wno-return-type
3797 Warn whenever a function is defined with a return type that defaults
3798 to @code{int}. Also warn about any @code{return} statement with no
3799 return value in a function whose return type is not @code{void}
3800 (falling off the end of the function body is considered returning
3801 without a value), and about a @code{return} statement with an
3802 expression in a function whose return type is @code{void}.
3804 For C++, a function without return type always produces a diagnostic
3805 message, even when @option{-Wno-return-type} is specified. The only
3806 exceptions are @samp{main} and functions defined in system headers.
3808 This warning is enabled by @option{-Wall}.
3813 Warn whenever a @code{switch} statement has an index of enumerated type
3814 and lacks a @code{case} for one or more of the named codes of that
3815 enumeration. (The presence of a @code{default} label prevents this
3816 warning.) @code{case} labels outside the enumeration range also
3817 provoke warnings when this option is used (even if there is a
3818 @code{default} label).
3819 This warning is enabled by @option{-Wall}.
3821 @item -Wswitch-default
3822 @opindex Wswitch-default
3823 @opindex Wno-switch-default
3824 Warn whenever a @code{switch} statement does not have a @code{default}
3828 @opindex Wswitch-enum
3829 @opindex Wno-switch-enum
3830 Warn whenever a @code{switch} statement has an index of enumerated type
3831 and lacks a @code{case} for one or more of the named codes of that
3832 enumeration. @code{case} labels outside the enumeration range also
3833 provoke warnings when this option is used. The only difference
3834 between @option{-Wswitch} and this option is that this option gives a
3835 warning about an omitted enumeration code even if there is a
3836 @code{default} label.
3838 @item -Wsync-nand @r{(C and C++ only)}
3840 @opindex Wno-sync-nand
3841 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3842 built-in functions are used. These functions changed semantics in GCC 4.4.
3846 @opindex Wno-trigraphs
3847 Warn if any trigraphs are encountered that might change the meaning of
3848 the program (trigraphs within comments are not warned about).
3849 This warning is enabled by @option{-Wall}.
3851 @item -Wunused-but-set-parameter
3852 @opindex Wunused-but-set-parameter
3853 @opindex Wno-unused-but-set-parameter
3854 Warn whenever a function parameter is assigned to, but otherwise unused
3855 (aside from its declaration).
3857 To suppress this warning use the @samp{unused} attribute
3858 (@pxref{Variable Attributes}).
3860 This warning is also enabled by @option{-Wunused} together with
3863 @item -Wunused-but-set-variable
3864 @opindex Wunused-but-set-variable
3865 @opindex Wno-unused-but-set-variable
3866 Warn whenever a local variable is assigned to, but otherwise unused
3867 (aside from its declaration).
3868 This warning is enabled by @option{-Wall}.
3870 To suppress this warning use the @samp{unused} attribute
3871 (@pxref{Variable Attributes}).
3873 This warning is also enabled by @option{-Wunused}, which is enabled
3876 @item -Wunused-function
3877 @opindex Wunused-function
3878 @opindex Wno-unused-function
3879 Warn whenever a static function is declared but not defined or a
3880 non-inline static function is unused.
3881 This warning is enabled by @option{-Wall}.
3883 @item -Wunused-label
3884 @opindex Wunused-label
3885 @opindex Wno-unused-label
3886 Warn whenever a label is declared but not used.
3887 This warning is enabled by @option{-Wall}.
3889 To suppress this warning use the @samp{unused} attribute
3890 (@pxref{Variable Attributes}).
3892 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3893 @opindex Wunused-local-typedefs
3894 Warn when a typedef locally defined in a function is not used.
3895 This warning is enabled by @option{-Wall}.
3897 @item -Wunused-parameter
3898 @opindex Wunused-parameter
3899 @opindex Wno-unused-parameter
3900 Warn whenever a function parameter is unused aside from its declaration.
3902 To suppress this warning use the @samp{unused} attribute
3903 (@pxref{Variable Attributes}).
3905 @item -Wno-unused-result
3906 @opindex Wunused-result
3907 @opindex Wno-unused-result
3908 Do not warn if a caller of a function marked with attribute
3909 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3910 its return value. The default is @option{-Wunused-result}.
3912 @item -Wunused-variable
3913 @opindex Wunused-variable
3914 @opindex Wno-unused-variable
3915 Warn whenever a local variable or non-constant static variable is unused
3916 aside from its declaration.
3917 This warning is enabled by @option{-Wall}.
3919 To suppress this warning use the @samp{unused} attribute
3920 (@pxref{Variable Attributes}).
3922 @item -Wunused-value
3923 @opindex Wunused-value
3924 @opindex Wno-unused-value
3925 Warn whenever a statement computes a result that is explicitly not
3926 used. To suppress this warning cast the unused expression to
3927 @samp{void}. This includes an expression-statement or the left-hand
3928 side of a comma expression that contains no side effects. For example,
3929 an expression such as @samp{x[i,j]} causes a warning, while
3930 @samp{x[(void)i,j]} does not.
3932 This warning is enabled by @option{-Wall}.
3937 All the above @option{-Wunused} options combined.
3939 In order to get a warning about an unused function parameter, you must
3940 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
3941 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
3943 @item -Wuninitialized
3944 @opindex Wuninitialized
3945 @opindex Wno-uninitialized
3946 Warn if an automatic variable is used without first being initialized
3947 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3948 warn if a non-static reference or non-static @samp{const} member
3949 appears in a class without constructors.
3951 If you want to warn about code that uses the uninitialized value of the
3952 variable in its own initializer, use the @option{-Winit-self} option.
3954 These warnings occur for individual uninitialized or clobbered
3955 elements of structure, union or array variables as well as for
3956 variables that are uninitialized or clobbered as a whole. They do
3957 not occur for variables or elements declared @code{volatile}. Because
3958 these warnings depend on optimization, the exact variables or elements
3959 for which there are warnings depends on the precise optimization
3960 options and version of GCC used.
3962 Note that there may be no warning about a variable that is used only
3963 to compute a value that itself is never used, because such
3964 computations may be deleted by data flow analysis before the warnings
3967 @item -Wmaybe-uninitialized
3968 @opindex Wmaybe-uninitialized
3969 @opindex Wno-maybe-uninitialized
3970 For an automatic variable, if there exists a path from the function
3971 entry to a use of the variable that is initialized, but there exist
3972 some other paths for which the variable is not initialized, the compiler
3973 emits a warning if it cannot prove the uninitialized paths are not
3974 executed at run time. These warnings are made optional because GCC is
3975 not smart enough to see all the reasons why the code might be correct
3976 in spite of appearing to have an error. Here is one example of how
3997 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3998 always initialized, but GCC doesn't know this. To suppress the
3999 warning, you need to provide a default case with assert(0) or
4002 @cindex @code{longjmp} warnings
4003 This option also warns when a non-volatile automatic variable might be
4004 changed by a call to @code{longjmp}. These warnings as well are possible
4005 only in optimizing compilation.
4007 The compiler sees only the calls to @code{setjmp}. It cannot know
4008 where @code{longjmp} will be called; in fact, a signal handler could
4009 call it at any point in the code. As a result, you may get a warning
4010 even when there is in fact no problem because @code{longjmp} cannot
4011 in fact be called at the place that would cause a problem.
4013 Some spurious warnings can be avoided if you declare all the functions
4014 you use that never return as @code{noreturn}. @xref{Function
4017 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4019 @item -Wunknown-pragmas
4020 @opindex Wunknown-pragmas
4021 @opindex Wno-unknown-pragmas
4022 @cindex warning for unknown pragmas
4023 @cindex unknown pragmas, warning
4024 @cindex pragmas, warning of unknown
4025 Warn when a @code{#pragma} directive is encountered that is not understood by
4026 GCC@. If this command-line option is used, warnings are even issued
4027 for unknown pragmas in system header files. This is not the case if
4028 the warnings are only enabled by the @option{-Wall} command-line option.
4031 @opindex Wno-pragmas
4033 Do not warn about misuses of pragmas, such as incorrect parameters,
4034 invalid syntax, or conflicts between pragmas. See also
4035 @option{-Wunknown-pragmas}.
4037 @item -Wstrict-aliasing
4038 @opindex Wstrict-aliasing
4039 @opindex Wno-strict-aliasing
4040 This option is only active when @option{-fstrict-aliasing} is active.
4041 It warns about code that might break the strict aliasing rules that the
4042 compiler is using for optimization. The warning does not catch all
4043 cases, but does attempt to catch the more common pitfalls. It is
4044 included in @option{-Wall}.
4045 It is equivalent to @option{-Wstrict-aliasing=3}
4047 @item -Wstrict-aliasing=n
4048 @opindex Wstrict-aliasing=n
4049 This option is only active when @option{-fstrict-aliasing} is active.
4050 It warns about code that might break the strict aliasing rules that the
4051 compiler is using for optimization.
4052 Higher levels correspond to higher accuracy (fewer false positives).
4053 Higher levels also correspond to more effort, similar to the way @option{-O}
4055 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4057 Level 1: Most aggressive, quick, least accurate.
4058 Possibly useful when higher levels
4059 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4060 false negatives. However, it has many false positives.
4061 Warns for all pointer conversions between possibly incompatible types,
4062 even if never dereferenced. Runs in the front end only.
4064 Level 2: Aggressive, quick, not too precise.
4065 May still have many false positives (not as many as level 1 though),
4066 and few false negatives (but possibly more than level 1).
4067 Unlike level 1, it only warns when an address is taken. Warns about
4068 incomplete types. Runs in the front end only.
4070 Level 3 (default for @option{-Wstrict-aliasing}):
4071 Should have very few false positives and few false
4072 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4073 Takes care of the common pun+dereference pattern in the front end:
4074 @code{*(int*)&some_float}.
4075 If optimization is enabled, it also runs in the back end, where it deals
4076 with multiple statement cases using flow-sensitive points-to information.
4077 Only warns when the converted pointer is dereferenced.
4078 Does not warn about incomplete types.
4080 @item -Wstrict-overflow
4081 @itemx -Wstrict-overflow=@var{n}
4082 @opindex Wstrict-overflow
4083 @opindex Wno-strict-overflow
4084 This option is only active when @option{-fstrict-overflow} is active.
4085 It warns about cases where the compiler optimizes based on the
4086 assumption that signed overflow does not occur. Note that it does not
4087 warn about all cases where the code might overflow: it only warns
4088 about cases where the compiler implements some optimization. Thus
4089 this warning depends on the optimization level.
4091 An optimization that assumes that signed overflow does not occur is
4092 perfectly safe if the values of the variables involved are such that
4093 overflow never does, in fact, occur. Therefore this warning can
4094 easily give a false positive: a warning about code that is not
4095 actually a problem. To help focus on important issues, several
4096 warning levels are defined. No warnings are issued for the use of
4097 undefined signed overflow when estimating how many iterations a loop
4098 requires, in particular when determining whether a loop will be
4102 @item -Wstrict-overflow=1
4103 Warn about cases that are both questionable and easy to avoid. For
4104 example, with @option{-fstrict-overflow}, the compiler simplifies
4105 @code{x + 1 > x} to @code{1}. This level of
4106 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4107 are not, and must be explicitly requested.
4109 @item -Wstrict-overflow=2
4110 Also warn about other cases where a comparison is simplified to a
4111 constant. For example: @code{abs (x) >= 0}. This can only be
4112 simplified when @option{-fstrict-overflow} is in effect, because
4113 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4114 zero. @option{-Wstrict-overflow} (with no level) is the same as
4115 @option{-Wstrict-overflow=2}.
4117 @item -Wstrict-overflow=3
4118 Also warn about other cases where a comparison is simplified. For
4119 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4121 @item -Wstrict-overflow=4
4122 Also warn about other simplifications not covered by the above cases.
4123 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4125 @item -Wstrict-overflow=5
4126 Also warn about cases where the compiler reduces the magnitude of a
4127 constant involved in a comparison. For example: @code{x + 2 > y} is
4128 simplified to @code{x + 1 >= y}. This is reported only at the
4129 highest warning level because this simplification applies to many
4130 comparisons, so this warning level gives a very large number of
4134 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4135 @opindex Wsuggest-attribute=
4136 @opindex Wno-suggest-attribute=
4137 Warn for cases where adding an attribute may be beneficial. The
4138 attributes currently supported are listed below.
4141 @item -Wsuggest-attribute=pure
4142 @itemx -Wsuggest-attribute=const
4143 @itemx -Wsuggest-attribute=noreturn
4144 @opindex Wsuggest-attribute=pure
4145 @opindex Wno-suggest-attribute=pure
4146 @opindex Wsuggest-attribute=const
4147 @opindex Wno-suggest-attribute=const
4148 @opindex Wsuggest-attribute=noreturn
4149 @opindex Wno-suggest-attribute=noreturn
4151 Warn about functions that might be candidates for attributes
4152 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4153 functions visible in other compilation units or (in the case of @code{pure} and
4154 @code{const}) if it cannot prove that the function returns normally. A function
4155 returns normally if it doesn't contain an infinite loop or return abnormally
4156 by throwing, calling @code{abort()} or trapping. This analysis requires option
4157 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4158 higher. Higher optimization levels improve the accuracy of the analysis.
4160 @item -Wsuggest-attribute=format
4161 @itemx -Wmissing-format-attribute
4162 @opindex Wsuggest-attribute=format
4163 @opindex Wmissing-format-attribute
4164 @opindex Wno-suggest-attribute=format
4165 @opindex Wno-missing-format-attribute
4169 Warn about function pointers that might be candidates for @code{format}
4170 attributes. Note these are only possible candidates, not absolute ones.
4171 GCC guesses that function pointers with @code{format} attributes that
4172 are used in assignment, initialization, parameter passing or return
4173 statements should have a corresponding @code{format} attribute in the
4174 resulting type. I.e.@: the left-hand side of the assignment or
4175 initialization, the type of the parameter variable, or the return type
4176 of the containing function respectively should also have a @code{format}
4177 attribute to avoid the warning.
4179 GCC also warns about function definitions that might be
4180 candidates for @code{format} attributes. Again, these are only
4181 possible candidates. GCC guesses that @code{format} attributes
4182 might be appropriate for any function that calls a function like
4183 @code{vprintf} or @code{vscanf}, but this might not always be the
4184 case, and some functions for which @code{format} attributes are
4185 appropriate may not be detected.
4188 @item -Warray-bounds
4189 @opindex Wno-array-bounds
4190 @opindex Warray-bounds
4191 This option is only active when @option{-ftree-vrp} is active
4192 (default for @option{-O2} and above). It warns about subscripts to arrays
4193 that are always out of bounds. This warning is enabled by @option{-Wall}.
4195 @item -Wno-div-by-zero
4196 @opindex Wno-div-by-zero
4197 @opindex Wdiv-by-zero
4198 Do not warn about compile-time integer division by zero. Floating-point
4199 division by zero is not warned about, as it can be a legitimate way of
4200 obtaining infinities and NaNs.
4202 @item -Wsystem-headers
4203 @opindex Wsystem-headers
4204 @opindex Wno-system-headers
4205 @cindex warnings from system headers
4206 @cindex system headers, warnings from
4207 Print warning messages for constructs found in system header files.
4208 Warnings from system headers are normally suppressed, on the assumption
4209 that they usually do not indicate real problems and would only make the
4210 compiler output harder to read. Using this command-line option tells
4211 GCC to emit warnings from system headers as if they occurred in user
4212 code. However, note that using @option{-Wall} in conjunction with this
4213 option does @emph{not} warn about unknown pragmas in system
4214 headers---for that, @option{-Wunknown-pragmas} must also be used.
4217 @opindex Wtrampolines
4218 @opindex Wno-trampolines
4219 Warn about trampolines generated for pointers to nested functions.
4221 A trampoline is a small piece of data or code that is created at run
4222 time on the stack when the address of a nested function is taken, and
4223 is used to call the nested function indirectly. For some targets, it
4224 is made up of data only and thus requires no special treatment. But,
4225 for most targets, it is made up of code and thus requires the stack
4226 to be made executable in order for the program to work properly.
4229 @opindex Wfloat-equal
4230 @opindex Wno-float-equal
4231 Warn if floating-point values are used in equality comparisons.
4233 The idea behind this is that sometimes it is convenient (for the
4234 programmer) to consider floating-point values as approximations to
4235 infinitely precise real numbers. If you are doing this, then you need
4236 to compute (by analyzing the code, or in some other way) the maximum or
4237 likely maximum error that the computation introduces, and allow for it
4238 when performing comparisons (and when producing output, but that's a
4239 different problem). In particular, instead of testing for equality, you
4240 should check to see whether the two values have ranges that overlap; and
4241 this is done with the relational operators, so equality comparisons are
4244 @item -Wtraditional @r{(C and Objective-C only)}
4245 @opindex Wtraditional
4246 @opindex Wno-traditional
4247 Warn about certain constructs that behave differently in traditional and
4248 ISO C@. Also warn about ISO C constructs that have no traditional C
4249 equivalent, and/or problematic constructs that should be avoided.
4253 Macro parameters that appear within string literals in the macro body.
4254 In traditional C macro replacement takes place within string literals,
4255 but in ISO C it does not.
4258 In traditional C, some preprocessor directives did not exist.
4259 Traditional preprocessors only considered a line to be a directive
4260 if the @samp{#} appeared in column 1 on the line. Therefore
4261 @option{-Wtraditional} warns about directives that traditional C
4262 understands but ignores because the @samp{#} does not appear as the
4263 first character on the line. It also suggests you hide directives like
4264 @samp{#pragma} not understood by traditional C by indenting them. Some
4265 traditional implementations do not recognize @samp{#elif}, so this option
4266 suggests avoiding it altogether.
4269 A function-like macro that appears without arguments.
4272 The unary plus operator.
4275 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4276 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4277 constants.) Note, these suffixes appear in macros defined in the system
4278 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4279 Use of these macros in user code might normally lead to spurious
4280 warnings, however GCC's integrated preprocessor has enough context to
4281 avoid warning in these cases.
4284 A function declared external in one block and then used after the end of
4288 A @code{switch} statement has an operand of type @code{long}.
4291 A non-@code{static} function declaration follows a @code{static} one.
4292 This construct is not accepted by some traditional C compilers.
4295 The ISO type of an integer constant has a different width or
4296 signedness from its traditional type. This warning is only issued if
4297 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4298 typically represent bit patterns, are not warned about.
4301 Usage of ISO string concatenation is detected.
4304 Initialization of automatic aggregates.
4307 Identifier conflicts with labels. Traditional C lacks a separate
4308 namespace for labels.
4311 Initialization of unions. If the initializer is zero, the warning is
4312 omitted. This is done under the assumption that the zero initializer in
4313 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4314 initializer warnings and relies on default initialization to zero in the
4318 Conversions by prototypes between fixed/floating-point values and vice
4319 versa. The absence of these prototypes when compiling with traditional
4320 C causes serious problems. This is a subset of the possible
4321 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4324 Use of ISO C style function definitions. This warning intentionally is
4325 @emph{not} issued for prototype declarations or variadic functions
4326 because these ISO C features appear in your code when using
4327 libiberty's traditional C compatibility macros, @code{PARAMS} and
4328 @code{VPARAMS}. This warning is also bypassed for nested functions
4329 because that feature is already a GCC extension and thus not relevant to
4330 traditional C compatibility.
4333 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4334 @opindex Wtraditional-conversion
4335 @opindex Wno-traditional-conversion
4336 Warn if a prototype causes a type conversion that is different from what
4337 would happen to the same argument in the absence of a prototype. This
4338 includes conversions of fixed point to floating and vice versa, and
4339 conversions changing the width or signedness of a fixed-point argument
4340 except when the same as the default promotion.
4342 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4343 @opindex Wdeclaration-after-statement
4344 @opindex Wno-declaration-after-statement
4345 Warn when a declaration is found after a statement in a block. This
4346 construct, known from C++, was introduced with ISO C99 and is by default
4347 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4348 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4353 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4355 @item -Wno-endif-labels
4356 @opindex Wno-endif-labels
4357 @opindex Wendif-labels
4358 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4363 Warn whenever a local variable or type declaration shadows another variable,
4364 parameter, type, or class member (in C++), or whenever a built-in function
4365 is shadowed. Note that in C++, the compiler warns if a local variable
4366 shadows an explicit typedef, but not if it shadows a struct/class/enum.
4368 @item -Wlarger-than=@var{len}
4369 @opindex Wlarger-than=@var{len}
4370 @opindex Wlarger-than-@var{len}
4371 Warn whenever an object of larger than @var{len} bytes is defined.
4373 @item -Wframe-larger-than=@var{len}
4374 @opindex Wframe-larger-than
4375 Warn if the size of a function frame is larger than @var{len} bytes.
4376 The computation done to determine the stack frame size is approximate
4377 and not conservative.
4378 The actual requirements may be somewhat greater than @var{len}
4379 even if you do not get a warning. In addition, any space allocated
4380 via @code{alloca}, variable-length arrays, or related constructs
4381 is not included by the compiler when determining
4382 whether or not to issue a warning.
4384 @item -Wno-free-nonheap-object
4385 @opindex Wno-free-nonheap-object
4386 @opindex Wfree-nonheap-object
4387 Do not warn when attempting to free an object that was not allocated
4390 @item -Wstack-usage=@var{len}
4391 @opindex Wstack-usage
4392 Warn if the stack usage of a function might be larger than @var{len} bytes.
4393 The computation done to determine the stack usage is conservative.
4394 Any space allocated via @code{alloca}, variable-length arrays, or related
4395 constructs is included by the compiler when determining whether or not to
4398 The message is in keeping with the output of @option{-fstack-usage}.
4402 If the stack usage is fully static but exceeds the specified amount, it's:
4405 warning: stack usage is 1120 bytes
4408 If the stack usage is (partly) dynamic but bounded, it's:
4411 warning: stack usage might be 1648 bytes
4414 If the stack usage is (partly) dynamic and not bounded, it's:
4417 warning: stack usage might be unbounded
4421 @item -Wunsafe-loop-optimizations
4422 @opindex Wunsafe-loop-optimizations
4423 @opindex Wno-unsafe-loop-optimizations
4424 Warn if the loop cannot be optimized because the compiler cannot
4425 assume anything on the bounds of the loop indices. With
4426 @option{-funsafe-loop-optimizations} warn if the compiler makes
4429 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4430 @opindex Wno-pedantic-ms-format
4431 @opindex Wpedantic-ms-format
4432 When used in combination with @option{-Wformat}
4433 and @option{-pedantic} without GNU extensions, this option
4434 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4435 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4436 which depend on the MS runtime.
4438 @item -Wpointer-arith
4439 @opindex Wpointer-arith
4440 @opindex Wno-pointer-arith
4441 Warn about anything that depends on the ``size of'' a function type or
4442 of @code{void}. GNU C assigns these types a size of 1, for
4443 convenience in calculations with @code{void *} pointers and pointers
4444 to functions. In C++, warn also when an arithmetic operation involves
4445 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4448 @opindex Wtype-limits
4449 @opindex Wno-type-limits
4450 Warn if a comparison is always true or always false due to the limited
4451 range of the data type, but do not warn for constant expressions. For
4452 example, warn if an unsigned variable is compared against zero with
4453 @samp{<} or @samp{>=}. This warning is also enabled by
4456 @item -Wbad-function-cast @r{(C and Objective-C only)}
4457 @opindex Wbad-function-cast
4458 @opindex Wno-bad-function-cast
4459 Warn whenever a function call is cast to a non-matching type.
4460 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4462 @item -Wc++-compat @r{(C and Objective-C only)}
4463 Warn about ISO C constructs that are outside of the common subset of
4464 ISO C and ISO C++, e.g.@: request for implicit conversion from
4465 @code{void *} to a pointer to non-@code{void} type.
4467 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4468 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4469 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4470 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4471 enabled by @option{-Wall}.
4475 @opindex Wno-cast-qual
4476 Warn whenever a pointer is cast so as to remove a type qualifier from
4477 the target type. For example, warn if a @code{const char *} is cast
4478 to an ordinary @code{char *}.
4480 Also warn when making a cast that introduces a type qualifier in an
4481 unsafe way. For example, casting @code{char **} to @code{const char **}
4482 is unsafe, as in this example:
4485 /* p is char ** value. */
4486 const char **q = (const char **) p;
4487 /* Assignment of readonly string to const char * is OK. */
4489 /* Now char** pointer points to read-only memory. */
4494 @opindex Wcast-align
4495 @opindex Wno-cast-align
4496 Warn whenever a pointer is cast such that the required alignment of the
4497 target is increased. For example, warn if a @code{char *} is cast to
4498 an @code{int *} on machines where integers can only be accessed at
4499 two- or four-byte boundaries.
4501 @item -Wwrite-strings
4502 @opindex Wwrite-strings
4503 @opindex Wno-write-strings
4504 When compiling C, give string constants the type @code{const
4505 char[@var{length}]} so that copying the address of one into a
4506 non-@code{const} @code{char *} pointer produces a warning. These
4507 warnings help you find at compile time code that can try to write
4508 into a string constant, but only if you have been very careful about
4509 using @code{const} in declarations and prototypes. Otherwise, it is
4510 just a nuisance. This is why we did not make @option{-Wall} request
4513 When compiling C++, warn about the deprecated conversion from string
4514 literals to @code{char *}. This warning is enabled by default for C++
4519 @opindex Wno-clobbered
4520 Warn for variables that might be changed by @samp{longjmp} or
4521 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4523 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4524 @opindex Wconditionally-supported
4525 @opindex Wno-conditionally-supported
4526 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4529 @opindex Wconversion
4530 @opindex Wno-conversion
4531 Warn for implicit conversions that may alter a value. This includes
4532 conversions between real and integer, like @code{abs (x)} when
4533 @code{x} is @code{double}; conversions between signed and unsigned,
4534 like @code{unsigned ui = -1}; and conversions to smaller types, like
4535 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4536 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4537 changed by the conversion like in @code{abs (2.0)}. Warnings about
4538 conversions between signed and unsigned integers can be disabled by
4539 using @option{-Wno-sign-conversion}.
4541 For C++, also warn for confusing overload resolution for user-defined
4542 conversions; and conversions that never use a type conversion
4543 operator: conversions to @code{void}, the same type, a base class or a
4544 reference to them. Warnings about conversions between signed and
4545 unsigned integers are disabled by default in C++ unless
4546 @option{-Wsign-conversion} is explicitly enabled.
4548 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4549 @opindex Wconversion-null
4550 @opindex Wno-conversion-null
4551 Do not warn for conversions between @code{NULL} and non-pointer
4552 types. @option{-Wconversion-null} is enabled by default.
4554 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4555 @opindex Wzero-as-null-pointer-constant
4556 @opindex Wno-zero-as-null-pointer-constant
4557 Warn when a literal '0' is used as null pointer constant. This can
4558 be useful to facilitate the conversion to @code{nullptr} in C++11.
4562 @opindex Wno-date-time
4563 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4564 are encountered as they might prevent bit-wise-identical reproducible
4567 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4568 @opindex Wdelete-incomplete
4569 @opindex Wno-delete-incomplete
4570 Warn when deleting a pointer to incomplete type, which may cause
4571 undefined behavior at runtime. This warning is enabled by default.
4573 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4574 @opindex Wuseless-cast
4575 @opindex Wno-useless-cast
4576 Warn when an expression is casted to its own type.
4579 @opindex Wempty-body
4580 @opindex Wno-empty-body
4581 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4582 while} statement. This warning is also enabled by @option{-Wextra}.
4584 @item -Wenum-compare
4585 @opindex Wenum-compare
4586 @opindex Wno-enum-compare
4587 Warn about a comparison between values of different enumerated types.
4588 In C++ enumeral mismatches in conditional expressions are also
4589 diagnosed and the warning is enabled by default. In C this warning is
4590 enabled by @option{-Wall}.
4592 @item -Wjump-misses-init @r{(C, Objective-C only)}
4593 @opindex Wjump-misses-init
4594 @opindex Wno-jump-misses-init
4595 Warn if a @code{goto} statement or a @code{switch} statement jumps
4596 forward across the initialization of a variable, or jumps backward to a
4597 label after the variable has been initialized. This only warns about
4598 variables that are initialized when they are declared. This warning is
4599 only supported for C and Objective-C; in C++ this sort of branch is an
4602 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4603 can be disabled with the @option{-Wno-jump-misses-init} option.
4605 @item -Wsign-compare
4606 @opindex Wsign-compare
4607 @opindex Wno-sign-compare
4608 @cindex warning for comparison of signed and unsigned values
4609 @cindex comparison of signed and unsigned values, warning
4610 @cindex signed and unsigned values, comparison warning
4611 Warn when a comparison between signed and unsigned values could produce
4612 an incorrect result when the signed value is converted to unsigned.
4613 This warning is also enabled by @option{-Wextra}; to get the other warnings
4614 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4616 @item -Wsign-conversion
4617 @opindex Wsign-conversion
4618 @opindex Wno-sign-conversion
4619 Warn for implicit conversions that may change the sign of an integer
4620 value, like assigning a signed integer expression to an unsigned
4621 integer variable. An explicit cast silences the warning. In C, this
4622 option is enabled also by @option{-Wconversion}.
4624 @item -Wfloat-conversion
4625 @opindex Wfloat-conversion
4626 @opindex Wno-float-conversion
4627 Warn for implicit conversions that reduce the precision of a real value.
4628 This includes conversions from real to integer, and from higher precision
4629 real to lower precision real values. This option is also enabled by
4630 @option{-Wconversion}.
4632 @item -Wsizeof-pointer-memaccess
4633 @opindex Wsizeof-pointer-memaccess
4634 @opindex Wno-sizeof-pointer-memaccess
4635 Warn for suspicious length parameters to certain string and memory built-in
4636 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4637 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4638 but a pointer, and suggests a possible fix, or about
4639 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4644 @opindex Wno-address
4645 Warn about suspicious uses of memory addresses. These include using
4646 the address of a function in a conditional expression, such as
4647 @code{void func(void); if (func)}, and comparisons against the memory
4648 address of a string literal, such as @code{if (x == "abc")}. Such
4649 uses typically indicate a programmer error: the address of a function
4650 always evaluates to true, so their use in a conditional usually
4651 indicate that the programmer forgot the parentheses in a function
4652 call; and comparisons against string literals result in unspecified
4653 behavior and are not portable in C, so they usually indicate that the
4654 programmer intended to use @code{strcmp}. This warning is enabled by
4658 @opindex Wlogical-op
4659 @opindex Wno-logical-op
4660 Warn about suspicious uses of logical operators in expressions.
4661 This includes using logical operators in contexts where a
4662 bit-wise operator is likely to be expected.
4664 @item -Waggregate-return
4665 @opindex Waggregate-return
4666 @opindex Wno-aggregate-return
4667 Warn if any functions that return structures or unions are defined or
4668 called. (In languages where you can return an array, this also elicits
4671 @item -Wno-aggressive-loop-optimizations
4672 @opindex Wno-aggressive-loop-optimizations
4673 @opindex Waggressive-loop-optimizations
4674 Warn if in a loop with constant number of iterations the compiler detects
4675 undefined behavior in some statement during one or more of the iterations.
4677 @item -Wno-attributes
4678 @opindex Wno-attributes
4679 @opindex Wattributes
4680 Do not warn if an unexpected @code{__attribute__} is used, such as
4681 unrecognized attributes, function attributes applied to variables,
4682 etc. This does not stop errors for incorrect use of supported
4685 @item -Wno-builtin-macro-redefined
4686 @opindex Wno-builtin-macro-redefined
4687 @opindex Wbuiltin-macro-redefined
4688 Do not warn if certain built-in macros are redefined. This suppresses
4689 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4690 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4692 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4693 @opindex Wstrict-prototypes
4694 @opindex Wno-strict-prototypes
4695 Warn if a function is declared or defined without specifying the
4696 argument types. (An old-style function definition is permitted without
4697 a warning if preceded by a declaration that specifies the argument
4700 @item -Wold-style-declaration @r{(C and Objective-C only)}
4701 @opindex Wold-style-declaration
4702 @opindex Wno-old-style-declaration
4703 Warn for obsolescent usages, according to the C Standard, in a
4704 declaration. For example, warn if storage-class specifiers like
4705 @code{static} are not the first things in a declaration. This warning
4706 is also enabled by @option{-Wextra}.
4708 @item -Wold-style-definition @r{(C and Objective-C only)}
4709 @opindex Wold-style-definition
4710 @opindex Wno-old-style-definition
4711 Warn if an old-style function definition is used. A warning is given
4712 even if there is a previous prototype.
4714 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4715 @opindex Wmissing-parameter-type
4716 @opindex Wno-missing-parameter-type
4717 A function parameter is declared without a type specifier in K&R-style
4724 This warning is also enabled by @option{-Wextra}.
4726 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4727 @opindex Wmissing-prototypes
4728 @opindex Wno-missing-prototypes
4729 Warn if a global function is defined without a previous prototype
4730 declaration. This warning is issued even if the definition itself
4731 provides a prototype. Use this option to detect global functions
4732 that do not have a matching prototype declaration in a header file.
4733 This option is not valid for C++ because all function declarations
4734 provide prototypes and a non-matching declaration will declare an
4735 overload rather than conflict with an earlier declaration.
4736 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
4738 @item -Wmissing-declarations
4739 @opindex Wmissing-declarations
4740 @opindex Wno-missing-declarations
4741 Warn if a global function is defined without a previous declaration.
4742 Do so even if the definition itself provides a prototype.
4743 Use this option to detect global functions that are not declared in
4744 header files. In C, no warnings are issued for functions with previous
4745 non-prototype declarations; use @option{-Wmissing-prototype} to detect
4746 missing prototypes. In C++, no warnings are issued for function templates,
4747 or for inline functions, or for functions in anonymous namespaces.
4749 @item -Wmissing-field-initializers
4750 @opindex Wmissing-field-initializers
4751 @opindex Wno-missing-field-initializers
4755 Warn if a structure's initializer has some fields missing. For
4756 example, the following code causes such a warning, because
4757 @code{x.h} is implicitly zero:
4760 struct s @{ int f, g, h; @};
4761 struct s x = @{ 3, 4 @};
4764 This option does not warn about designated initializers, so the following
4765 modification does not trigger a warning:
4768 struct s @{ int f, g, h; @};
4769 struct s x = @{ .f = 3, .g = 4 @};
4772 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4773 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
4775 @item -Wno-multichar
4776 @opindex Wno-multichar
4778 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4779 Usually they indicate a typo in the user's code, as they have
4780 implementation-defined values, and should not be used in portable code.
4782 @item -Wnormalized=<none|id|nfc|nfkc>
4783 @opindex Wnormalized=
4786 @cindex character set, input normalization
4787 In ISO C and ISO C++, two identifiers are different if they are
4788 different sequences of characters. However, sometimes when characters
4789 outside the basic ASCII character set are used, you can have two
4790 different character sequences that look the same. To avoid confusion,
4791 the ISO 10646 standard sets out some @dfn{normalization rules} which
4792 when applied ensure that two sequences that look the same are turned into
4793 the same sequence. GCC can warn you if you are using identifiers that
4794 have not been normalized; this option controls that warning.
4796 There are four levels of warning supported by GCC@. The default is
4797 @option{-Wnormalized=nfc}, which warns about any identifier that is
4798 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4799 recommended form for most uses.
4801 Unfortunately, there are some characters allowed in identifiers by
4802 ISO C and ISO C++ that, when turned into NFC, are not allowed in
4803 identifiers. That is, there's no way to use these symbols in portable
4804 ISO C or C++ and have all your identifiers in NFC@.
4805 @option{-Wnormalized=id} suppresses the warning for these characters.
4806 It is hoped that future versions of the standards involved will correct
4807 this, which is why this option is not the default.
4809 You can switch the warning off for all characters by writing
4810 @option{-Wnormalized=none}. You should only do this if you
4811 are using some other normalization scheme (like ``D''), because
4812 otherwise you can easily create bugs that are literally impossible to see.
4814 Some characters in ISO 10646 have distinct meanings but look identical
4815 in some fonts or display methodologies, especially once formatting has
4816 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4817 LETTER N'', displays just like a regular @code{n} that has been
4818 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4819 normalization scheme to convert all these into a standard form as
4820 well, and GCC warns if your code is not in NFKC if you use
4821 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4822 about every identifier that contains the letter O because it might be
4823 confused with the digit 0, and so is not the default, but may be
4824 useful as a local coding convention if the programming environment
4825 cannot be fixed to display these characters distinctly.
4827 @item -Wno-deprecated
4828 @opindex Wno-deprecated
4829 @opindex Wdeprecated
4830 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4832 @item -Wno-deprecated-declarations
4833 @opindex Wno-deprecated-declarations
4834 @opindex Wdeprecated-declarations
4835 Do not warn about uses of functions (@pxref{Function Attributes}),
4836 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4837 Attributes}) marked as deprecated by using the @code{deprecated}
4841 @opindex Wno-overflow
4843 Do not warn about compile-time overflow in constant expressions.
4846 @opindex Wopenm-simd
4847 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
4848 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
4849 be used to relax the cost model.
4851 @item -Woverride-init @r{(C and Objective-C only)}
4852 @opindex Woverride-init
4853 @opindex Wno-override-init
4857 Warn if an initialized field without side effects is overridden when
4858 using designated initializers (@pxref{Designated Inits, , Designated
4861 This warning is included in @option{-Wextra}. To get other
4862 @option{-Wextra} warnings without this one, use @option{-Wextra
4863 -Wno-override-init}.
4868 Warn if a structure is given the packed attribute, but the packed
4869 attribute has no effect on the layout or size of the structure.
4870 Such structures may be mis-aligned for little benefit. For
4871 instance, in this code, the variable @code{f.x} in @code{struct bar}
4872 is misaligned even though @code{struct bar} does not itself
4873 have the packed attribute:
4880 @} __attribute__((packed));
4888 @item -Wpacked-bitfield-compat
4889 @opindex Wpacked-bitfield-compat
4890 @opindex Wno-packed-bitfield-compat
4891 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4892 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4893 the change can lead to differences in the structure layout. GCC
4894 informs you when the offset of such a field has changed in GCC 4.4.
4895 For example there is no longer a 4-bit padding between field @code{a}
4896 and @code{b} in this structure:
4903 @} __attribute__ ((packed));
4906 This warning is enabled by default. Use
4907 @option{-Wno-packed-bitfield-compat} to disable this warning.
4912 Warn if padding is included in a structure, either to align an element
4913 of the structure or to align the whole structure. Sometimes when this
4914 happens it is possible to rearrange the fields of the structure to
4915 reduce the padding and so make the structure smaller.
4917 @item -Wredundant-decls
4918 @opindex Wredundant-decls
4919 @opindex Wno-redundant-decls
4920 Warn if anything is declared more than once in the same scope, even in
4921 cases where multiple declaration is valid and changes nothing.
4923 @item -Wnested-externs @r{(C and Objective-C only)}
4924 @opindex Wnested-externs
4925 @opindex Wno-nested-externs
4926 Warn if an @code{extern} declaration is encountered within a function.
4928 @item -Wno-inherited-variadic-ctor
4929 @opindex Winherited-variadic-ctor
4930 @opindex Wno-inherited-variadic-ctor
4931 Suppress warnings about use of C++11 inheriting constructors when the
4932 base class inherited from has a C variadic constructor; the warning is
4933 on by default because the ellipsis is not inherited.
4938 Warn if a function that is declared as inline cannot be inlined.
4939 Even with this option, the compiler does not warn about failures to
4940 inline functions declared in system headers.
4942 The compiler uses a variety of heuristics to determine whether or not
4943 to inline a function. For example, the compiler takes into account
4944 the size of the function being inlined and the amount of inlining
4945 that has already been done in the current function. Therefore,
4946 seemingly insignificant changes in the source program can cause the
4947 warnings produced by @option{-Winline} to appear or disappear.
4949 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4950 @opindex Wno-invalid-offsetof
4951 @opindex Winvalid-offsetof
4952 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4953 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4954 to a non-POD type is undefined. In existing C++ implementations,
4955 however, @samp{offsetof} typically gives meaningful results even when
4956 applied to certain kinds of non-POD types (such as a simple
4957 @samp{struct} that fails to be a POD type only by virtue of having a
4958 constructor). This flag is for users who are aware that they are
4959 writing nonportable code and who have deliberately chosen to ignore the
4962 The restrictions on @samp{offsetof} may be relaxed in a future version
4963 of the C++ standard.
4965 @item -Wno-int-to-pointer-cast
4966 @opindex Wno-int-to-pointer-cast
4967 @opindex Wint-to-pointer-cast
4968 Suppress warnings from casts to pointer type of an integer of a
4969 different size. In C++, casting to a pointer type of smaller size is
4970 an error. @option{Wint-to-pointer-cast} is enabled by default.
4973 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4974 @opindex Wno-pointer-to-int-cast
4975 @opindex Wpointer-to-int-cast
4976 Suppress warnings from casts from a pointer to an integer type of a
4980 @opindex Winvalid-pch
4981 @opindex Wno-invalid-pch
4982 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4983 the search path but can't be used.
4987 @opindex Wno-long-long
4988 Warn if @samp{long long} type is used. This is enabled by either
4989 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
4990 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4992 @item -Wvariadic-macros
4993 @opindex Wvariadic-macros
4994 @opindex Wno-variadic-macros
4995 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4996 alternate syntax when in pedantic ISO C99 mode. This is default.
4997 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
5001 @opindex Wno-varargs
5002 Warn upon questionable usage of the macros used to handle variable
5003 arguments like @samp{va_start}. This is default. To inhibit the
5004 warning messages, use @option{-Wno-varargs}.
5006 @item -Wvector-operation-performance
5007 @opindex Wvector-operation-performance
5008 @opindex Wno-vector-operation-performance
5009 Warn if vector operation is not implemented via SIMD capabilities of the
5010 architecture. Mainly useful for the performance tuning.
5011 Vector operation can be implemented @code{piecewise}, which means that the
5012 scalar operation is performed on every vector element;
5013 @code{in parallel}, which means that the vector operation is implemented
5014 using scalars of wider type, which normally is more performance efficient;
5015 and @code{as a single scalar}, which means that vector fits into a
5018 @item -Wno-virtual-move-assign
5019 @opindex Wvirtual-move-assign
5020 @opindex Wno-virtual-move-assign
5021 Suppress warnings about inheriting from a virtual base with a
5022 non-trivial C++11 move assignment operator. This is dangerous because
5023 if the virtual base is reachable along more than one path, it will be
5024 moved multiple times, which can mean both objects end up in the
5025 moved-from state. If the move assignment operator is written to avoid
5026 moving from a moved-from object, this warning can be disabled.
5031 Warn if variable length array is used in the code.
5032 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5033 the variable length array.
5035 @item -Wvolatile-register-var
5036 @opindex Wvolatile-register-var
5037 @opindex Wno-volatile-register-var
5038 Warn if a register variable is declared volatile. The volatile
5039 modifier does not inhibit all optimizations that may eliminate reads
5040 and/or writes to register variables. This warning is enabled by
5043 @item -Wdisabled-optimization
5044 @opindex Wdisabled-optimization
5045 @opindex Wno-disabled-optimization
5046 Warn if a requested optimization pass is disabled. This warning does
5047 not generally indicate that there is anything wrong with your code; it
5048 merely indicates that GCC's optimizers are unable to handle the code
5049 effectively. Often, the problem is that your code is too big or too
5050 complex; GCC refuses to optimize programs when the optimization
5051 itself is likely to take inordinate amounts of time.
5053 @item -Wpointer-sign @r{(C and Objective-C only)}
5054 @opindex Wpointer-sign
5055 @opindex Wno-pointer-sign
5056 Warn for pointer argument passing or assignment with different signedness.
5057 This option is only supported for C and Objective-C@. It is implied by
5058 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5059 @option{-Wno-pointer-sign}.
5061 @item -Wstack-protector
5062 @opindex Wstack-protector
5063 @opindex Wno-stack-protector
5064 This option is only active when @option{-fstack-protector} is active. It
5065 warns about functions that are not protected against stack smashing.
5067 @item -Woverlength-strings
5068 @opindex Woverlength-strings
5069 @opindex Wno-overlength-strings
5070 Warn about string constants that are longer than the ``minimum
5071 maximum'' length specified in the C standard. Modern compilers
5072 generally allow string constants that are much longer than the
5073 standard's minimum limit, but very portable programs should avoid
5074 using longer strings.
5076 The limit applies @emph{after} string constant concatenation, and does
5077 not count the trailing NUL@. In C90, the limit was 509 characters; in
5078 C99, it was raised to 4095. C++98 does not specify a normative
5079 minimum maximum, so we do not diagnose overlength strings in C++@.
5081 This option is implied by @option{-Wpedantic}, and can be disabled with
5082 @option{-Wno-overlength-strings}.
5084 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5085 @opindex Wunsuffixed-float-constants
5087 Issue a warning for any floating constant that does not have
5088 a suffix. When used together with @option{-Wsystem-headers} it
5089 warns about such constants in system header files. This can be useful
5090 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5091 from the decimal floating-point extension to C99.
5094 @node Debugging Options
5095 @section Options for Debugging Your Program or GCC
5096 @cindex options, debugging
5097 @cindex debugging information options
5099 GCC has various special options that are used for debugging
5100 either your program or GCC:
5105 Produce debugging information in the operating system's native format
5106 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5109 On most systems that use stabs format, @option{-g} enables use of extra
5110 debugging information that only GDB can use; this extra information
5111 makes debugging work better in GDB but probably makes other debuggers
5113 refuse to read the program. If you want to control for certain whether
5114 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5115 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5117 GCC allows you to use @option{-g} with
5118 @option{-O}. The shortcuts taken by optimized code may occasionally
5119 produce surprising results: some variables you declared may not exist
5120 at all; flow of control may briefly move where you did not expect it;
5121 some statements may not be executed because they compute constant
5122 results or their values are already at hand; some statements may
5123 execute in different places because they have been moved out of loops.
5125 Nevertheless it proves possible to debug optimized output. This makes
5126 it reasonable to use the optimizer for programs that might have bugs.
5128 The following options are useful when GCC is generated with the
5129 capability for more than one debugging format.
5132 @opindex gsplit-dwarf
5133 Separate as much dwarf debugging information as possible into a
5134 separate output file with the extension .dwo. This option allows
5135 the build system to avoid linking files with debug information. To
5136 be useful, this option requires a debugger capable of reading .dwo
5141 Produce debugging information for use by GDB@. This means to use the
5142 most expressive format available (DWARF 2, stabs, or the native format
5143 if neither of those are supported), including GDB extensions if at all
5148 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5150 @item -ggnu-pubnames
5151 @opindex ggnu-pubnames
5152 Generate .debug_pubnames and .debug_pubtypes sections in a format
5153 suitable for conversion into a GDB@ index. This option is only useful
5154 with a linker that can produce GDB@ index version 7.
5158 Produce debugging information in stabs format (if that is supported),
5159 without GDB extensions. This is the format used by DBX on most BSD
5160 systems. On MIPS, Alpha and System V Release 4 systems this option
5161 produces stabs debugging output that is not understood by DBX or SDB@.
5162 On System V Release 4 systems this option requires the GNU assembler.
5164 @item -feliminate-unused-debug-symbols
5165 @opindex feliminate-unused-debug-symbols
5166 Produce debugging information in stabs format (if that is supported),
5167 for only symbols that are actually used.
5169 @item -femit-class-debug-always
5170 Instead of emitting debugging information for a C++ class in only one
5171 object file, emit it in all object files using the class. This option
5172 should be used only with debuggers that are unable to handle the way GCC
5173 normally emits debugging information for classes because using this
5174 option increases the size of debugging information by as much as a
5177 @item -fdebug-types-section
5178 @opindex fdebug-types-section
5179 @opindex fno-debug-types-section
5180 When using DWARF Version 4 or higher, type DIEs can be put into
5181 their own @code{.debug_types} section instead of making them part of the
5182 @code{.debug_info} section. It is more efficient to put them in a separate
5183 comdat sections since the linker can then remove duplicates.
5184 But not all DWARF consumers support @code{.debug_types} sections yet
5185 and on some objects @code{.debug_types} produces larger instead of smaller
5186 debugging information.
5190 Produce debugging information in stabs format (if that is supported),
5191 using GNU extensions understood only by the GNU debugger (GDB)@. The
5192 use of these extensions is likely to make other debuggers crash or
5193 refuse to read the program.
5197 Produce debugging information in COFF format (if that is supported).
5198 This is the format used by SDB on most System V systems prior to
5203 Produce debugging information in XCOFF format (if that is supported).
5204 This is the format used by the DBX debugger on IBM RS/6000 systems.
5208 Produce debugging information in XCOFF format (if that is supported),
5209 using GNU extensions understood only by the GNU debugger (GDB)@. The
5210 use of these extensions is likely to make other debuggers crash or
5211 refuse to read the program, and may cause assemblers other than the GNU
5212 assembler (GAS) to fail with an error.
5214 @item -gdwarf-@var{version}
5215 @opindex gdwarf-@var{version}
5216 Produce debugging information in DWARF format (if that is supported).
5217 The value of @var{version} may be either 2, 3 or 4; the default version
5218 for most targets is 4.
5220 Note that with DWARF Version 2, some ports require and always
5221 use some non-conflicting DWARF 3 extensions in the unwind tables.
5223 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5224 for maximum benefit.
5226 @item -grecord-gcc-switches
5227 @opindex grecord-gcc-switches
5228 This switch causes the command-line options used to invoke the
5229 compiler that may affect code generation to be appended to the
5230 DW_AT_producer attribute in DWARF debugging information. The options
5231 are concatenated with spaces separating them from each other and from
5232 the compiler version. See also @option{-frecord-gcc-switches} for another
5233 way of storing compiler options into the object file. This is the default.
5235 @item -gno-record-gcc-switches
5236 @opindex gno-record-gcc-switches
5237 Disallow appending command-line options to the DW_AT_producer attribute
5238 in DWARF debugging information.
5240 @item -gstrict-dwarf
5241 @opindex gstrict-dwarf
5242 Disallow using extensions of later DWARF standard version than selected
5243 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5244 DWARF extensions from later standard versions is allowed.
5246 @item -gno-strict-dwarf
5247 @opindex gno-strict-dwarf
5248 Allow using extensions of later DWARF standard version than selected with
5249 @option{-gdwarf-@var{version}}.
5253 Produce debugging information in Alpha/VMS debug format (if that is
5254 supported). This is the format used by DEBUG on Alpha/VMS systems.
5257 @itemx -ggdb@var{level}
5258 @itemx -gstabs@var{level}
5259 @itemx -gcoff@var{level}
5260 @itemx -gxcoff@var{level}
5261 @itemx -gvms@var{level}
5262 Request debugging information and also use @var{level} to specify how
5263 much information. The default level is 2.
5265 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5268 Level 1 produces minimal information, enough for making backtraces in
5269 parts of the program that you don't plan to debug. This includes
5270 descriptions of functions and external variables, and line number
5271 tables, but no information about local variables.
5273 Level 3 includes extra information, such as all the macro definitions
5274 present in the program. Some debuggers support macro expansion when
5275 you use @option{-g3}.
5277 @option{-gdwarf-2} does not accept a concatenated debug level, because
5278 GCC used to support an option @option{-gdwarf} that meant to generate
5279 debug information in version 1 of the DWARF format (which is very
5280 different from version 2), and it would have been too confusing. That
5281 debug format is long obsolete, but the option cannot be changed now.
5282 Instead use an additional @option{-g@var{level}} option to change the
5283 debug level for DWARF.
5287 Turn off generation of debug info, if leaving out this option
5288 generates it, or turn it on at level 2 otherwise. The position of this
5289 argument in the command line does not matter; it takes effect after all
5290 other options are processed, and it does so only once, no matter how
5291 many times it is given. This is mainly intended to be used with
5292 @option{-fcompare-debug}.
5294 @item -fsanitize=address
5295 @opindex fsanitize=address
5296 Enable AddressSanitizer, a fast memory error detector.
5297 Memory access instructions will be instrumented to detect
5298 out-of-bounds and use-after-free bugs.
5299 See @uref{http://code.google.com/p/address-sanitizer/} for
5300 more details. The run-time behavior can be influenced using the
5301 @env{ASAN_OPTIONS} environment variable; see
5302 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5303 a list of supported options.
5305 @item -fsanitize=thread
5306 @opindex fsanitize=thread
5307 Enable ThreadSanitizer, a fast data race detector.
5308 Memory access instructions will be instrumented to detect
5309 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5310 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5311 environment variable; see
5312 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5315 @item -fsanitize=leak
5316 @opindex fsanitize=leak
5317 Enable LeakSanitizer, a memory leak detector.
5318 This option only matters for linking of executables and if neither
5319 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5320 case it will link the executable against a library that overrides @code{malloc}
5321 and other allocator functions. See
5322 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5323 details. The run-time behavior can be influenced using the
5324 @env{LSAN_OPTIONS} environment variable.
5326 @item -fsanitize=undefined
5327 @opindex fsanitize=undefined
5328 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5329 Various computations will be instrumented to detect undefined behavior
5330 at runtime. Current suboptions are:
5334 @item -fsanitize=shift
5335 @opindex fsanitize=shift
5337 This option enables checking that the result of a shift operation is
5338 not undefined. Note that what exactly is considered undefined differs
5339 slightly between C and C++, as well as between ISO C90 and C99, etc.
5341 @item -fsanitize=integer-divide-by-zero
5342 @opindex fsanitize=integer-divide-by-zero
5344 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5346 @item -fsanitize=unreachable
5347 @opindex fsanitize=unreachable
5349 With this option, the compiler will turn the @code{__builtin_unreachable}
5350 call into a diagnostics message call instead. When reaching the
5351 @code{__builtin_unreachable} call, the behavior is undefined.
5353 @item -fsanitize=vla-bound
5354 @opindex fsanitize=vla-bound
5356 This option instructs the compiler to check that the size of a variable
5357 length array is positive. This option does not have any effect in
5358 @option{-std=c++1y} mode, as the standard requires the exception be thrown
5361 @item -fsanitize=null
5362 @opindex fsanitize=null
5364 This option enables pointer checking. Particularly, the application
5365 built with this option turned on will issue an error message when it
5366 tries to dereference a NULL pointer, or if a reference (possibly an
5367 rvalue reference) is bound to a NULL pointer.
5369 @item -fsanitize=return
5370 @opindex fsanitize=return
5372 This option enables return statement checking. Programs
5373 built with this option turned on will issue an error message
5374 when the end of a non-void function is reached without actually
5375 returning a value. This option works in C++ only.
5377 @item -fsanitize=signed-integer-overflow
5378 @opindex fsanitize=signed-integer-overflow
5380 This option enables signed integer overflow checking. We check that
5381 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5382 does not overflow in the signed arithmetics. Note, integer promotion
5383 rules must be taken into account. That is, the following is not an
5386 signed char a = SCHAR_MAX;
5392 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5393 @option{-fsanitize=undefined} gives a diagnostic message.
5394 This currently works only for the C family of languages.
5396 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5397 @opindex fdump-final-insns
5398 Dump the final internal representation (RTL) to @var{file}. If the
5399 optional argument is omitted (or if @var{file} is @code{.}), the name
5400 of the dump file is determined by appending @code{.gkd} to the
5401 compilation output file name.
5403 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5404 @opindex fcompare-debug
5405 @opindex fno-compare-debug
5406 If no error occurs during compilation, run the compiler a second time,
5407 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5408 passed to the second compilation. Dump the final internal
5409 representation in both compilations, and print an error if they differ.
5411 If the equal sign is omitted, the default @option{-gtoggle} is used.
5413 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5414 and nonzero, implicitly enables @option{-fcompare-debug}. If
5415 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5416 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5419 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5420 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5421 of the final representation and the second compilation, preventing even
5422 @env{GCC_COMPARE_DEBUG} from taking effect.
5424 To verify full coverage during @option{-fcompare-debug} testing, set
5425 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
5426 which GCC rejects as an invalid option in any actual compilation
5427 (rather than preprocessing, assembly or linking). To get just a
5428 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5429 not overridden} will do.
5431 @item -fcompare-debug-second
5432 @opindex fcompare-debug-second
5433 This option is implicitly passed to the compiler for the second
5434 compilation requested by @option{-fcompare-debug}, along with options to
5435 silence warnings, and omitting other options that would cause
5436 side-effect compiler outputs to files or to the standard output. Dump
5437 files and preserved temporary files are renamed so as to contain the
5438 @code{.gk} additional extension during the second compilation, to avoid
5439 overwriting those generated by the first.
5441 When this option is passed to the compiler driver, it causes the
5442 @emph{first} compilation to be skipped, which makes it useful for little
5443 other than debugging the compiler proper.
5445 @item -feliminate-dwarf2-dups
5446 @opindex feliminate-dwarf2-dups
5447 Compress DWARF 2 debugging information by eliminating duplicated
5448 information about each symbol. This option only makes sense when
5449 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5451 @item -femit-struct-debug-baseonly
5452 @opindex femit-struct-debug-baseonly
5453 Emit debug information for struct-like types
5454 only when the base name of the compilation source file
5455 matches the base name of file in which the struct is defined.
5457 This option substantially reduces the size of debugging information,
5458 but at significant potential loss in type information to the debugger.
5459 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5460 See @option{-femit-struct-debug-detailed} for more detailed control.
5462 This option works only with DWARF 2.
5464 @item -femit-struct-debug-reduced
5465 @opindex femit-struct-debug-reduced
5466 Emit debug information for struct-like types
5467 only when the base name of the compilation source file
5468 matches the base name of file in which the type is defined,
5469 unless the struct is a template or defined in a system header.
5471 This option significantly reduces the size of debugging information,
5472 with some potential loss in type information to the debugger.
5473 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5474 See @option{-femit-struct-debug-detailed} for more detailed control.
5476 This option works only with DWARF 2.
5478 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5479 Specify the struct-like types
5480 for which the compiler generates debug information.
5481 The intent is to reduce duplicate struct debug information
5482 between different object files within the same program.
5484 This option is a detailed version of
5485 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5486 which serves for most needs.
5488 A specification has the syntax@*
5489 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5491 The optional first word limits the specification to
5492 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5493 A struct type is used directly when it is the type of a variable, member.
5494 Indirect uses arise through pointers to structs.
5495 That is, when use of an incomplete struct is valid, the use is indirect.
5497 @samp{struct one direct; struct two * indirect;}.
5499 The optional second word limits the specification to
5500 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5501 Generic structs are a bit complicated to explain.
5502 For C++, these are non-explicit specializations of template classes,
5503 or non-template classes within the above.
5504 Other programming languages have generics,
5505 but @option{-femit-struct-debug-detailed} does not yet implement them.
5507 The third word specifies the source files for those
5508 structs for which the compiler should emit debug information.
5509 The values @samp{none} and @samp{any} have the normal meaning.
5510 The value @samp{base} means that
5511 the base of name of the file in which the type declaration appears
5512 must match the base of the name of the main compilation file.
5513 In practice, this means that when compiling @file{foo.c}, debug information
5514 is generated for types declared in that file and @file{foo.h},
5515 but not other header files.
5516 The value @samp{sys} means those types satisfying @samp{base}
5517 or declared in system or compiler headers.
5519 You may need to experiment to determine the best settings for your application.
5521 The default is @option{-femit-struct-debug-detailed=all}.
5523 This option works only with DWARF 2.
5525 @item -fno-merge-debug-strings
5526 @opindex fmerge-debug-strings
5527 @opindex fno-merge-debug-strings
5528 Direct the linker to not merge together strings in the debugging
5529 information that are identical in different object files. Merging is
5530 not supported by all assemblers or linkers. Merging decreases the size
5531 of the debug information in the output file at the cost of increasing
5532 link processing time. Merging is enabled by default.
5534 @item -fdebug-prefix-map=@var{old}=@var{new}
5535 @opindex fdebug-prefix-map
5536 When compiling files in directory @file{@var{old}}, record debugging
5537 information describing them as in @file{@var{new}} instead.
5539 @item -fno-dwarf2-cfi-asm
5540 @opindex fdwarf2-cfi-asm
5541 @opindex fno-dwarf2-cfi-asm
5542 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5543 instead of using GAS @code{.cfi_*} directives.
5545 @cindex @command{prof}
5548 Generate extra code to write profile information suitable for the
5549 analysis program @command{prof}. You must use this option when compiling
5550 the source files you want data about, and you must also use it when
5553 @cindex @command{gprof}
5556 Generate extra code to write profile information suitable for the
5557 analysis program @command{gprof}. You must use this option when compiling
5558 the source files you want data about, and you must also use it when
5563 Makes the compiler print out each function name as it is compiled, and
5564 print some statistics about each pass when it finishes.
5567 @opindex ftime-report
5568 Makes the compiler print some statistics about the time consumed by each
5569 pass when it finishes.
5572 @opindex fmem-report
5573 Makes the compiler print some statistics about permanent memory
5574 allocation when it finishes.
5576 @item -fmem-report-wpa
5577 @opindex fmem-report-wpa
5578 Makes the compiler print some statistics about permanent memory
5579 allocation for the WPA phase only.
5581 @item -fpre-ipa-mem-report
5582 @opindex fpre-ipa-mem-report
5583 @item -fpost-ipa-mem-report
5584 @opindex fpost-ipa-mem-report
5585 Makes the compiler print some statistics about permanent memory
5586 allocation before or after interprocedural optimization.
5588 @item -fprofile-report
5589 @opindex fprofile-report
5590 Makes the compiler print some statistics about consistency of the
5591 (estimated) profile and effect of individual passes.
5594 @opindex fstack-usage
5595 Makes the compiler output stack usage information for the program, on a
5596 per-function basis. The filename for the dump is made by appending
5597 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5598 the output file, if explicitly specified and it is not an executable,
5599 otherwise it is the basename of the source file. An entry is made up
5604 The name of the function.
5608 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5611 The qualifier @code{static} means that the function manipulates the stack
5612 statically: a fixed number of bytes are allocated for the frame on function
5613 entry and released on function exit; no stack adjustments are otherwise made
5614 in the function. The second field is this fixed number of bytes.
5616 The qualifier @code{dynamic} means that the function manipulates the stack
5617 dynamically: in addition to the static allocation described above, stack
5618 adjustments are made in the body of the function, for example to push/pop
5619 arguments around function calls. If the qualifier @code{bounded} is also
5620 present, the amount of these adjustments is bounded at compile time and
5621 the second field is an upper bound of the total amount of stack used by
5622 the function. If it is not present, the amount of these adjustments is
5623 not bounded at compile time and the second field only represents the
5626 @item -fprofile-arcs
5627 @opindex fprofile-arcs
5628 Add code so that program flow @dfn{arcs} are instrumented. During
5629 execution the program records how many times each branch and call is
5630 executed and how many times it is taken or returns. When the compiled
5631 program exits it saves this data to a file called
5632 @file{@var{auxname}.gcda} for each source file. The data may be used for
5633 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5634 test coverage analysis (@option{-ftest-coverage}). Each object file's
5635 @var{auxname} is generated from the name of the output file, if
5636 explicitly specified and it is not the final executable, otherwise it is
5637 the basename of the source file. In both cases any suffix is removed
5638 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5639 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5640 @xref{Cross-profiling}.
5642 @cindex @command{gcov}
5646 This option is used to compile and link code instrumented for coverage
5647 analysis. The option is a synonym for @option{-fprofile-arcs}
5648 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5649 linking). See the documentation for those options for more details.
5654 Compile the source files with @option{-fprofile-arcs} plus optimization
5655 and code generation options. For test coverage analysis, use the
5656 additional @option{-ftest-coverage} option. You do not need to profile
5657 every source file in a program.
5660 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5661 (the latter implies the former).
5664 Run the program on a representative workload to generate the arc profile
5665 information. This may be repeated any number of times. You can run
5666 concurrent instances of your program, and provided that the file system
5667 supports locking, the data files will be correctly updated. Also
5668 @code{fork} calls are detected and correctly handled (double counting
5672 For profile-directed optimizations, compile the source files again with
5673 the same optimization and code generation options plus
5674 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5675 Control Optimization}).
5678 For test coverage analysis, use @command{gcov} to produce human readable
5679 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5680 @command{gcov} documentation for further information.
5684 With @option{-fprofile-arcs}, for each function of your program GCC
5685 creates a program flow graph, then finds a spanning tree for the graph.
5686 Only arcs that are not on the spanning tree have to be instrumented: the
5687 compiler adds code to count the number of times that these arcs are
5688 executed. When an arc is the only exit or only entrance to a block, the
5689 instrumentation code can be added to the block; otherwise, a new basic
5690 block must be created to hold the instrumentation code.
5693 @item -ftest-coverage
5694 @opindex ftest-coverage
5695 Produce a notes file that the @command{gcov} code-coverage utility
5696 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5697 show program coverage. Each source file's note file is called
5698 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5699 above for a description of @var{auxname} and instructions on how to
5700 generate test coverage data. Coverage data matches the source files
5701 more closely if you do not optimize.
5703 @item -fdbg-cnt-list
5704 @opindex fdbg-cnt-list
5705 Print the name and the counter upper bound for all debug counters.
5708 @item -fdbg-cnt=@var{counter-value-list}
5710 Set the internal debug counter upper bound. @var{counter-value-list}
5711 is a comma-separated list of @var{name}:@var{value} pairs
5712 which sets the upper bound of each debug counter @var{name} to @var{value}.
5713 All debug counters have the initial upper bound of @code{UINT_MAX};
5714 thus @code{dbg_cnt()} returns true always unless the upper bound
5715 is set by this option.
5716 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
5717 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
5719 @item -fenable-@var{kind}-@var{pass}
5720 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5724 This is a set of options that are used to explicitly disable/enable
5725 optimization passes. These options are intended for use for debugging GCC.
5726 Compiler users should use regular options for enabling/disabling
5731 @item -fdisable-ipa-@var{pass}
5732 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
5733 statically invoked in the compiler multiple times, the pass name should be
5734 appended with a sequential number starting from 1.
5736 @item -fdisable-rtl-@var{pass}
5737 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
5738 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
5739 statically invoked in the compiler multiple times, the pass name should be
5740 appended with a sequential number starting from 1. @var{range-list} is a
5741 comma-separated list of function ranges or assembler names. Each range is a number
5742 pair separated by a colon. The range is inclusive in both ends. If the range
5743 is trivial, the number pair can be simplified as a single number. If the
5744 function's call graph node's @var{uid} falls within one of the specified ranges,
5745 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5746 function header of a dump file, and the pass names can be dumped by using
5747 option @option{-fdump-passes}.
5749 @item -fdisable-tree-@var{pass}
5750 @itemx -fdisable-tree-@var{pass}=@var{range-list}
5751 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5754 @item -fenable-ipa-@var{pass}
5755 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
5756 statically invoked in the compiler multiple times, the pass name should be
5757 appended with a sequential number starting from 1.
5759 @item -fenable-rtl-@var{pass}
5760 @itemx -fenable-rtl-@var{pass}=@var{range-list}
5761 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
5762 description and examples.
5764 @item -fenable-tree-@var{pass}
5765 @itemx -fenable-tree-@var{pass}=@var{range-list}
5766 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5767 of option arguments.
5771 Here are some examples showing uses of these options.
5775 # disable ccp1 for all functions
5777 # disable complete unroll for function whose cgraph node uid is 1
5778 -fenable-tree-cunroll=1
5779 # disable gcse2 for functions at the following ranges [1,1],
5780 # [300,400], and [400,1000]
5781 # disable gcse2 for functions foo and foo2
5782 -fdisable-rtl-gcse2=foo,foo2
5783 # disable early inlining
5784 -fdisable-tree-einline
5785 # disable ipa inlining
5786 -fdisable-ipa-inline
5787 # enable tree full unroll
5788 -fenable-tree-unroll
5792 @item -d@var{letters}
5793 @itemx -fdump-rtl-@var{pass}
5794 @itemx -fdump-rtl-@var{pass}=@var{filename}
5796 @opindex fdump-rtl-@var{pass}
5797 Says to make debugging dumps during compilation at times specified by
5798 @var{letters}. This is used for debugging the RTL-based passes of the
5799 compiler. The file names for most of the dumps are made by appending
5800 a pass number and a word to the @var{dumpname}, and the files are
5801 created in the directory of the output file. In case of
5802 @option{=@var{filename}} option, the dump is output on the given file
5803 instead of the pass numbered dump files. Note that the pass number is
5804 computed statically as passes get registered into the pass manager.
5805 Thus the numbering is not related to the dynamic order of execution of
5806 passes. In particular, a pass installed by a plugin could have a
5807 number over 200 even if it executed quite early. @var{dumpname} is
5808 generated from the name of the output file, if explicitly specified
5809 and it is not an executable, otherwise it is the basename of the
5810 source file. These switches may have different effects when
5811 @option{-E} is used for preprocessing.
5813 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5814 @option{-d} option @var{letters}. Here are the possible
5815 letters for use in @var{pass} and @var{letters}, and their meanings:
5819 @item -fdump-rtl-alignments
5820 @opindex fdump-rtl-alignments
5821 Dump after branch alignments have been computed.
5823 @item -fdump-rtl-asmcons
5824 @opindex fdump-rtl-asmcons
5825 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5827 @item -fdump-rtl-auto_inc_dec
5828 @opindex fdump-rtl-auto_inc_dec
5829 Dump after auto-inc-dec discovery. This pass is only run on
5830 architectures that have auto inc or auto dec instructions.
5832 @item -fdump-rtl-barriers
5833 @opindex fdump-rtl-barriers
5834 Dump after cleaning up the barrier instructions.
5836 @item -fdump-rtl-bbpart
5837 @opindex fdump-rtl-bbpart
5838 Dump after partitioning hot and cold basic blocks.
5840 @item -fdump-rtl-bbro
5841 @opindex fdump-rtl-bbro
5842 Dump after block reordering.
5844 @item -fdump-rtl-btl1
5845 @itemx -fdump-rtl-btl2
5846 @opindex fdump-rtl-btl2
5847 @opindex fdump-rtl-btl2
5848 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5849 after the two branch
5850 target load optimization passes.
5852 @item -fdump-rtl-bypass
5853 @opindex fdump-rtl-bypass
5854 Dump after jump bypassing and control flow optimizations.
5856 @item -fdump-rtl-combine
5857 @opindex fdump-rtl-combine
5858 Dump after the RTL instruction combination pass.
5860 @item -fdump-rtl-compgotos
5861 @opindex fdump-rtl-compgotos
5862 Dump after duplicating the computed gotos.
5864 @item -fdump-rtl-ce1
5865 @itemx -fdump-rtl-ce2
5866 @itemx -fdump-rtl-ce3
5867 @opindex fdump-rtl-ce1
5868 @opindex fdump-rtl-ce2
5869 @opindex fdump-rtl-ce3
5870 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5871 @option{-fdump-rtl-ce3} enable dumping after the three
5872 if conversion passes.
5874 @item -fdump-rtl-cprop_hardreg
5875 @opindex fdump-rtl-cprop_hardreg
5876 Dump after hard register copy propagation.
5878 @item -fdump-rtl-csa
5879 @opindex fdump-rtl-csa
5880 Dump after combining stack adjustments.
5882 @item -fdump-rtl-cse1
5883 @itemx -fdump-rtl-cse2
5884 @opindex fdump-rtl-cse1
5885 @opindex fdump-rtl-cse2
5886 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5887 the two common subexpression elimination passes.
5889 @item -fdump-rtl-dce
5890 @opindex fdump-rtl-dce
5891 Dump after the standalone dead code elimination passes.
5893 @item -fdump-rtl-dbr
5894 @opindex fdump-rtl-dbr
5895 Dump after delayed branch scheduling.
5897 @item -fdump-rtl-dce1
5898 @itemx -fdump-rtl-dce2
5899 @opindex fdump-rtl-dce1
5900 @opindex fdump-rtl-dce2
5901 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5902 the two dead store elimination passes.
5905 @opindex fdump-rtl-eh
5906 Dump after finalization of EH handling code.
5908 @item -fdump-rtl-eh_ranges
5909 @opindex fdump-rtl-eh_ranges
5910 Dump after conversion of EH handling range regions.
5912 @item -fdump-rtl-expand
5913 @opindex fdump-rtl-expand
5914 Dump after RTL generation.
5916 @item -fdump-rtl-fwprop1
5917 @itemx -fdump-rtl-fwprop2
5918 @opindex fdump-rtl-fwprop1
5919 @opindex fdump-rtl-fwprop2
5920 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5921 dumping after the two forward propagation passes.
5923 @item -fdump-rtl-gcse1
5924 @itemx -fdump-rtl-gcse2
5925 @opindex fdump-rtl-gcse1
5926 @opindex fdump-rtl-gcse2
5927 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5928 after global common subexpression elimination.
5930 @item -fdump-rtl-init-regs
5931 @opindex fdump-rtl-init-regs
5932 Dump after the initialization of the registers.
5934 @item -fdump-rtl-initvals
5935 @opindex fdump-rtl-initvals
5936 Dump after the computation of the initial value sets.
5938 @item -fdump-rtl-into_cfglayout
5939 @opindex fdump-rtl-into_cfglayout
5940 Dump after converting to cfglayout mode.
5942 @item -fdump-rtl-ira
5943 @opindex fdump-rtl-ira
5944 Dump after iterated register allocation.
5946 @item -fdump-rtl-jump
5947 @opindex fdump-rtl-jump
5948 Dump after the second jump optimization.
5950 @item -fdump-rtl-loop2
5951 @opindex fdump-rtl-loop2
5952 @option{-fdump-rtl-loop2} enables dumping after the rtl
5953 loop optimization passes.
5955 @item -fdump-rtl-mach
5956 @opindex fdump-rtl-mach
5957 Dump after performing the machine dependent reorganization pass, if that
5960 @item -fdump-rtl-mode_sw
5961 @opindex fdump-rtl-mode_sw
5962 Dump after removing redundant mode switches.
5964 @item -fdump-rtl-rnreg
5965 @opindex fdump-rtl-rnreg
5966 Dump after register renumbering.
5968 @item -fdump-rtl-outof_cfglayout
5969 @opindex fdump-rtl-outof_cfglayout
5970 Dump after converting from cfglayout mode.
5972 @item -fdump-rtl-peephole2
5973 @opindex fdump-rtl-peephole2
5974 Dump after the peephole pass.
5976 @item -fdump-rtl-postreload
5977 @opindex fdump-rtl-postreload
5978 Dump after post-reload optimizations.
5980 @item -fdump-rtl-pro_and_epilogue
5981 @opindex fdump-rtl-pro_and_epilogue
5982 Dump after generating the function prologues and epilogues.
5984 @item -fdump-rtl-sched1
5985 @itemx -fdump-rtl-sched2
5986 @opindex fdump-rtl-sched1
5987 @opindex fdump-rtl-sched2
5988 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5989 after the basic block scheduling passes.
5991 @item -fdump-rtl-ree
5992 @opindex fdump-rtl-ree
5993 Dump after sign/zero extension elimination.
5995 @item -fdump-rtl-seqabstr
5996 @opindex fdump-rtl-seqabstr
5997 Dump after common sequence discovery.
5999 @item -fdump-rtl-shorten
6000 @opindex fdump-rtl-shorten
6001 Dump after shortening branches.
6003 @item -fdump-rtl-sibling
6004 @opindex fdump-rtl-sibling
6005 Dump after sibling call optimizations.
6007 @item -fdump-rtl-split1
6008 @itemx -fdump-rtl-split2
6009 @itemx -fdump-rtl-split3
6010 @itemx -fdump-rtl-split4
6011 @itemx -fdump-rtl-split5
6012 @opindex fdump-rtl-split1
6013 @opindex fdump-rtl-split2
6014 @opindex fdump-rtl-split3
6015 @opindex fdump-rtl-split4
6016 @opindex fdump-rtl-split5
6017 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
6018 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
6019 @option{-fdump-rtl-split5} enable dumping after five rounds of
6020 instruction splitting.
6022 @item -fdump-rtl-sms
6023 @opindex fdump-rtl-sms
6024 Dump after modulo scheduling. This pass is only run on some
6027 @item -fdump-rtl-stack
6028 @opindex fdump-rtl-stack
6029 Dump after conversion from GCC's ``flat register file'' registers to the
6030 x87's stack-like registers. This pass is only run on x86 variants.
6032 @item -fdump-rtl-subreg1
6033 @itemx -fdump-rtl-subreg2
6034 @opindex fdump-rtl-subreg1
6035 @opindex fdump-rtl-subreg2
6036 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6037 the two subreg expansion passes.
6039 @item -fdump-rtl-unshare
6040 @opindex fdump-rtl-unshare
6041 Dump after all rtl has been unshared.
6043 @item -fdump-rtl-vartrack
6044 @opindex fdump-rtl-vartrack
6045 Dump after variable tracking.
6047 @item -fdump-rtl-vregs
6048 @opindex fdump-rtl-vregs
6049 Dump after converting virtual registers to hard registers.
6051 @item -fdump-rtl-web
6052 @opindex fdump-rtl-web
6053 Dump after live range splitting.
6055 @item -fdump-rtl-regclass
6056 @itemx -fdump-rtl-subregs_of_mode_init
6057 @itemx -fdump-rtl-subregs_of_mode_finish
6058 @itemx -fdump-rtl-dfinit
6059 @itemx -fdump-rtl-dfinish
6060 @opindex fdump-rtl-regclass
6061 @opindex fdump-rtl-subregs_of_mode_init
6062 @opindex fdump-rtl-subregs_of_mode_finish
6063 @opindex fdump-rtl-dfinit
6064 @opindex fdump-rtl-dfinish
6065 These dumps are defined but always produce empty files.
6068 @itemx -fdump-rtl-all
6070 @opindex fdump-rtl-all
6071 Produce all the dumps listed above.
6075 Annotate the assembler output with miscellaneous debugging information.
6079 Dump all macro definitions, at the end of preprocessing, in addition to
6084 Produce a core dump whenever an error occurs.
6088 Annotate the assembler output with a comment indicating which
6089 pattern and alternative is used. The length of each instruction is
6094 Dump the RTL in the assembler output as a comment before each instruction.
6095 Also turns on @option{-dp} annotation.
6099 Just generate RTL for a function instead of compiling it. Usually used
6100 with @option{-fdump-rtl-expand}.
6104 @opindex fdump-noaddr
6105 When doing debugging dumps, suppress address output. This makes it more
6106 feasible to use diff on debugging dumps for compiler invocations with
6107 different compiler binaries and/or different
6108 text / bss / data / heap / stack / dso start locations.
6110 @item -fdump-unnumbered
6111 @opindex fdump-unnumbered
6112 When doing debugging dumps, suppress instruction numbers and address output.
6113 This makes it more feasible to use diff on debugging dumps for compiler
6114 invocations with different options, in particular with and without
6117 @item -fdump-unnumbered-links
6118 @opindex fdump-unnumbered-links
6119 When doing debugging dumps (see @option{-d} option above), suppress
6120 instruction numbers for the links to the previous and next instructions
6123 @item -fdump-translation-unit @r{(C++ only)}
6124 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6125 @opindex fdump-translation-unit
6126 Dump a representation of the tree structure for the entire translation
6127 unit to a file. The file name is made by appending @file{.tu} to the
6128 source file name, and the file is created in the same directory as the
6129 output file. If the @samp{-@var{options}} form is used, @var{options}
6130 controls the details of the dump as described for the
6131 @option{-fdump-tree} options.
6133 @item -fdump-class-hierarchy @r{(C++ only)}
6134 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6135 @opindex fdump-class-hierarchy
6136 Dump a representation of each class's hierarchy and virtual function
6137 table layout to a file. The file name is made by appending
6138 @file{.class} to the source file name, and the file is created in the
6139 same directory as the output file. If the @samp{-@var{options}} form
6140 is used, @var{options} controls the details of the dump as described
6141 for the @option{-fdump-tree} options.
6143 @item -fdump-ipa-@var{switch}
6145 Control the dumping at various stages of inter-procedural analysis
6146 language tree to a file. The file name is generated by appending a
6147 switch specific suffix to the source file name, and the file is created
6148 in the same directory as the output file. The following dumps are
6153 Enables all inter-procedural analysis dumps.
6156 Dumps information about call-graph optimization, unused function removal,
6157 and inlining decisions.
6160 Dump after function inlining.
6165 @opindex fdump-passes
6166 Dump the list of optimization passes that are turned on and off by
6167 the current command-line options.
6169 @item -fdump-statistics-@var{option}
6170 @opindex fdump-statistics
6171 Enable and control dumping of pass statistics in a separate file. The
6172 file name is generated by appending a suffix ending in
6173 @samp{.statistics} to the source file name, and the file is created in
6174 the same directory as the output file. If the @samp{-@var{option}}
6175 form is used, @samp{-stats} causes counters to be summed over the
6176 whole compilation unit while @samp{-details} dumps every event as
6177 the passes generate them. The default with no option is to sum
6178 counters for each function compiled.
6180 @item -fdump-tree-@var{switch}
6181 @itemx -fdump-tree-@var{switch}-@var{options}
6182 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6184 Control the dumping at various stages of processing the intermediate
6185 language tree to a file. The file name is generated by appending a
6186 switch-specific suffix to the source file name, and the file is
6187 created in the same directory as the output file. In case of
6188 @option{=@var{filename}} option, the dump is output on the given file
6189 instead of the auto named dump files. If the @samp{-@var{options}}
6190 form is used, @var{options} is a list of @samp{-} separated options
6191 which control the details of the dump. Not all options are applicable
6192 to all dumps; those that are not meaningful are ignored. The
6193 following options are available
6197 Print the address of each node. Usually this is not meaningful as it
6198 changes according to the environment and source file. Its primary use
6199 is for tying up a dump file with a debug environment.
6201 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6202 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6203 use working backward from mangled names in the assembly file.
6205 When dumping front-end intermediate representations, inhibit dumping
6206 of members of a scope or body of a function merely because that scope
6207 has been reached. Only dump such items when they are directly reachable
6210 When dumping pretty-printed trees, this option inhibits dumping the
6211 bodies of control structures.
6213 When dumping RTL, print the RTL in slim (condensed) form instead of
6214 the default LISP-like representation.
6216 Print a raw representation of the tree. By default, trees are
6217 pretty-printed into a C-like representation.
6219 Enable more detailed dumps (not honored by every dump option). Also
6220 include information from the optimization passes.
6222 Enable dumping various statistics about the pass (not honored by every dump
6225 Enable showing basic block boundaries (disabled in raw dumps).
6227 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6228 dump a representation of the control flow graph suitable for viewing with
6229 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6230 the file is pretty-printed as a subgraph, so that GraphViz can render them
6231 all in a single plot.
6233 This option currently only works for RTL dumps, and the RTL is always
6234 dumped in slim form.
6236 Enable showing virtual operands for every statement.
6238 Enable showing line numbers for statements.
6240 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6242 Enable showing the tree dump for each statement.
6244 Enable showing the EH region number holding each statement.
6246 Enable showing scalar evolution analysis details.
6248 Enable showing optimization information (only available in certain
6251 Enable showing missed optimization information (only available in certain
6254 Enable other detailed optimization information (only available in
6256 @item =@var{filename}
6257 Instead of an auto named dump file, output into the given file
6258 name. The file names @file{stdout} and @file{stderr} are treated
6259 specially and are considered already open standard streams. For
6263 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6264 -fdump-tree-pre=stderr file.c
6267 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6268 output on to @file{stderr}. If two conflicting dump filenames are
6269 given for the same pass, then the latter option overrides the earlier
6273 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6274 and @option{lineno}.
6277 Turn on all optimization options, i.e., @option{optimized},
6278 @option{missed}, and @option{note}.
6281 The following tree dumps are possible:
6285 @opindex fdump-tree-original
6286 Dump before any tree based optimization, to @file{@var{file}.original}.
6289 @opindex fdump-tree-optimized
6290 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6293 @opindex fdump-tree-gimple
6294 Dump each function before and after the gimplification pass to a file. The
6295 file name is made by appending @file{.gimple} to the source file name.
6298 @opindex fdump-tree-cfg
6299 Dump the control flow graph of each function to a file. The file name is
6300 made by appending @file{.cfg} to the source file name.
6303 @opindex fdump-tree-ch
6304 Dump each function after copying loop headers. The file name is made by
6305 appending @file{.ch} to the source file name.
6308 @opindex fdump-tree-ssa
6309 Dump SSA related information to a file. The file name is made by appending
6310 @file{.ssa} to the source file name.
6313 @opindex fdump-tree-alias
6314 Dump aliasing information for each function. The file name is made by
6315 appending @file{.alias} to the source file name.
6318 @opindex fdump-tree-ccp
6319 Dump each function after CCP@. The file name is made by appending
6320 @file{.ccp} to the source file name.
6323 @opindex fdump-tree-storeccp
6324 Dump each function after STORE-CCP@. The file name is made by appending
6325 @file{.storeccp} to the source file name.
6328 @opindex fdump-tree-pre
6329 Dump trees after partial redundancy elimination. The file name is made
6330 by appending @file{.pre} to the source file name.
6333 @opindex fdump-tree-fre
6334 Dump trees after full redundancy elimination. The file name is made
6335 by appending @file{.fre} to the source file name.
6338 @opindex fdump-tree-copyprop
6339 Dump trees after copy propagation. The file name is made
6340 by appending @file{.copyprop} to the source file name.
6342 @item store_copyprop
6343 @opindex fdump-tree-store_copyprop
6344 Dump trees after store copy-propagation. The file name is made
6345 by appending @file{.store_copyprop} to the source file name.
6348 @opindex fdump-tree-dce
6349 Dump each function after dead code elimination. The file name is made by
6350 appending @file{.dce} to the source file name.
6353 @opindex fdump-tree-sra
6354 Dump each function after performing scalar replacement of aggregates. The
6355 file name is made by appending @file{.sra} to the source file name.
6358 @opindex fdump-tree-sink
6359 Dump each function after performing code sinking. The file name is made
6360 by appending @file{.sink} to the source file name.
6363 @opindex fdump-tree-dom
6364 Dump each function after applying dominator tree optimizations. The file
6365 name is made by appending @file{.dom} to the source file name.
6368 @opindex fdump-tree-dse
6369 Dump each function after applying dead store elimination. The file
6370 name is made by appending @file{.dse} to the source file name.
6373 @opindex fdump-tree-phiopt
6374 Dump each function after optimizing PHI nodes into straightline code. The file
6375 name is made by appending @file{.phiopt} to the source file name.
6378 @opindex fdump-tree-forwprop
6379 Dump each function after forward propagating single use variables. The file
6380 name is made by appending @file{.forwprop} to the source file name.
6383 @opindex fdump-tree-copyrename
6384 Dump each function after applying the copy rename optimization. The file
6385 name is made by appending @file{.copyrename} to the source file name.
6388 @opindex fdump-tree-nrv
6389 Dump each function after applying the named return value optimization on
6390 generic trees. The file name is made by appending @file{.nrv} to the source
6394 @opindex fdump-tree-vect
6395 Dump each function after applying vectorization of loops. The file name is
6396 made by appending @file{.vect} to the source file name.
6399 @opindex fdump-tree-slp
6400 Dump each function after applying vectorization of basic blocks. The file name
6401 is made by appending @file{.slp} to the source file name.
6404 @opindex fdump-tree-vrp
6405 Dump each function after Value Range Propagation (VRP). The file name
6406 is made by appending @file{.vrp} to the source file name.
6409 @opindex fdump-tree-all
6410 Enable all the available tree dumps with the flags provided in this option.
6414 @itemx -fopt-info-@var{options}
6415 @itemx -fopt-info-@var{options}=@var{filename}
6417 Controls optimization dumps from various optimization passes. If the
6418 @samp{-@var{options}} form is used, @var{options} is a list of
6419 @samp{-} separated options to select the dump details and
6420 optimizations. If @var{options} is not specified, it defaults to
6421 @option{optimized} for details and @option{optall} for optimization
6422 groups. If the @var{filename} is not specified, it defaults to
6423 @file{stderr}. Note that the output @var{filename} will be overwritten
6424 in case of multiple translation units. If a combined output from
6425 multiple translation units is desired, @file{stderr} should be used
6428 The options can be divided into two groups, 1) options describing the
6429 verbosity of the dump, and 2) options describing which optimizations
6430 should be included. The options from both the groups can be freely
6431 mixed as they are non-overlapping. However, in case of any conflicts,
6432 the latter options override the earlier options on the command
6433 line. Though multiple -fopt-info options are accepted, only one of
6434 them can have @option{=filename}. If other filenames are provided then
6435 all but the first one are ignored.
6437 The dump verbosity has the following options
6441 Print information when an optimization is successfully applied. It is
6442 up to a pass to decide which information is relevant. For example, the
6443 vectorizer passes print the source location of loops which got
6444 successfully vectorized.
6446 Print information about missed optimizations. Individual passes
6447 control which information to include in the output. For example,
6450 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
6453 will print information about missed optimization opportunities from
6454 vectorization passes on stderr.
6456 Print verbose information about optimizations, such as certain
6457 transformations, more detailed messages about decisions etc.
6459 Print detailed optimization information. This includes
6460 @var{optimized}, @var{missed}, and @var{note}.
6463 The second set of options describes a group of optimizations and may
6464 include one or more of the following.
6468 Enable dumps from all interprocedural optimizations.
6470 Enable dumps from all loop optimizations.
6472 Enable dumps from all inlining optimizations.
6474 Enable dumps from all vectorization optimizations.
6476 Enable dumps from all optimizations. This is a superset of
6477 the optimization groups listed above.
6482 gcc -O3 -fopt-info-missed=missed.all
6485 outputs missed optimization report from all the passes into
6490 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
6493 will output information about missed optimizations as well as
6494 optimized locations from all the inlining passes into
6497 If the @var{filename} is provided, then the dumps from all the
6498 applicable optimizations are concatenated into the @file{filename}.
6499 Otherwise the dump is output onto @file{stderr}. If @var{options} is
6500 omitted, it defaults to @option{all-optall}, which means dump all
6501 available optimization info from all the passes. In the following
6502 example, all optimization info is output on to @file{stderr}.
6508 Note that @option{-fopt-info-vec-missed} behaves the same as
6509 @option{-fopt-info-missed-vec}.
6511 As another example, consider
6514 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
6517 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
6518 in conflict since only one output file is allowed. In this case, only
6519 the first option takes effect and the subsequent options are
6520 ignored. Thus only the @file{vec.miss} is produced which contains
6521 dumps from the vectorizer about missed opportunities.
6523 @item -frandom-seed=@var{string}
6524 @opindex frandom-seed
6525 This option provides a seed that GCC uses in place of
6526 random numbers in generating certain symbol names
6527 that have to be different in every compiled file. It is also used to
6528 place unique stamps in coverage data files and the object files that
6529 produce them. You can use the @option{-frandom-seed} option to produce
6530 reproducibly identical object files.
6532 The @var{string} should be different for every file you compile.
6534 @item -fsched-verbose=@var{n}
6535 @opindex fsched-verbose
6536 On targets that use instruction scheduling, this option controls the
6537 amount of debugging output the scheduler prints. This information is
6538 written to standard error, unless @option{-fdump-rtl-sched1} or
6539 @option{-fdump-rtl-sched2} is specified, in which case it is output
6540 to the usual dump listing file, @file{.sched1} or @file{.sched2}
6541 respectively. However for @var{n} greater than nine, the output is
6542 always printed to standard error.
6544 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
6545 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
6546 For @var{n} greater than one, it also output basic block probabilities,
6547 detailed ready list information and unit/insn info. For @var{n} greater
6548 than two, it includes RTL at abort point, control-flow and regions info.
6549 And for @var{n} over four, @option{-fsched-verbose} also includes
6553 @itemx -save-temps=cwd
6555 Store the usual ``temporary'' intermediate files permanently; place them
6556 in the current directory and name them based on the source file. Thus,
6557 compiling @file{foo.c} with @option{-c -save-temps} produces files
6558 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
6559 preprocessed @file{foo.i} output file even though the compiler now
6560 normally uses an integrated preprocessor.
6562 When used in combination with the @option{-x} command-line option,
6563 @option{-save-temps} is sensible enough to avoid over writing an
6564 input source file with the same extension as an intermediate file.
6565 The corresponding intermediate file may be obtained by renaming the
6566 source file before using @option{-save-temps}.
6568 If you invoke GCC in parallel, compiling several different source
6569 files that share a common base name in different subdirectories or the
6570 same source file compiled for multiple output destinations, it is
6571 likely that the different parallel compilers will interfere with each
6572 other, and overwrite the temporary files. For instance:
6575 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
6576 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
6579 may result in @file{foo.i} and @file{foo.o} being written to
6580 simultaneously by both compilers.
6582 @item -save-temps=obj
6583 @opindex save-temps=obj
6584 Store the usual ``temporary'' intermediate files permanently. If the
6585 @option{-o} option is used, the temporary files are based on the
6586 object file. If the @option{-o} option is not used, the
6587 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
6592 gcc -save-temps=obj -c foo.c
6593 gcc -save-temps=obj -c bar.c -o dir/xbar.o
6594 gcc -save-temps=obj foobar.c -o dir2/yfoobar
6598 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
6599 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
6600 @file{dir2/yfoobar.o}.
6602 @item -time@r{[}=@var{file}@r{]}
6604 Report the CPU time taken by each subprocess in the compilation
6605 sequence. For C source files, this is the compiler proper and assembler
6606 (plus the linker if linking is done).
6608 Without the specification of an output file, the output looks like this:
6615 The first number on each line is the ``user time'', that is time spent
6616 executing the program itself. The second number is ``system time'',
6617 time spent executing operating system routines on behalf of the program.
6618 Both numbers are in seconds.
6620 With the specification of an output file, the output is appended to the
6621 named file, and it looks like this:
6624 0.12 0.01 cc1 @var{options}
6625 0.00 0.01 as @var{options}
6628 The ``user time'' and the ``system time'' are moved before the program
6629 name, and the options passed to the program are displayed, so that one
6630 can later tell what file was being compiled, and with which options.
6632 @item -fvar-tracking
6633 @opindex fvar-tracking
6634 Run variable tracking pass. It computes where variables are stored at each
6635 position in code. Better debugging information is then generated
6636 (if the debugging information format supports this information).
6638 It is enabled by default when compiling with optimization (@option{-Os},
6639 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6640 the debug info format supports it.
6642 @item -fvar-tracking-assignments
6643 @opindex fvar-tracking-assignments
6644 @opindex fno-var-tracking-assignments
6645 Annotate assignments to user variables early in the compilation and
6646 attempt to carry the annotations over throughout the compilation all the
6647 way to the end, in an attempt to improve debug information while
6648 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6650 It can be enabled even if var-tracking is disabled, in which case
6651 annotations are created and maintained, but discarded at the end.
6653 @item -fvar-tracking-assignments-toggle
6654 @opindex fvar-tracking-assignments-toggle
6655 @opindex fno-var-tracking-assignments-toggle
6656 Toggle @option{-fvar-tracking-assignments}, in the same way that
6657 @option{-gtoggle} toggles @option{-g}.
6659 @item -print-file-name=@var{library}
6660 @opindex print-file-name
6661 Print the full absolute name of the library file @var{library} that
6662 would be used when linking---and don't do anything else. With this
6663 option, GCC does not compile or link anything; it just prints the
6666 @item -print-multi-directory
6667 @opindex print-multi-directory
6668 Print the directory name corresponding to the multilib selected by any
6669 other switches present in the command line. This directory is supposed
6670 to exist in @env{GCC_EXEC_PREFIX}.
6672 @item -print-multi-lib
6673 @opindex print-multi-lib
6674 Print the mapping from multilib directory names to compiler switches
6675 that enable them. The directory name is separated from the switches by
6676 @samp{;}, and each switch starts with an @samp{@@} instead of the
6677 @samp{-}, without spaces between multiple switches. This is supposed to
6678 ease shell processing.
6680 @item -print-multi-os-directory
6681 @opindex print-multi-os-directory
6682 Print the path to OS libraries for the selected
6683 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6684 present in the @file{lib} subdirectory and no multilibs are used, this is
6685 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6686 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6687 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6688 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6690 @item -print-multiarch
6691 @opindex print-multiarch
6692 Print the path to OS libraries for the selected multiarch,
6693 relative to some @file{lib} subdirectory.
6695 @item -print-prog-name=@var{program}
6696 @opindex print-prog-name
6697 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6699 @item -print-libgcc-file-name
6700 @opindex print-libgcc-file-name
6701 Same as @option{-print-file-name=libgcc.a}.
6703 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6704 but you do want to link with @file{libgcc.a}. You can do:
6707 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6710 @item -print-search-dirs
6711 @opindex print-search-dirs
6712 Print the name of the configured installation directory and a list of
6713 program and library directories @command{gcc} searches---and don't do anything else.
6715 This is useful when @command{gcc} prints the error message
6716 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6717 To resolve this you either need to put @file{cpp0} and the other compiler
6718 components where @command{gcc} expects to find them, or you can set the environment
6719 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6720 Don't forget the trailing @samp{/}.
6721 @xref{Environment Variables}.
6723 @item -print-sysroot
6724 @opindex print-sysroot
6725 Print the target sysroot directory that is used during
6726 compilation. This is the target sysroot specified either at configure
6727 time or using the @option{--sysroot} option, possibly with an extra
6728 suffix that depends on compilation options. If no target sysroot is
6729 specified, the option prints nothing.
6731 @item -print-sysroot-headers-suffix
6732 @opindex print-sysroot-headers-suffix
6733 Print the suffix added to the target sysroot when searching for
6734 headers, or give an error if the compiler is not configured with such
6735 a suffix---and don't do anything else.
6738 @opindex dumpmachine
6739 Print the compiler's target machine (for example,
6740 @samp{i686-pc-linux-gnu})---and don't do anything else.
6743 @opindex dumpversion
6744 Print the compiler version (for example, @samp{3.0})---and don't do
6749 Print the compiler's built-in specs---and don't do anything else. (This
6750 is used when GCC itself is being built.) @xref{Spec Files}.
6752 @item -fno-eliminate-unused-debug-types
6753 @opindex feliminate-unused-debug-types
6754 @opindex fno-eliminate-unused-debug-types
6755 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
6756 output for types that are nowhere used in the source file being compiled.
6757 Sometimes it is useful to have GCC emit debugging
6758 information for all types declared in a compilation
6759 unit, regardless of whether or not they are actually used
6760 in that compilation unit, for example
6761 if, in the debugger, you want to cast a value to a type that is
6762 not actually used in your program (but is declared). More often,
6763 however, this results in a significant amount of wasted space.
6766 @node Optimize Options
6767 @section Options That Control Optimization
6768 @cindex optimize options
6769 @cindex options, optimization
6771 These options control various sorts of optimizations.
6773 Without any optimization option, the compiler's goal is to reduce the
6774 cost of compilation and to make debugging produce the expected
6775 results. Statements are independent: if you stop the program with a
6776 breakpoint between statements, you can then assign a new value to any
6777 variable or change the program counter to any other statement in the
6778 function and get exactly the results you expect from the source
6781 Turning on optimization flags makes the compiler attempt to improve
6782 the performance and/or code size at the expense of compilation time
6783 and possibly the ability to debug the program.
6785 The compiler performs optimization based on the knowledge it has of the
6786 program. Compiling multiple files at once to a single output file mode allows
6787 the compiler to use information gained from all of the files when compiling
6790 Not all optimizations are controlled directly by a flag. Only
6791 optimizations that have a flag are listed in this section.
6793 Most optimizations are only enabled if an @option{-O} level is set on
6794 the command line. Otherwise they are disabled, even if individual
6795 optimization flags are specified.
6797 Depending on the target and how GCC was configured, a slightly different
6798 set of optimizations may be enabled at each @option{-O} level than
6799 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6800 to find out the exact set of optimizations that are enabled at each level.
6801 @xref{Overall Options}, for examples.
6808 Optimize. Optimizing compilation takes somewhat more time, and a lot
6809 more memory for a large function.
6811 With @option{-O}, the compiler tries to reduce code size and execution
6812 time, without performing any optimizations that take a great deal of
6815 @option{-O} turns on the following optimization flags:
6819 -fcprop-registers @gol
6822 -fdelayed-branch @gol
6824 -fguess-branch-probability @gol
6825 -fif-conversion2 @gol
6826 -fif-conversion @gol
6827 -fipa-pure-const @gol
6829 -fipa-reference @gol
6831 -fsplit-wide-types @gol
6833 -ftree-builtin-call-dce @gol
6836 -ftree-copyrename @gol
6838 -ftree-dominator-opts @gol
6840 -ftree-forwprop @gol
6849 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6850 where doing so does not interfere with debugging.
6854 Optimize even more. GCC performs nearly all supported optimizations
6855 that do not involve a space-speed tradeoff.
6856 As compared to @option{-O}, this option increases both compilation time
6857 and the performance of the generated code.
6859 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6860 also turns on the following optimization flags:
6861 @gccoptlist{-fthread-jumps @gol
6862 -falign-functions -falign-jumps @gol
6863 -falign-loops -falign-labels @gol
6866 -fcse-follow-jumps -fcse-skip-blocks @gol
6867 -fdelete-null-pointer-checks @gol
6868 -fdevirtualize -fdevirtualize-speculatively @gol
6869 -fexpensive-optimizations @gol
6870 -fgcse -fgcse-lm @gol
6871 -fhoist-adjacent-loads @gol
6872 -finline-small-functions @gol
6873 -findirect-inlining @gol
6875 -fisolate-erroneous-paths-dereference @gol
6876 -foptimize-sibling-calls @gol
6877 -fpartial-inlining @gol
6879 -freorder-blocks -freorder-functions @gol
6880 -frerun-cse-after-loop @gol
6881 -fsched-interblock -fsched-spec @gol
6882 -fschedule-insns -fschedule-insns2 @gol
6883 -fstrict-aliasing -fstrict-overflow @gol
6884 -ftree-switch-conversion -ftree-tail-merge @gol
6888 Please note the warning under @option{-fgcse} about
6889 invoking @option{-O2} on programs that use computed gotos.
6893 Optimize yet more. @option{-O3} turns on all optimizations specified
6894 by @option{-O2} and also turns on the @option{-finline-functions},
6895 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6896 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
6897 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
6898 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
6902 Reduce compilation time and make debugging produce the expected
6903 results. This is the default.
6907 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6908 do not typically increase code size. It also performs further
6909 optimizations designed to reduce code size.
6911 @option{-Os} disables the following optimization flags:
6912 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6913 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6914 -fprefetch-loop-arrays}
6918 Disregard strict standards compliance. @option{-Ofast} enables all
6919 @option{-O3} optimizations. It also enables optimizations that are not
6920 valid for all standard-compliant programs.
6921 It turns on @option{-ffast-math} and the Fortran-specific
6922 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6926 Optimize debugging experience. @option{-Og} enables optimizations
6927 that do not interfere with debugging. It should be the optimization
6928 level of choice for the standard edit-compile-debug cycle, offering
6929 a reasonable level of optimization while maintaining fast compilation
6930 and a good debugging experience.
6932 If you use multiple @option{-O} options, with or without level numbers,
6933 the last such option is the one that is effective.
6936 Options of the form @option{-f@var{flag}} specify machine-independent
6937 flags. Most flags have both positive and negative forms; the negative
6938 form of @option{-ffoo} is @option{-fno-foo}. In the table
6939 below, only one of the forms is listed---the one you typically
6940 use. You can figure out the other form by either removing @samp{no-}
6943 The following options control specific optimizations. They are either
6944 activated by @option{-O} options or are related to ones that are. You
6945 can use the following flags in the rare cases when ``fine-tuning'' of
6946 optimizations to be performed is desired.
6949 @item -fno-defer-pop
6950 @opindex fno-defer-pop
6951 Always pop the arguments to each function call as soon as that function
6952 returns. For machines that must pop arguments after a function call,
6953 the compiler normally lets arguments accumulate on the stack for several
6954 function calls and pops them all at once.
6956 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6958 @item -fforward-propagate
6959 @opindex fforward-propagate
6960 Perform a forward propagation pass on RTL@. The pass tries to combine two
6961 instructions and checks if the result can be simplified. If loop unrolling
6962 is active, two passes are performed and the second is scheduled after
6965 This option is enabled by default at optimization levels @option{-O},
6966 @option{-O2}, @option{-O3}, @option{-Os}.
6968 @item -ffp-contract=@var{style}
6969 @opindex ffp-contract
6970 @option{-ffp-contract=off} disables floating-point expression contraction.
6971 @option{-ffp-contract=fast} enables floating-point expression contraction
6972 such as forming of fused multiply-add operations if the target has
6973 native support for them.
6974 @option{-ffp-contract=on} enables floating-point expression contraction
6975 if allowed by the language standard. This is currently not implemented
6976 and treated equal to @option{-ffp-contract=off}.
6978 The default is @option{-ffp-contract=fast}.
6980 @item -fomit-frame-pointer
6981 @opindex fomit-frame-pointer
6982 Don't keep the frame pointer in a register for functions that
6983 don't need one. This avoids the instructions to save, set up and
6984 restore frame pointers; it also makes an extra register available
6985 in many functions. @strong{It also makes debugging impossible on
6988 On some machines, such as the VAX, this flag has no effect, because
6989 the standard calling sequence automatically handles the frame pointer
6990 and nothing is saved by pretending it doesn't exist. The
6991 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6992 whether a target machine supports this flag. @xref{Registers,,Register
6993 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6995 Starting with GCC version 4.6, the default setting (when not optimizing for
6996 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
6997 @option{-fomit-frame-pointer}. The default can be reverted to
6998 @option{-fno-omit-frame-pointer} by configuring GCC with the
6999 @option{--enable-frame-pointer} configure option.
7001 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7003 @item -foptimize-sibling-calls
7004 @opindex foptimize-sibling-calls
7005 Optimize sibling and tail recursive calls.
7007 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7011 Do not expand any functions inline apart from those marked with
7012 the @code{always_inline} attribute. This is the default when not
7015 Single functions can be exempted from inlining by marking them
7016 with the @code{noinline} attribute.
7018 @item -finline-small-functions
7019 @opindex finline-small-functions
7020 Integrate functions into their callers when their body is smaller than expected
7021 function call code (so overall size of program gets smaller). The compiler
7022 heuristically decides which functions are simple enough to be worth integrating
7023 in this way. This inlining applies to all functions, even those not declared
7026 Enabled at level @option{-O2}.
7028 @item -findirect-inlining
7029 @opindex findirect-inlining
7030 Inline also indirect calls that are discovered to be known at compile
7031 time thanks to previous inlining. This option has any effect only
7032 when inlining itself is turned on by the @option{-finline-functions}
7033 or @option{-finline-small-functions} options.
7035 Enabled at level @option{-O2}.
7037 @item -finline-functions
7038 @opindex finline-functions
7039 Consider all functions for inlining, even if they are not declared inline.
7040 The compiler heuristically decides which functions are worth integrating
7043 If all calls to a given function are integrated, and the function is
7044 declared @code{static}, then the function is normally not output as
7045 assembler code in its own right.
7047 Enabled at level @option{-O3}.
7049 @item -finline-functions-called-once
7050 @opindex finline-functions-called-once
7051 Consider all @code{static} functions called once for inlining into their
7052 caller even if they are not marked @code{inline}. If a call to a given
7053 function is integrated, then the function is not output as assembler code
7056 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7058 @item -fearly-inlining
7059 @opindex fearly-inlining
7060 Inline functions marked by @code{always_inline} and functions whose body seems
7061 smaller than the function call overhead early before doing
7062 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7063 makes profiling significantly cheaper and usually inlining faster on programs
7064 having large chains of nested wrapper functions.
7070 Perform interprocedural scalar replacement of aggregates, removal of
7071 unused parameters and replacement of parameters passed by reference
7072 by parameters passed by value.
7074 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7076 @item -finline-limit=@var{n}
7077 @opindex finline-limit
7078 By default, GCC limits the size of functions that can be inlined. This flag
7079 allows coarse control of this limit. @var{n} is the size of functions that
7080 can be inlined in number of pseudo instructions.
7082 Inlining is actually controlled by a number of parameters, which may be
7083 specified individually by using @option{--param @var{name}=@var{value}}.
7084 The @option{-finline-limit=@var{n}} option sets some of these parameters
7088 @item max-inline-insns-single
7089 is set to @var{n}/2.
7090 @item max-inline-insns-auto
7091 is set to @var{n}/2.
7094 See below for a documentation of the individual
7095 parameters controlling inlining and for the defaults of these parameters.
7097 @emph{Note:} there may be no value to @option{-finline-limit} that results
7098 in default behavior.
7100 @emph{Note:} pseudo instruction represents, in this particular context, an
7101 abstract measurement of function's size. In no way does it represent a count
7102 of assembly instructions and as such its exact meaning might change from one
7103 release to an another.
7105 @item -fno-keep-inline-dllexport
7106 @opindex -fno-keep-inline-dllexport
7107 This is a more fine-grained version of @option{-fkeep-inline-functions},
7108 which applies only to functions that are declared using the @code{dllexport}
7109 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7112 @item -fkeep-inline-functions
7113 @opindex fkeep-inline-functions
7114 In C, emit @code{static} functions that are declared @code{inline}
7115 into the object file, even if the function has been inlined into all
7116 of its callers. This switch does not affect functions using the
7117 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7118 inline functions into the object file.
7120 @item -fkeep-static-consts
7121 @opindex fkeep-static-consts
7122 Emit variables declared @code{static const} when optimization isn't turned
7123 on, even if the variables aren't referenced.
7125 GCC enables this option by default. If you want to force the compiler to
7126 check if a variable is referenced, regardless of whether or not
7127 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7129 @item -fmerge-constants
7130 @opindex fmerge-constants
7131 Attempt to merge identical constants (string constants and floating-point
7132 constants) across compilation units.
7134 This option is the default for optimized compilation if the assembler and
7135 linker support it. Use @option{-fno-merge-constants} to inhibit this
7138 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7140 @item -fmerge-all-constants
7141 @opindex fmerge-all-constants
7142 Attempt to merge identical constants and identical variables.
7144 This option implies @option{-fmerge-constants}. In addition to
7145 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7146 arrays or initialized constant variables with integral or floating-point
7147 types. Languages like C or C++ require each variable, including multiple
7148 instances of the same variable in recursive calls, to have distinct locations,
7149 so using this option results in non-conforming
7152 @item -fmodulo-sched
7153 @opindex fmodulo-sched
7154 Perform swing modulo scheduling immediately before the first scheduling
7155 pass. This pass looks at innermost loops and reorders their
7156 instructions by overlapping different iterations.
7158 @item -fmodulo-sched-allow-regmoves
7159 @opindex fmodulo-sched-allow-regmoves
7160 Perform more aggressive SMS-based modulo scheduling with register moves
7161 allowed. By setting this flag certain anti-dependences edges are
7162 deleted, which triggers the generation of reg-moves based on the
7163 life-range analysis. This option is effective only with
7164 @option{-fmodulo-sched} enabled.
7166 @item -fno-branch-count-reg
7167 @opindex fno-branch-count-reg
7168 Do not use ``decrement and branch'' instructions on a count register,
7169 but instead generate a sequence of instructions that decrement a
7170 register, compare it against zero, then branch based upon the result.
7171 This option is only meaningful on architectures that support such
7172 instructions, which include x86, PowerPC, IA-64 and S/390.
7174 The default is @option{-fbranch-count-reg}.
7176 @item -fno-function-cse
7177 @opindex fno-function-cse
7178 Do not put function addresses in registers; make each instruction that
7179 calls a constant function contain the function's address explicitly.
7181 This option results in less efficient code, but some strange hacks
7182 that alter the assembler output may be confused by the optimizations
7183 performed when this option is not used.
7185 The default is @option{-ffunction-cse}
7187 @item -fno-zero-initialized-in-bss
7188 @opindex fno-zero-initialized-in-bss
7189 If the target supports a BSS section, GCC by default puts variables that
7190 are initialized to zero into BSS@. This can save space in the resulting
7193 This option turns off this behavior because some programs explicitly
7194 rely on variables going to the data section---e.g., so that the
7195 resulting executable can find the beginning of that section and/or make
7196 assumptions based on that.
7198 The default is @option{-fzero-initialized-in-bss}.
7200 @item -fthread-jumps
7201 @opindex fthread-jumps
7202 Perform optimizations that check to see if a jump branches to a
7203 location where another comparison subsumed by the first is found. If
7204 so, the first branch is redirected to either the destination of the
7205 second branch or a point immediately following it, depending on whether
7206 the condition is known to be true or false.
7208 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7210 @item -fsplit-wide-types
7211 @opindex fsplit-wide-types
7212 When using a type that occupies multiple registers, such as @code{long
7213 long} on a 32-bit system, split the registers apart and allocate them
7214 independently. This normally generates better code for those types,
7215 but may make debugging more difficult.
7217 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7220 @item -fcse-follow-jumps
7221 @opindex fcse-follow-jumps
7222 In common subexpression elimination (CSE), scan through jump instructions
7223 when the target of the jump is not reached by any other path. For
7224 example, when CSE encounters an @code{if} statement with an
7225 @code{else} clause, CSE follows the jump when the condition
7228 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7230 @item -fcse-skip-blocks
7231 @opindex fcse-skip-blocks
7232 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7233 follow jumps that conditionally skip over blocks. When CSE
7234 encounters a simple @code{if} statement with no else clause,
7235 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7236 body of the @code{if}.
7238 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7240 @item -frerun-cse-after-loop
7241 @opindex frerun-cse-after-loop
7242 Re-run common subexpression elimination after loop optimizations are
7245 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7249 Perform a global common subexpression elimination pass.
7250 This pass also performs global constant and copy propagation.
7252 @emph{Note:} When compiling a program using computed gotos, a GCC
7253 extension, you may get better run-time performance if you disable
7254 the global common subexpression elimination pass by adding
7255 @option{-fno-gcse} to the command line.
7257 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7261 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7262 attempts to move loads that are only killed by stores into themselves. This
7263 allows a loop containing a load/store sequence to be changed to a load outside
7264 the loop, and a copy/store within the loop.
7266 Enabled by default when @option{-fgcse} is enabled.
7270 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7271 global common subexpression elimination. This pass attempts to move
7272 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7273 loops containing a load/store sequence can be changed to a load before
7274 the loop and a store after the loop.
7276 Not enabled at any optimization level.
7280 When @option{-fgcse-las} is enabled, the global common subexpression
7281 elimination pass eliminates redundant loads that come after stores to the
7282 same memory location (both partial and full redundancies).
7284 Not enabled at any optimization level.
7286 @item -fgcse-after-reload
7287 @opindex fgcse-after-reload
7288 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7289 pass is performed after reload. The purpose of this pass is to clean up
7292 @item -faggressive-loop-optimizations
7293 @opindex faggressive-loop-optimizations
7294 This option tells the loop optimizer to use language constraints to
7295 derive bounds for the number of iterations of a loop. This assumes that
7296 loop code does not invoke undefined behavior by for example causing signed
7297 integer overflows or out-of-bound array accesses. The bounds for the
7298 number of iterations of a loop are used to guide loop unrolling and peeling
7299 and loop exit test optimizations.
7300 This option is enabled by default.
7302 @item -funsafe-loop-optimizations
7303 @opindex funsafe-loop-optimizations
7304 This option tells the loop optimizer to assume that loop indices do not
7305 overflow, and that loops with nontrivial exit condition are not
7306 infinite. This enables a wider range of loop optimizations even if
7307 the loop optimizer itself cannot prove that these assumptions are valid.
7308 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7309 if it finds this kind of loop.
7311 @item -fcrossjumping
7312 @opindex fcrossjumping
7313 Perform cross-jumping transformation.
7314 This transformation unifies equivalent code and saves code size. The
7315 resulting code may or may not perform better than without cross-jumping.
7317 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7319 @item -fauto-inc-dec
7320 @opindex fauto-inc-dec
7321 Combine increments or decrements of addresses with memory accesses.
7322 This pass is always skipped on architectures that do not have
7323 instructions to support this. Enabled by default at @option{-O} and
7324 higher on architectures that support this.
7328 Perform dead code elimination (DCE) on RTL@.
7329 Enabled by default at @option{-O} and higher.
7333 Perform dead store elimination (DSE) on RTL@.
7334 Enabled by default at @option{-O} and higher.
7336 @item -fif-conversion
7337 @opindex fif-conversion
7338 Attempt to transform conditional jumps into branch-less equivalents. This
7339 includes use of conditional moves, min, max, set flags and abs instructions, and
7340 some tricks doable by standard arithmetics. The use of conditional execution
7341 on chips where it is available is controlled by @code{if-conversion2}.
7343 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7345 @item -fif-conversion2
7346 @opindex fif-conversion2
7347 Use conditional execution (where available) to transform conditional jumps into
7348 branch-less equivalents.
7350 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7352 @item -fdeclone-ctor-dtor
7353 @opindex fdeclone-ctor-dtor
7354 The C++ ABI requires multiple entry points for constructors and
7355 destructors: one for a base subobject, one for a complete object, and
7356 one for a virtual destructor that calls operator delete afterwards.
7357 For a hierarchy with virtual bases, the base and complete variants are
7358 clones, which means two copies of the function. With this option, the
7359 base and complete variants are changed to be thunks that call a common
7362 Enabled by @option{-Os}.
7364 @item -fdelete-null-pointer-checks
7365 @opindex fdelete-null-pointer-checks
7366 Assume that programs cannot safely dereference null pointers, and that
7367 no code or data element resides there. This enables simple constant
7368 folding optimizations at all optimization levels. In addition, other
7369 optimization passes in GCC use this flag to control global dataflow
7370 analyses that eliminate useless checks for null pointers; these assume
7371 that if a pointer is checked after it has already been dereferenced,
7374 Note however that in some environments this assumption is not true.
7375 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7376 for programs that depend on that behavior.
7378 Some targets, especially embedded ones, disable this option at all levels.
7379 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7380 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7381 are enabled independently at different optimization levels.
7383 @item -fdevirtualize
7384 @opindex fdevirtualize
7385 Attempt to convert calls to virtual functions to direct calls. This
7386 is done both within a procedure and interprocedurally as part of
7387 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
7388 propagation (@option{-fipa-cp}).
7389 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7391 @item -fdevirtualize-speculatively
7392 @opindex fdevirtualize-speculatively
7393 Attempt to convert calls to virtual functions to speculative direct calls.
7394 Based on the analysis of the type inheritance graph, determine for a given call
7395 the set of likely targets. If the set is small, preferably of size 1, change
7396 the call into an conditional deciding on direct and indirect call. The
7397 speculative calls enable more optimizations, such as inlining. When they seem
7398 useless after further optimization, they are converted back into original form.
7400 @item -fexpensive-optimizations
7401 @opindex fexpensive-optimizations
7402 Perform a number of minor optimizations that are relatively expensive.
7404 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7408 Attempt to remove redundant extension instructions. This is especially
7409 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7410 registers after writing to their lower 32-bit half.
7412 Enabled for x86 at levels @option{-O2}, @option{-O3}.
7414 @item -flive-range-shrinkage
7415 @opindex flive-range-shrinkage
7416 Attempt to decrease register pressure through register live range
7417 shrinkage. This is helpful for fast processors with small or moderate
7420 @item -fira-algorithm=@var{algorithm}
7421 Use the specified coloring algorithm for the integrated register
7422 allocator. The @var{algorithm} argument can be @samp{priority}, which
7423 specifies Chow's priority coloring, or @samp{CB}, which specifies
7424 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7425 for all architectures, but for those targets that do support it, it is
7426 the default because it generates better code.
7428 @item -fira-region=@var{region}
7429 Use specified regions for the integrated register allocator. The
7430 @var{region} argument should be one of the following:
7435 Use all loops as register allocation regions.
7436 This can give the best results for machines with a small and/or
7437 irregular register set.
7440 Use all loops except for loops with small register pressure
7441 as the regions. This value usually gives
7442 the best results in most cases and for most architectures,
7443 and is enabled by default when compiling with optimization for speed
7444 (@option{-O}, @option{-O2}, @dots{}).
7447 Use all functions as a single region.
7448 This typically results in the smallest code size, and is enabled by default for
7449 @option{-Os} or @option{-O0}.
7453 @item -fira-hoist-pressure
7454 @opindex fira-hoist-pressure
7455 Use IRA to evaluate register pressure in the code hoisting pass for
7456 decisions to hoist expressions. This option usually results in smaller
7457 code, but it can slow the compiler down.
7459 This option is enabled at level @option{-Os} for all targets.
7461 @item -fira-loop-pressure
7462 @opindex fira-loop-pressure
7463 Use IRA to evaluate register pressure in loops for decisions to move
7464 loop invariants. This option usually results in generation
7465 of faster and smaller code on machines with large register files (>= 32
7466 registers), but it can slow the compiler down.
7468 This option is enabled at level @option{-O3} for some targets.
7470 @item -fno-ira-share-save-slots
7471 @opindex fno-ira-share-save-slots
7472 Disable sharing of stack slots used for saving call-used hard
7473 registers living through a call. Each hard register gets a
7474 separate stack slot, and as a result function stack frames are
7477 @item -fno-ira-share-spill-slots
7478 @opindex fno-ira-share-spill-slots
7479 Disable sharing of stack slots allocated for pseudo-registers. Each
7480 pseudo-register that does not get a hard register gets a separate
7481 stack slot, and as a result function stack frames are larger.
7483 @item -fira-verbose=@var{n}
7484 @opindex fira-verbose
7485 Control the verbosity of the dump file for the integrated register allocator.
7486 The default value is 5. If the value @var{n} is greater or equal to 10,
7487 the dump output is sent to stderr using the same format as @var{n} minus 10.
7489 @item -fdelayed-branch
7490 @opindex fdelayed-branch
7491 If supported for the target machine, attempt to reorder instructions
7492 to exploit instruction slots available after delayed branch
7495 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7497 @item -fschedule-insns
7498 @opindex fschedule-insns
7499 If supported for the target machine, attempt to reorder instructions to
7500 eliminate execution stalls due to required data being unavailable. This
7501 helps machines that have slow floating point or memory load instructions
7502 by allowing other instructions to be issued until the result of the load
7503 or floating-point instruction is required.
7505 Enabled at levels @option{-O2}, @option{-O3}.
7507 @item -fschedule-insns2
7508 @opindex fschedule-insns2
7509 Similar to @option{-fschedule-insns}, but requests an additional pass of
7510 instruction scheduling after register allocation has been done. This is
7511 especially useful on machines with a relatively small number of
7512 registers and where memory load instructions take more than one cycle.
7514 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7516 @item -fno-sched-interblock
7517 @opindex fno-sched-interblock
7518 Don't schedule instructions across basic blocks. This is normally
7519 enabled by default when scheduling before register allocation, i.e.@:
7520 with @option{-fschedule-insns} or at @option{-O2} or higher.
7522 @item -fno-sched-spec
7523 @opindex fno-sched-spec
7524 Don't allow speculative motion of non-load instructions. This is normally
7525 enabled by default when scheduling before register allocation, i.e.@:
7526 with @option{-fschedule-insns} or at @option{-O2} or higher.
7528 @item -fsched-pressure
7529 @opindex fsched-pressure
7530 Enable register pressure sensitive insn scheduling before register
7531 allocation. This only makes sense when scheduling before register
7532 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7533 @option{-O2} or higher. Usage of this option can improve the
7534 generated code and decrease its size by preventing register pressure
7535 increase above the number of available hard registers and subsequent
7536 spills in register allocation.
7538 @item -fsched-spec-load
7539 @opindex fsched-spec-load
7540 Allow speculative motion of some load instructions. This only makes
7541 sense when scheduling before register allocation, i.e.@: with
7542 @option{-fschedule-insns} or at @option{-O2} or higher.
7544 @item -fsched-spec-load-dangerous
7545 @opindex fsched-spec-load-dangerous
7546 Allow speculative motion of more load instructions. This only makes
7547 sense when scheduling before register allocation, i.e.@: with
7548 @option{-fschedule-insns} or at @option{-O2} or higher.
7550 @item -fsched-stalled-insns
7551 @itemx -fsched-stalled-insns=@var{n}
7552 @opindex fsched-stalled-insns
7553 Define how many insns (if any) can be moved prematurely from the queue
7554 of stalled insns into the ready list during the second scheduling pass.
7555 @option{-fno-sched-stalled-insns} means that no insns are moved
7556 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7557 on how many queued insns can be moved prematurely.
7558 @option{-fsched-stalled-insns} without a value is equivalent to
7559 @option{-fsched-stalled-insns=1}.
7561 @item -fsched-stalled-insns-dep
7562 @itemx -fsched-stalled-insns-dep=@var{n}
7563 @opindex fsched-stalled-insns-dep
7564 Define how many insn groups (cycles) are examined for a dependency
7565 on a stalled insn that is a candidate for premature removal from the queue
7566 of stalled insns. This has an effect only during the second scheduling pass,
7567 and only if @option{-fsched-stalled-insns} is used.
7568 @option{-fno-sched-stalled-insns-dep} is equivalent to
7569 @option{-fsched-stalled-insns-dep=0}.
7570 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7571 @option{-fsched-stalled-insns-dep=1}.
7573 @item -fsched2-use-superblocks
7574 @opindex fsched2-use-superblocks
7575 When scheduling after register allocation, use superblock scheduling.
7576 This allows motion across basic block boundaries,
7577 resulting in faster schedules. This option is experimental, as not all machine
7578 descriptions used by GCC model the CPU closely enough to avoid unreliable
7579 results from the algorithm.
7581 This only makes sense when scheduling after register allocation, i.e.@: with
7582 @option{-fschedule-insns2} or at @option{-O2} or higher.
7584 @item -fsched-group-heuristic
7585 @opindex fsched-group-heuristic
7586 Enable the group heuristic in the scheduler. This heuristic favors
7587 the instruction that belongs to a schedule group. This is enabled
7588 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7589 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7591 @item -fsched-critical-path-heuristic
7592 @opindex fsched-critical-path-heuristic
7593 Enable the critical-path heuristic in the scheduler. This heuristic favors
7594 instructions on the critical path. This is enabled by default when
7595 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7596 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7598 @item -fsched-spec-insn-heuristic
7599 @opindex fsched-spec-insn-heuristic
7600 Enable the speculative instruction heuristic in the scheduler. This
7601 heuristic favors speculative instructions with greater dependency weakness.
7602 This is enabled by default when scheduling is enabled, i.e.@:
7603 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7604 or at @option{-O2} or higher.
7606 @item -fsched-rank-heuristic
7607 @opindex fsched-rank-heuristic
7608 Enable the rank heuristic in the scheduler. This heuristic favors
7609 the instruction belonging to a basic block with greater size or frequency.
7610 This is enabled by default when scheduling is enabled, i.e.@:
7611 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7612 at @option{-O2} or higher.
7614 @item -fsched-last-insn-heuristic
7615 @opindex fsched-last-insn-heuristic
7616 Enable the last-instruction heuristic in the scheduler. This heuristic
7617 favors the instruction that is less dependent on the last instruction
7618 scheduled. This is enabled by default when scheduling is enabled,
7619 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7620 at @option{-O2} or higher.
7622 @item -fsched-dep-count-heuristic
7623 @opindex fsched-dep-count-heuristic
7624 Enable the dependent-count heuristic in the scheduler. This heuristic
7625 favors the instruction that has more instructions depending on it.
7626 This is enabled by default when scheduling is enabled, i.e.@:
7627 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7628 at @option{-O2} or higher.
7630 @item -freschedule-modulo-scheduled-loops
7631 @opindex freschedule-modulo-scheduled-loops
7632 Modulo scheduling is performed before traditional scheduling. If a loop
7633 is modulo scheduled, later scheduling passes may change its schedule.
7634 Use this option to control that behavior.
7636 @item -fselective-scheduling
7637 @opindex fselective-scheduling
7638 Schedule instructions using selective scheduling algorithm. Selective
7639 scheduling runs instead of the first scheduler pass.
7641 @item -fselective-scheduling2
7642 @opindex fselective-scheduling2
7643 Schedule instructions using selective scheduling algorithm. Selective
7644 scheduling runs instead of the second scheduler pass.
7646 @item -fsel-sched-pipelining
7647 @opindex fsel-sched-pipelining
7648 Enable software pipelining of innermost loops during selective scheduling.
7649 This option has no effect unless one of @option{-fselective-scheduling} or
7650 @option{-fselective-scheduling2} is turned on.
7652 @item -fsel-sched-pipelining-outer-loops
7653 @opindex fsel-sched-pipelining-outer-loops
7654 When pipelining loops during selective scheduling, also pipeline outer loops.
7655 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7658 @opindex fshrink-wrap
7659 Emit function prologues only before parts of the function that need it,
7660 rather than at the top of the function. This flag is enabled by default at
7661 @option{-O} and higher.
7663 @item -fcaller-saves
7664 @opindex fcaller-saves
7665 Enable allocation of values to registers that are clobbered by
7666 function calls, by emitting extra instructions to save and restore the
7667 registers around such calls. Such allocation is done only when it
7668 seems to result in better code.
7670 This option is always enabled by default on certain machines, usually
7671 those which have no call-preserved registers to use instead.
7673 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7675 @item -fcombine-stack-adjustments
7676 @opindex fcombine-stack-adjustments
7677 Tracks stack adjustments (pushes and pops) and stack memory references
7678 and then tries to find ways to combine them.
7680 Enabled by default at @option{-O1} and higher.
7682 @item -fconserve-stack
7683 @opindex fconserve-stack
7684 Attempt to minimize stack usage. The compiler attempts to use less
7685 stack space, even if that makes the program slower. This option
7686 implies setting the @option{large-stack-frame} parameter to 100
7687 and the @option{large-stack-frame-growth} parameter to 400.
7689 @item -ftree-reassoc
7690 @opindex ftree-reassoc
7691 Perform reassociation on trees. This flag is enabled by default
7692 at @option{-O} and higher.
7696 Perform partial redundancy elimination (PRE) on trees. This flag is
7697 enabled by default at @option{-O2} and @option{-O3}.
7699 @item -ftree-partial-pre
7700 @opindex ftree-partial-pre
7701 Make partial redundancy elimination (PRE) more aggressive. This flag is
7702 enabled by default at @option{-O3}.
7704 @item -ftree-forwprop
7705 @opindex ftree-forwprop
7706 Perform forward propagation on trees. This flag is enabled by default
7707 at @option{-O} and higher.
7711 Perform full redundancy elimination (FRE) on trees. The difference
7712 between FRE and PRE is that FRE only considers expressions
7713 that are computed on all paths leading to the redundant computation.
7714 This analysis is faster than PRE, though it exposes fewer redundancies.
7715 This flag is enabled by default at @option{-O} and higher.
7717 @item -ftree-phiprop
7718 @opindex ftree-phiprop
7719 Perform hoisting of loads from conditional pointers on trees. This
7720 pass is enabled by default at @option{-O} and higher.
7722 @item -fhoist-adjacent-loads
7723 @opindex hoist-adjacent-loads
7724 Speculatively hoist loads from both branches of an if-then-else if the
7725 loads are from adjacent locations in the same structure and the target
7726 architecture has a conditional move instruction. This flag is enabled
7727 by default at @option{-O2} and higher.
7729 @item -ftree-copy-prop
7730 @opindex ftree-copy-prop
7731 Perform copy propagation on trees. This pass eliminates unnecessary
7732 copy operations. This flag is enabled by default at @option{-O} and
7735 @item -fipa-pure-const
7736 @opindex fipa-pure-const
7737 Discover which functions are pure or constant.
7738 Enabled by default at @option{-O} and higher.
7740 @item -fipa-reference
7741 @opindex fipa-reference
7742 Discover which static variables do not escape the
7744 Enabled by default at @option{-O} and higher.
7748 Perform interprocedural pointer analysis and interprocedural modification
7749 and reference analysis. This option can cause excessive memory and
7750 compile-time usage on large compilation units. It is not enabled by
7751 default at any optimization level.
7754 @opindex fipa-profile
7755 Perform interprocedural profile propagation. The functions called only from
7756 cold functions are marked as cold. Also functions executed once (such as
7757 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7758 functions and loop less parts of functions executed once are then optimized for
7760 Enabled by default at @option{-O} and higher.
7764 Perform interprocedural constant propagation.
7765 This optimization analyzes the program to determine when values passed
7766 to functions are constants and then optimizes accordingly.
7767 This optimization can substantially increase performance
7768 if the application has constants passed to functions.
7769 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7771 @item -fipa-cp-clone
7772 @opindex fipa-cp-clone
7773 Perform function cloning to make interprocedural constant propagation stronger.
7774 When enabled, interprocedural constant propagation performs function cloning
7775 when externally visible function can be called with constant arguments.
7776 Because this optimization can create multiple copies of functions,
7777 it may significantly increase code size
7778 (see @option{--param ipcp-unit-growth=@var{value}}).
7779 This flag is enabled by default at @option{-O3}.
7781 @item -fisolate-erroneous-paths-dereference
7782 Detect paths which trigger erroneous or undefined behaviour due to
7783 dereferencing a NULL pointer. Isolate those paths from the main control
7784 flow and turn the statement with erroneous or undefined behaviour into a trap.
7786 @item -fisolate-erroneous-paths-attribute
7787 Detect paths which trigger erroneous or undefined behaviour due a NULL value
7788 being used in a way which is forbidden by a @code{returns_nonnull} or @code{nonnull}
7789 attribute. Isolate those paths from the main control flow and turn the
7790 statement with erroneous or undefined behaviour into a trap. This is not
7791 currently enabled, but may be enabled by @code{-O2} in the future.
7795 Perform forward store motion on trees. This flag is
7796 enabled by default at @option{-O} and higher.
7798 @item -ftree-bit-ccp
7799 @opindex ftree-bit-ccp
7800 Perform sparse conditional bit constant propagation on trees and propagate
7801 pointer alignment information.
7802 This pass only operates on local scalar variables and is enabled by default
7803 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7807 Perform sparse conditional constant propagation (CCP) on trees. This
7808 pass only operates on local scalar variables and is enabled by default
7809 at @option{-O} and higher.
7811 @item -ftree-switch-conversion
7812 Perform conversion of simple initializations in a switch to
7813 initializations from a scalar array. This flag is enabled by default
7814 at @option{-O2} and higher.
7816 @item -ftree-tail-merge
7817 Look for identical code sequences. When found, replace one with a jump to the
7818 other. This optimization is known as tail merging or cross jumping. This flag
7819 is enabled by default at @option{-O2} and higher. The compilation time
7821 be limited using @option{max-tail-merge-comparisons} parameter and
7822 @option{max-tail-merge-iterations} parameter.
7826 Perform dead code elimination (DCE) on trees. This flag is enabled by
7827 default at @option{-O} and higher.
7829 @item -ftree-builtin-call-dce
7830 @opindex ftree-builtin-call-dce
7831 Perform conditional dead code elimination (DCE) for calls to built-in functions
7832 that may set @code{errno} but are otherwise side-effect free. This flag is
7833 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7836 @item -ftree-dominator-opts
7837 @opindex ftree-dominator-opts
7838 Perform a variety of simple scalar cleanups (constant/copy
7839 propagation, redundancy elimination, range propagation and expression
7840 simplification) based on a dominator tree traversal. This also
7841 performs jump threading (to reduce jumps to jumps). This flag is
7842 enabled by default at @option{-O} and higher.
7846 Perform dead store elimination (DSE) on trees. A dead store is a store into
7847 a memory location that is later overwritten by another store without
7848 any intervening loads. In this case the earlier store can be deleted. This
7849 flag is enabled by default at @option{-O} and higher.
7853 Perform loop header copying on trees. This is beneficial since it increases
7854 effectiveness of code motion optimizations. It also saves one jump. This flag
7855 is enabled by default at @option{-O} and higher. It is not enabled
7856 for @option{-Os}, since it usually increases code size.
7858 @item -ftree-loop-optimize
7859 @opindex ftree-loop-optimize
7860 Perform loop optimizations on trees. This flag is enabled by default
7861 at @option{-O} and higher.
7863 @item -ftree-loop-linear
7864 @opindex ftree-loop-linear
7865 Perform loop interchange transformations on tree. Same as
7866 @option{-floop-interchange}. To use this code transformation, GCC has
7867 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7868 enable the Graphite loop transformation infrastructure.
7870 @item -floop-interchange
7871 @opindex floop-interchange
7872 Perform loop interchange transformations on loops. Interchanging two
7873 nested loops switches the inner and outer loops. For example, given a
7878 A(J, I) = A(J, I) * C
7882 loop interchange transforms the loop as if it were written:
7886 A(J, I) = A(J, I) * C
7890 which can be beneficial when @code{N} is larger than the caches,
7891 because in Fortran, the elements of an array are stored in memory
7892 contiguously by column, and the original loop iterates over rows,
7893 potentially creating at each access a cache miss. This optimization
7894 applies to all the languages supported by GCC and is not limited to
7895 Fortran. To use this code transformation, GCC has to be configured
7896 with @option{--with-ppl} and @option{--with-cloog} to enable the
7897 Graphite loop transformation infrastructure.
7899 @item -floop-strip-mine
7900 @opindex floop-strip-mine
7901 Perform loop strip mining transformations on loops. Strip mining
7902 splits a loop into two nested loops. The outer loop has strides
7903 equal to the strip size and the inner loop has strides of the
7904 original loop within a strip. The strip length can be changed
7905 using the @option{loop-block-tile-size} parameter. For example,
7912 loop strip mining transforms the loop as if it were written:
7915 DO I = II, min (II + 50, N)
7920 This optimization applies to all the languages supported by GCC and is
7921 not limited to Fortran. To use this code transformation, GCC has to
7922 be configured with @option{--with-ppl} and @option{--with-cloog} to
7923 enable the Graphite loop transformation infrastructure.
7926 @opindex floop-block
7927 Perform loop blocking transformations on loops. Blocking strip mines
7928 each loop in the loop nest such that the memory accesses of the
7929 element loops fit inside caches. The strip length can be changed
7930 using the @option{loop-block-tile-size} parameter. For example, given
7935 A(J, I) = B(I) + C(J)
7939 loop blocking transforms the loop as if it were written:
7943 DO I = II, min (II + 50, N)
7944 DO J = JJ, min (JJ + 50, M)
7945 A(J, I) = B(I) + C(J)
7951 which can be beneficial when @code{M} is larger than the caches,
7952 because the innermost loop iterates over a smaller amount of data
7953 which can be kept in the caches. This optimization applies to all the
7954 languages supported by GCC and is not limited to Fortran. To use this
7955 code transformation, GCC has to be configured with @option{--with-ppl}
7956 and @option{--with-cloog} to enable the Graphite loop transformation
7959 @item -fgraphite-identity
7960 @opindex fgraphite-identity
7961 Enable the identity transformation for graphite. For every SCoP we generate
7962 the polyhedral representation and transform it back to gimple. Using
7963 @option{-fgraphite-identity} we can check the costs or benefits of the
7964 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7965 are also performed by the code generator CLooG, like index splitting and
7966 dead code elimination in loops.
7968 @item -floop-nest-optimize
7969 @opindex floop-nest-optimize
7970 Enable the ISL based loop nest optimizer. This is a generic loop nest
7971 optimizer based on the Pluto optimization algorithms. It calculates a loop
7972 structure optimized for data-locality and parallelism. This option
7975 @item -floop-parallelize-all
7976 @opindex floop-parallelize-all
7977 Use the Graphite data dependence analysis to identify loops that can
7978 be parallelized. Parallelize all the loops that can be analyzed to
7979 not contain loop carried dependences without checking that it is
7980 profitable to parallelize the loops.
7982 @item -fcheck-data-deps
7983 @opindex fcheck-data-deps
7984 Compare the results of several data dependence analyzers. This option
7985 is used for debugging the data dependence analyzers.
7987 @item -ftree-loop-if-convert
7988 Attempt to transform conditional jumps in the innermost loops to
7989 branch-less equivalents. The intent is to remove control-flow from
7990 the innermost loops in order to improve the ability of the
7991 vectorization pass to handle these loops. This is enabled by default
7992 if vectorization is enabled.
7994 @item -ftree-loop-if-convert-stores
7995 Attempt to also if-convert conditional jumps containing memory writes.
7996 This transformation can be unsafe for multi-threaded programs as it
7997 transforms conditional memory writes into unconditional memory writes.
8000 for (i = 0; i < N; i++)
8006 for (i = 0; i < N; i++)
8007 A[i] = cond ? expr : A[i];
8009 potentially producing data races.
8011 @item -ftree-loop-distribution
8012 Perform loop distribution. This flag can improve cache performance on
8013 big loop bodies and allow further loop optimizations, like
8014 parallelization or vectorization, to take place. For example, the loop
8031 @item -ftree-loop-distribute-patterns
8032 Perform loop distribution of patterns that can be code generated with
8033 calls to a library. This flag is enabled by default at @option{-O3}.
8035 This pass distributes the initialization loops and generates a call to
8036 memset zero. For example, the loop
8052 and the initialization loop is transformed into a call to memset zero.
8054 @item -ftree-loop-im
8055 @opindex ftree-loop-im
8056 Perform loop invariant motion on trees. This pass moves only invariants that
8057 are hard to handle at RTL level (function calls, operations that expand to
8058 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8059 operands of conditions that are invariant out of the loop, so that we can use
8060 just trivial invariantness analysis in loop unswitching. The pass also includes
8063 @item -ftree-loop-ivcanon
8064 @opindex ftree-loop-ivcanon
8065 Create a canonical counter for number of iterations in loops for which
8066 determining number of iterations requires complicated analysis. Later
8067 optimizations then may determine the number easily. Useful especially
8068 in connection with unrolling.
8072 Perform induction variable optimizations (strength reduction, induction
8073 variable merging and induction variable elimination) on trees.
8075 @item -ftree-parallelize-loops=n
8076 @opindex ftree-parallelize-loops
8077 Parallelize loops, i.e., split their iteration space to run in n threads.
8078 This is only possible for loops whose iterations are independent
8079 and can be arbitrarily reordered. The optimization is only
8080 profitable on multiprocessor machines, for loops that are CPU-intensive,
8081 rather than constrained e.g.@: by memory bandwidth. This option
8082 implies @option{-pthread}, and thus is only supported on targets
8083 that have support for @option{-pthread}.
8087 Perform function-local points-to analysis on trees. This flag is
8088 enabled by default at @option{-O} and higher.
8092 Perform scalar replacement of aggregates. This pass replaces structure
8093 references with scalars to prevent committing structures to memory too
8094 early. This flag is enabled by default at @option{-O} and higher.
8096 @item -ftree-copyrename
8097 @opindex ftree-copyrename
8098 Perform copy renaming on trees. This pass attempts to rename compiler
8099 temporaries to other variables at copy locations, usually resulting in
8100 variable names which more closely resemble the original variables. This flag
8101 is enabled by default at @option{-O} and higher.
8103 @item -ftree-coalesce-inlined-vars
8104 @opindex ftree-coalesce-inlined-vars
8105 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8106 combine small user-defined variables too, but only if they were inlined
8107 from other functions. It is a more limited form of
8108 @option{-ftree-coalesce-vars}. This may harm debug information of such
8109 inlined variables, but it will keep variables of the inlined-into
8110 function apart from each other, such that they are more likely to
8111 contain the expected values in a debugging session. This was the
8112 default in GCC versions older than 4.7.
8114 @item -ftree-coalesce-vars
8115 @opindex ftree-coalesce-vars
8116 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8117 combine small user-defined variables too, instead of just compiler
8118 temporaries. This may severely limit the ability to debug an optimized
8119 program compiled with @option{-fno-var-tracking-assignments}. In the
8120 negated form, this flag prevents SSA coalescing of user variables,
8121 including inlined ones. This option is enabled by default.
8125 Perform temporary expression replacement during the SSA->normal phase. Single
8126 use/single def temporaries are replaced at their use location with their
8127 defining expression. This results in non-GIMPLE code, but gives the expanders
8128 much more complex trees to work on resulting in better RTL generation. This is
8129 enabled by default at @option{-O} and higher.
8133 Perform straight-line strength reduction on trees. This recognizes related
8134 expressions involving multiplications and replaces them by less expensive
8135 calculations when possible. This is enabled by default at @option{-O} and
8138 @item -ftree-vectorize
8139 @opindex ftree-vectorize
8140 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8141 and @option{-ftree-slp-vectorize} if not explicitly specified.
8143 @item -ftree-loop-vectorize
8144 @opindex ftree-loop-vectorize
8145 Perform loop vectorization on trees. This flag is enabled by default at
8146 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8148 @item -ftree-slp-vectorize
8149 @opindex ftree-slp-vectorize
8150 Perform basic block vectorization on trees. This flag is enabled by default at
8151 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8153 @item -fvect-cost-model=@var{model}
8154 @opindex fvect-cost-model
8155 Alter the cost model used for vectorization. The @var{model} argument
8156 should be one of @code{unlimited}, @code{dynamic} or @code{cheap}.
8157 With the @code{unlimited} model the vectorized code-path is assumed
8158 to be profitable while with the @code{dynamic} model a runtime check
8159 will guard the vectorized code-path to enable it only for iteration
8160 counts that will likely execute faster than when executing the original
8161 scalar loop. The @code{cheap} model will disable vectorization of
8162 loops where doing so would be cost prohibitive for example due to
8163 required runtime checks for data dependence or alignment but otherwise
8164 is equal to the @code{dynamic} model.
8165 The default cost model depends on other optimization flags and is
8166 either @code{dynamic} or @code{cheap}.
8168 @item -fsimd-cost-model=@var{model}
8169 @opindex fsimd-cost-model
8170 Alter the cost model used for vectorization of loops marked with the OpenMP
8171 or Cilk Plus simd directive. The @var{model} argument should be one of
8172 @code{unlimited}, @code{dynamic}, @code{cheap}. All values of @var{model}
8173 have the same meaning as described in @option{-fvect-cost-model} and by
8174 default a cost model defined with @option{-fvect-cost-model} is used.
8178 Perform Value Range Propagation on trees. This is similar to the
8179 constant propagation pass, but instead of values, ranges of values are
8180 propagated. This allows the optimizers to remove unnecessary range
8181 checks like array bound checks and null pointer checks. This is
8182 enabled by default at @option{-O2} and higher. Null pointer check
8183 elimination is only done if @option{-fdelete-null-pointer-checks} is
8188 Perform tail duplication to enlarge superblock size. This transformation
8189 simplifies the control flow of the function allowing other optimizations to do
8192 @item -funroll-loops
8193 @opindex funroll-loops
8194 Unroll loops whose number of iterations can be determined at compile
8195 time or upon entry to the loop. @option{-funroll-loops} implies
8196 @option{-frerun-cse-after-loop}. This option makes code larger,
8197 and may or may not make it run faster.
8199 @item -funroll-all-loops
8200 @opindex funroll-all-loops
8201 Unroll all loops, even if their number of iterations is uncertain when
8202 the loop is entered. This usually makes programs run more slowly.
8203 @option{-funroll-all-loops} implies the same options as
8204 @option{-funroll-loops},
8206 @item -fsplit-ivs-in-unroller
8207 @opindex fsplit-ivs-in-unroller
8208 Enables expression of values of induction variables in later iterations
8209 of the unrolled loop using the value in the first iteration. This breaks
8210 long dependency chains, thus improving efficiency of the scheduling passes.
8212 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8213 same effect. However, that is not reliable in cases where the loop body
8214 is more complicated than a single basic block. It also does not work at all
8215 on some architectures due to restrictions in the CSE pass.
8217 This optimization is enabled by default.
8219 @item -fvariable-expansion-in-unroller
8220 @opindex fvariable-expansion-in-unroller
8221 With this option, the compiler creates multiple copies of some
8222 local variables when unrolling a loop, which can result in superior code.
8224 @item -fpartial-inlining
8225 @opindex fpartial-inlining
8226 Inline parts of functions. This option has any effect only
8227 when inlining itself is turned on by the @option{-finline-functions}
8228 or @option{-finline-small-functions} options.
8230 Enabled at level @option{-O2}.
8232 @item -fpredictive-commoning
8233 @opindex fpredictive-commoning
8234 Perform predictive commoning optimization, i.e., reusing computations
8235 (especially memory loads and stores) performed in previous
8236 iterations of loops.
8238 This option is enabled at level @option{-O3}.
8240 @item -fprefetch-loop-arrays
8241 @opindex fprefetch-loop-arrays
8242 If supported by the target machine, generate instructions to prefetch
8243 memory to improve the performance of loops that access large arrays.
8245 This option may generate better or worse code; results are highly
8246 dependent on the structure of loops within the source code.
8248 Disabled at level @option{-Os}.
8251 @itemx -fno-peephole2
8252 @opindex fno-peephole
8253 @opindex fno-peephole2
8254 Disable any machine-specific peephole optimizations. The difference
8255 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8256 are implemented in the compiler; some targets use one, some use the
8257 other, a few use both.
8259 @option{-fpeephole} is enabled by default.
8260 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8262 @item -fno-guess-branch-probability
8263 @opindex fno-guess-branch-probability
8264 Do not guess branch probabilities using heuristics.
8266 GCC uses heuristics to guess branch probabilities if they are
8267 not provided by profiling feedback (@option{-fprofile-arcs}). These
8268 heuristics are based on the control flow graph. If some branch probabilities
8269 are specified by @samp{__builtin_expect}, then the heuristics are
8270 used to guess branch probabilities for the rest of the control flow graph,
8271 taking the @samp{__builtin_expect} info into account. The interactions
8272 between the heuristics and @samp{__builtin_expect} can be complex, and in
8273 some cases, it may be useful to disable the heuristics so that the effects
8274 of @samp{__builtin_expect} are easier to understand.
8276 The default is @option{-fguess-branch-probability} at levels
8277 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8279 @item -freorder-blocks
8280 @opindex freorder-blocks
8281 Reorder basic blocks in the compiled function in order to reduce number of
8282 taken branches and improve code locality.
8284 Enabled at levels @option{-O2}, @option{-O3}.
8286 @item -freorder-blocks-and-partition
8287 @opindex freorder-blocks-and-partition
8288 In addition to reordering basic blocks in the compiled function, in order
8289 to reduce number of taken branches, partitions hot and cold basic blocks
8290 into separate sections of the assembly and .o files, to improve
8291 paging and cache locality performance.
8293 This optimization is automatically turned off in the presence of
8294 exception handling, for linkonce sections, for functions with a user-defined
8295 section attribute and on any architecture that does not support named
8298 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8300 @item -freorder-functions
8301 @opindex freorder-functions
8302 Reorder functions in the object file in order to
8303 improve code locality. This is implemented by using special
8304 subsections @code{.text.hot} for most frequently executed functions and
8305 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8306 the linker so object file format must support named sections and linker must
8307 place them in a reasonable way.
8309 Also profile feedback must be available to make this option effective. See
8310 @option{-fprofile-arcs} for details.
8312 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8314 @item -fstrict-aliasing
8315 @opindex fstrict-aliasing
8316 Allow the compiler to assume the strictest aliasing rules applicable to
8317 the language being compiled. For C (and C++), this activates
8318 optimizations based on the type of expressions. In particular, an
8319 object of one type is assumed never to reside at the same address as an
8320 object of a different type, unless the types are almost the same. For
8321 example, an @code{unsigned int} can alias an @code{int}, but not a
8322 @code{void*} or a @code{double}. A character type may alias any other
8325 @anchor{Type-punning}Pay special attention to code like this:
8338 The practice of reading from a different union member than the one most
8339 recently written to (called ``type-punning'') is common. Even with
8340 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8341 is accessed through the union type. So, the code above works as
8342 expected. @xref{Structures unions enumerations and bit-fields
8343 implementation}. However, this code might not:
8354 Similarly, access by taking the address, casting the resulting pointer
8355 and dereferencing the result has undefined behavior, even if the cast
8356 uses a union type, e.g.:
8360 return ((union a_union *) &d)->i;
8364 The @option{-fstrict-aliasing} option is enabled at levels
8365 @option{-O2}, @option{-O3}, @option{-Os}.
8367 @item -fstrict-overflow
8368 @opindex fstrict-overflow
8369 Allow the compiler to assume strict signed overflow rules, depending
8370 on the language being compiled. For C (and C++) this means that
8371 overflow when doing arithmetic with signed numbers is undefined, which
8372 means that the compiler may assume that it does not happen. This
8373 permits various optimizations. For example, the compiler assumes
8374 that an expression like @code{i + 10 > i} is always true for
8375 signed @code{i}. This assumption is only valid if signed overflow is
8376 undefined, as the expression is false if @code{i + 10} overflows when
8377 using twos complement arithmetic. When this option is in effect any
8378 attempt to determine whether an operation on signed numbers
8379 overflows must be written carefully to not actually involve overflow.
8381 This option also allows the compiler to assume strict pointer
8382 semantics: given a pointer to an object, if adding an offset to that
8383 pointer does not produce a pointer to the same object, the addition is
8384 undefined. This permits the compiler to conclude that @code{p + u >
8385 p} is always true for a pointer @code{p} and unsigned integer
8386 @code{u}. This assumption is only valid because pointer wraparound is
8387 undefined, as the expression is false if @code{p + u} overflows using
8388 twos complement arithmetic.
8390 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8391 that integer signed overflow is fully defined: it wraps. When
8392 @option{-fwrapv} is used, there is no difference between
8393 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8394 integers. With @option{-fwrapv} certain types of overflow are
8395 permitted. For example, if the compiler gets an overflow when doing
8396 arithmetic on constants, the overflowed value can still be used with
8397 @option{-fwrapv}, but not otherwise.
8399 The @option{-fstrict-overflow} option is enabled at levels
8400 @option{-O2}, @option{-O3}, @option{-Os}.
8402 @item -falign-functions
8403 @itemx -falign-functions=@var{n}
8404 @opindex falign-functions
8405 Align the start of functions to the next power-of-two greater than
8406 @var{n}, skipping up to @var{n} bytes. For instance,
8407 @option{-falign-functions=32} aligns functions to the next 32-byte
8408 boundary, but @option{-falign-functions=24} aligns to the next
8409 32-byte boundary only if this can be done by skipping 23 bytes or less.
8411 @option{-fno-align-functions} and @option{-falign-functions=1} are
8412 equivalent and mean that functions are not aligned.
8414 Some assemblers only support this flag when @var{n} is a power of two;
8415 in that case, it is rounded up.
8417 If @var{n} is not specified or is zero, use a machine-dependent default.
8419 Enabled at levels @option{-O2}, @option{-O3}.
8421 @item -falign-labels
8422 @itemx -falign-labels=@var{n}
8423 @opindex falign-labels
8424 Align all branch targets to a power-of-two boundary, skipping up to
8425 @var{n} bytes like @option{-falign-functions}. This option can easily
8426 make code slower, because it must insert dummy operations for when the
8427 branch target is reached in the usual flow of the code.
8429 @option{-fno-align-labels} and @option{-falign-labels=1} are
8430 equivalent and mean that labels are not aligned.
8432 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8433 are greater than this value, then their values are used instead.
8435 If @var{n} is not specified or is zero, use a machine-dependent default
8436 which is very likely to be @samp{1}, meaning no alignment.
8438 Enabled at levels @option{-O2}, @option{-O3}.
8441 @itemx -falign-loops=@var{n}
8442 @opindex falign-loops
8443 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8444 like @option{-falign-functions}. If the loops are
8445 executed many times, this makes up for any execution of the dummy
8448 @option{-fno-align-loops} and @option{-falign-loops=1} are
8449 equivalent and mean that loops are not aligned.
8451 If @var{n} is not specified or is zero, use a machine-dependent default.
8453 Enabled at levels @option{-O2}, @option{-O3}.
8456 @itemx -falign-jumps=@var{n}
8457 @opindex falign-jumps
8458 Align branch targets to a power-of-two boundary, for branch targets
8459 where the targets can only be reached by jumping, skipping up to @var{n}
8460 bytes like @option{-falign-functions}. In this case, no dummy operations
8463 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8464 equivalent and mean that loops are not aligned.
8466 If @var{n} is not specified or is zero, use a machine-dependent default.
8468 Enabled at levels @option{-O2}, @option{-O3}.
8470 @item -funit-at-a-time
8471 @opindex funit-at-a-time
8472 This option is left for compatibility reasons. @option{-funit-at-a-time}
8473 has no effect, while @option{-fno-unit-at-a-time} implies
8474 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8478 @item -fno-toplevel-reorder
8479 @opindex fno-toplevel-reorder
8480 Do not reorder top-level functions, variables, and @code{asm}
8481 statements. Output them in the same order that they appear in the
8482 input file. When this option is used, unreferenced static variables
8483 are not removed. This option is intended to support existing code
8484 that relies on a particular ordering. For new code, it is better to
8485 use attributes when possible.
8487 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8488 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8493 Constructs webs as commonly used for register allocation purposes and assign
8494 each web individual pseudo register. This allows the register allocation pass
8495 to operate on pseudos directly, but also strengthens several other optimization
8496 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8497 however, make debugging impossible, since variables no longer stay in a
8500 Enabled by default with @option{-funroll-loops}.
8502 @item -fwhole-program
8503 @opindex fwhole-program
8504 Assume that the current compilation unit represents the whole program being
8505 compiled. All public functions and variables with the exception of @code{main}
8506 and those merged by attribute @code{externally_visible} become static functions
8507 and in effect are optimized more aggressively by interprocedural optimizers.
8509 This option should not be used in combination with @code{-flto}.
8510 Instead relying on a linker plugin should provide safer and more precise
8513 @item -flto[=@var{n}]
8515 This option runs the standard link-time optimizer. When invoked
8516 with source code, it generates GIMPLE (one of GCC's internal
8517 representations) and writes it to special ELF sections in the object
8518 file. When the object files are linked together, all the function
8519 bodies are read from these ELF sections and instantiated as if they
8520 had been part of the same translation unit.
8522 To use the link-time optimizer, @option{-flto} and optimization
8523 options should be specified at compile time and during the final link.
8527 gcc -c -O2 -flto foo.c
8528 gcc -c -O2 -flto bar.c
8529 gcc -o myprog -flto -O2 foo.o bar.o
8532 The first two invocations to GCC save a bytecode representation
8533 of GIMPLE into special ELF sections inside @file{foo.o} and
8534 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8535 @file{foo.o} and @file{bar.o}, merges the two files into a single
8536 internal image, and compiles the result as usual. Since both
8537 @file{foo.o} and @file{bar.o} are merged into a single image, this
8538 causes all the interprocedural analyses and optimizations in GCC to
8539 work across the two files as if they were a single one. This means,
8540 for example, that the inliner is able to inline functions in
8541 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8543 Another (simpler) way to enable link-time optimization is:
8546 gcc -o myprog -flto -O2 foo.c bar.c
8549 The above generates bytecode for @file{foo.c} and @file{bar.c},
8550 merges them together into a single GIMPLE representation and optimizes
8551 them as usual to produce @file{myprog}.
8553 The only important thing to keep in mind is that to enable link-time
8554 optimizations you need to use the GCC driver to perform the link-step.
8555 GCC then automatically performs link-time optimization if any of the
8556 objects involved were compiled with the @option{-flto}. You generally
8557 should specify the optimization options to be used for link-time
8558 optimization though GCC will try to be clever at guessing an
8559 optimization level to use from the options used at compile-time
8560 if you fail to specify one at link-time. You can always override
8561 the automatic decision to do link-time optimization at link-time
8562 by passing @option{-fno-lto} to the link command.
8564 To make whole program optimization effective, it is necessary to make
8565 certain whole program assumptions. The compiler needs to know
8566 what functions and variables can be accessed by libraries and runtime
8567 outside of the link-time optimized unit. When supported by the linker,
8568 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8569 to the compiler about used and externally visible symbols. When
8570 the linker plugin is not available, @option{-fwhole-program} should be
8571 used to allow the compiler to make these assumptions, which leads
8572 to more aggressive optimization decisions.
8574 When @option{-fuse-linker-plugin} is not enabled then, when a file is
8575 compiled with @option{-flto}, the generated object file is larger than
8576 a regular object file because it contains GIMPLE bytecodes and the usual
8577 final code (see @option{-ffat-lto-objects}. This means that
8578 object files with LTO information can be linked as normal object
8579 files; if @option{-fno-lto} is passed to the linker, no
8580 interprocedural optimizations are applied. Note that when
8581 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
8582 but you cannot perform a regular, non-LTO link on them.
8584 Additionally, the optimization flags used to compile individual files
8585 are not necessarily related to those used at link time. For instance,
8588 gcc -c -O0 -ffat-lto-objects -flto foo.c
8589 gcc -c -O0 -ffat-lto-objects -flto bar.c
8590 gcc -o myprog -O3 foo.o bar.o
8593 This produces individual object files with unoptimized assembler
8594 code, but the resulting binary @file{myprog} is optimized at
8595 @option{-O3}. If, instead, the final binary is generated with
8596 @option{-fno-lto}, then @file{myprog} is not optimized.
8598 When producing the final binary, GCC only
8599 applies link-time optimizations to those files that contain bytecode.
8600 Therefore, you can mix and match object files and libraries with
8601 GIMPLE bytecodes and final object code. GCC automatically selects
8602 which files to optimize in LTO mode and which files to link without
8605 There are some code generation flags preserved by GCC when
8606 generating bytecodes, as they need to be used during the final link
8607 stage. Generally options specified at link-time override those
8608 specified at compile-time.
8610 If you do not specify an optimization level option @option{-O} at
8611 link-time then GCC will compute one based on the optimization levels
8612 used when compiling the object files. The highest optimization
8613 level will win here.
8615 Currently, the following options and their setting are take from
8616 the first object file that explicitely specified it:
8617 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8618 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8619 and all the @option{-m} target flags.
8621 Certain ABI changing flags are required to match in all compilation-units
8622 and trying to override this at link-time with a conflicting value
8623 is ignored. This includes options such as @option{-freg-struct-return}
8624 and @option{-fpcc-struct-return}.
8626 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8627 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8628 are passed through to the link stage and merged conservatively for
8629 conflicting translation units. Specifically
8630 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8631 precedence and for example @option{-ffp-contract=off} takes precedence
8632 over @option{-ffp-contract=fast}. You can override them at linke-time.
8634 It is recommended that you compile all the files participating in the
8635 same link with the same options and also specify those options at
8638 If LTO encounters objects with C linkage declared with incompatible
8639 types in separate translation units to be linked together (undefined
8640 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8641 issued. The behavior is still undefined at run time. Similar
8642 diagnostics may be raised for other languages.
8644 Another feature of LTO is that it is possible to apply interprocedural
8645 optimizations on files written in different languages:
8650 gfortran -c -flto baz.f90
8651 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8654 Notice that the final link is done with @command{g++} to get the C++
8655 runtime libraries and @option{-lgfortran} is added to get the Fortran
8656 runtime libraries. In general, when mixing languages in LTO mode, you
8657 should use the same link command options as when mixing languages in a
8658 regular (non-LTO) compilation.
8660 If object files containing GIMPLE bytecode are stored in a library archive, say
8661 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8662 are using a linker with plugin support. To create static libraries suitable
8663 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8664 and @code{ranlib}; to show the symbols of object files with GIMPLE bytecode, use
8665 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8666 and @command{nm} have been compiled with plugin support. At link time, use the the
8667 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8668 the LTO optimization process:
8671 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8674 With the linker plugin enabled, the linker extracts the needed
8675 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8676 to make them part of the aggregated GIMPLE image to be optimized.
8678 If you are not using a linker with plugin support and/or do not
8679 enable the linker plugin, then the objects inside @file{libfoo.a}
8680 are extracted and linked as usual, but they do not participate
8681 in the LTO optimization process. In order to make a static library suitable
8682 for both LTO optimization and usual linkage, compile its object files with
8683 @option{-flto} @code{-ffat-lto-objects}.
8685 Link-time optimizations do not require the presence of the whole program to
8686 operate. If the program does not require any symbols to be exported, it is
8687 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8688 the interprocedural optimizers to use more aggressive assumptions which may
8689 lead to improved optimization opportunities.
8690 Use of @option{-fwhole-program} is not needed when linker plugin is
8691 active (see @option{-fuse-linker-plugin}).
8693 The current implementation of LTO makes no
8694 attempt to generate bytecode that is portable between different
8695 types of hosts. The bytecode files are versioned and there is a
8696 strict version check, so bytecode files generated in one version of
8697 GCC will not work with an older or newer version of GCC.
8699 Link-time optimization does not work well with generation of debugging
8700 information. Combining @option{-flto} with
8701 @option{-g} is currently experimental and expected to produce unexpected
8704 If you specify the optional @var{n}, the optimization and code
8705 generation done at link time is executed in parallel using @var{n}
8706 parallel jobs by utilizing an installed @command{make} program. The
8707 environment variable @env{MAKE} may be used to override the program
8708 used. The default value for @var{n} is 1.
8710 You can also specify @option{-flto=jobserver} to use GNU make's
8711 job server mode to determine the number of parallel jobs. This
8712 is useful when the Makefile calling GCC is already executing in parallel.
8713 You must prepend a @samp{+} to the command recipe in the parent Makefile
8714 for this to work. This option likely only works if @env{MAKE} is
8717 @item -flto-partition=@var{alg}
8718 @opindex flto-partition
8719 Specify the partitioning algorithm used by the link-time optimizer.
8720 The value is either @code{1to1} to specify a partitioning mirroring
8721 the original source files or @code{balanced} to specify partitioning
8722 into equally sized chunks (whenever possible) or @code{max} to create
8723 new partition for every symbol where possible. Specifying @code{none}
8724 as an algorithm disables partitioning and streaming completely.
8725 The default value is @code{balanced}. While @code{1to1} can be used
8726 as an workaround for various code ordering issues, the @code{max}
8727 partitioning is intended for internal testing only.
8729 @item -flto-compression-level=@var{n}
8730 This option specifies the level of compression used for intermediate
8731 language written to LTO object files, and is only meaningful in
8732 conjunction with LTO mode (@option{-flto}). Valid
8733 values are 0 (no compression) to 9 (maximum compression). Values
8734 outside this range are clamped to either 0 or 9. If the option is not
8735 given, a default balanced compression setting is used.
8738 Prints a report with internal details on the workings of the link-time
8739 optimizer. The contents of this report vary from version to version.
8740 It is meant to be useful to GCC developers when processing object
8741 files in LTO mode (via @option{-flto}).
8743 Disabled by default.
8745 @item -flto-report-wpa
8746 Like @option{-flto-report}, but only print for the WPA phase of Link
8749 @item -fuse-linker-plugin
8750 Enables the use of a linker plugin during link-time optimization. This
8751 option relies on plugin support in the linker, which is available in gold
8752 or in GNU ld 2.21 or newer.
8754 This option enables the extraction of object files with GIMPLE bytecode out
8755 of library archives. This improves the quality of optimization by exposing
8756 more code to the link-time optimizer. This information specifies what
8757 symbols can be accessed externally (by non-LTO object or during dynamic
8758 linking). Resulting code quality improvements on binaries (and shared
8759 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
8760 See @option{-flto} for a description of the effect of this flag and how to
8763 This option is enabled by default when LTO support in GCC is enabled
8764 and GCC was configured for use with
8765 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8767 @item -ffat-lto-objects
8768 @opindex ffat-lto-objects
8769 Fat LTO objects are object files that contain both the intermediate language
8770 and the object code. This makes them usable for both LTO linking and normal
8771 linking. This option is effective only when compiling with @option{-flto}
8772 and is ignored at link time.
8774 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8775 requires the complete toolchain to be aware of LTO. It requires a linker with
8776 linker plugin support for basic functionality. Additionally,
8777 @command{nm}, @command{ar} and @command{ranlib}
8778 need to support linker plugins to allow a full-featured build environment
8779 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8780 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8781 to these tools. With non fat LTO makefiles need to be modified to use them.
8783 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8786 @item -fcompare-elim
8787 @opindex fcompare-elim
8788 After register allocation and post-register allocation instruction splitting,
8789 identify arithmetic instructions that compute processor flags similar to a
8790 comparison operation based on that arithmetic. If possible, eliminate the
8791 explicit comparison operation.
8793 This pass only applies to certain targets that cannot explicitly represent
8794 the comparison operation before register allocation is complete.
8796 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8799 @opindex fuse-ld=bfd
8800 Use the @command{bfd} linker instead of the default linker.
8803 @opindex fuse-ld=gold
8804 Use the @command{gold} linker instead of the default linker.
8806 @item -fcprop-registers
8807 @opindex fcprop-registers
8808 After register allocation and post-register allocation instruction splitting,
8809 perform a copy-propagation pass to try to reduce scheduling dependencies
8810 and occasionally eliminate the copy.
8812 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8814 @item -fprofile-correction
8815 @opindex fprofile-correction
8816 Profiles collected using an instrumented binary for multi-threaded programs may
8817 be inconsistent due to missed counter updates. When this option is specified,
8818 GCC uses heuristics to correct or smooth out such inconsistencies. By
8819 default, GCC emits an error message when an inconsistent profile is detected.
8821 @item -fprofile-dir=@var{path}
8822 @opindex fprofile-dir
8824 Set the directory to search for the profile data files in to @var{path}.
8825 This option affects only the profile data generated by
8826 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8827 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8828 and its related options. Both absolute and relative paths can be used.
8829 By default, GCC uses the current directory as @var{path}, thus the
8830 profile data file appears in the same directory as the object file.
8832 @item -fprofile-generate
8833 @itemx -fprofile-generate=@var{path}
8834 @opindex fprofile-generate
8836 Enable options usually used for instrumenting application to produce
8837 profile useful for later recompilation with profile feedback based
8838 optimization. You must use @option{-fprofile-generate} both when
8839 compiling and when linking your program.
8841 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
8843 If @var{path} is specified, GCC looks at the @var{path} to find
8844 the profile feedback data files. See @option{-fprofile-dir}.
8847 @itemx -fprofile-use=@var{path}
8848 @opindex fprofile-use
8849 Enable profile feedback directed optimizations, and optimizations
8850 generally profitable only with profile feedback available.
8852 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8853 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}, @code{-ftree-vectorize},
8854 @code{ftree-loop-distribute-patterns}
8856 By default, GCC emits an error message if the feedback profiles do not
8857 match the source code. This error can be turned into a warning by using
8858 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8861 If @var{path} is specified, GCC looks at the @var{path} to find
8862 the profile feedback data files. See @option{-fprofile-dir}.
8865 The following options control compiler behavior regarding floating-point
8866 arithmetic. These options trade off between speed and
8867 correctness. All must be specifically enabled.
8871 @opindex ffloat-store
8872 Do not store floating-point variables in registers, and inhibit other
8873 options that might change whether a floating-point value is taken from a
8876 @cindex floating-point precision
8877 This option prevents undesirable excess precision on machines such as
8878 the 68000 where the floating registers (of the 68881) keep more
8879 precision than a @code{double} is supposed to have. Similarly for the
8880 x86 architecture. For most programs, the excess precision does only
8881 good, but a few programs rely on the precise definition of IEEE floating
8882 point. Use @option{-ffloat-store} for such programs, after modifying
8883 them to store all pertinent intermediate computations into variables.
8885 @item -fexcess-precision=@var{style}
8886 @opindex fexcess-precision
8887 This option allows further control over excess precision on machines
8888 where floating-point registers have more precision than the IEEE
8889 @code{float} and @code{double} types and the processor does not
8890 support operations rounding to those types. By default,
8891 @option{-fexcess-precision=fast} is in effect; this means that
8892 operations are carried out in the precision of the registers and that
8893 it is unpredictable when rounding to the types specified in the source
8894 code takes place. When compiling C, if
8895 @option{-fexcess-precision=standard} is specified then excess
8896 precision follows the rules specified in ISO C99; in particular,
8897 both casts and assignments cause values to be rounded to their
8898 semantic types (whereas @option{-ffloat-store} only affects
8899 assignments). This option is enabled by default for C if a strict
8900 conformance option such as @option{-std=c99} is used.
8903 @option{-fexcess-precision=standard} is not implemented for languages
8904 other than C, and has no effect if
8905 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8906 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8907 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8908 semantics apply without excess precision, and in the latter, rounding
8913 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8914 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8915 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8917 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8919 This option is not turned on by any @option{-O} option besides
8920 @option{-Ofast} since it can result in incorrect output for programs
8921 that depend on an exact implementation of IEEE or ISO rules/specifications
8922 for math functions. It may, however, yield faster code for programs
8923 that do not require the guarantees of these specifications.
8925 @item -fno-math-errno
8926 @opindex fno-math-errno
8927 Do not set @code{errno} after calling math functions that are executed
8928 with a single instruction, e.g., @code{sqrt}. A program that relies on
8929 IEEE exceptions for math error handling may want to use this flag
8930 for speed while maintaining IEEE arithmetic compatibility.
8932 This option is not turned on by any @option{-O} option since
8933 it can result in incorrect output for programs that depend on
8934 an exact implementation of IEEE or ISO rules/specifications for
8935 math functions. It may, however, yield faster code for programs
8936 that do not require the guarantees of these specifications.
8938 The default is @option{-fmath-errno}.
8940 On Darwin systems, the math library never sets @code{errno}. There is
8941 therefore no reason for the compiler to consider the possibility that
8942 it might, and @option{-fno-math-errno} is the default.
8944 @item -funsafe-math-optimizations
8945 @opindex funsafe-math-optimizations
8947 Allow optimizations for floating-point arithmetic that (a) assume
8948 that arguments and results are valid and (b) may violate IEEE or
8949 ANSI standards. When used at link-time, it may include libraries
8950 or startup files that change the default FPU control word or other
8951 similar optimizations.
8953 This option is not turned on by any @option{-O} option since
8954 it can result in incorrect output for programs that depend on
8955 an exact implementation of IEEE or ISO rules/specifications for
8956 math functions. It may, however, yield faster code for programs
8957 that do not require the guarantees of these specifications.
8958 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8959 @option{-fassociative-math} and @option{-freciprocal-math}.
8961 The default is @option{-fno-unsafe-math-optimizations}.
8963 @item -fassociative-math
8964 @opindex fassociative-math
8966 Allow re-association of operands in series of floating-point operations.
8967 This violates the ISO C and C++ language standard by possibly changing
8968 computation result. NOTE: re-ordering may change the sign of zero as
8969 well as ignore NaNs and inhibit or create underflow or overflow (and
8970 thus cannot be used on code that relies on rounding behavior like
8971 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8972 and thus may not be used when ordered comparisons are required.
8973 This option requires that both @option{-fno-signed-zeros} and
8974 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8975 much sense with @option{-frounding-math}. For Fortran the option
8976 is automatically enabled when both @option{-fno-signed-zeros} and
8977 @option{-fno-trapping-math} are in effect.
8979 The default is @option{-fno-associative-math}.
8981 @item -freciprocal-math
8982 @opindex freciprocal-math
8984 Allow the reciprocal of a value to be used instead of dividing by
8985 the value if this enables optimizations. For example @code{x / y}
8986 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
8987 is subject to common subexpression elimination. Note that this loses
8988 precision and increases the number of flops operating on the value.
8990 The default is @option{-fno-reciprocal-math}.
8992 @item -ffinite-math-only
8993 @opindex ffinite-math-only
8994 Allow optimizations for floating-point arithmetic that assume
8995 that arguments and results are not NaNs or +-Infs.
8997 This option is not turned on by any @option{-O} option since
8998 it can result in incorrect output for programs that depend on
8999 an exact implementation of IEEE or ISO rules/specifications for
9000 math functions. It may, however, yield faster code for programs
9001 that do not require the guarantees of these specifications.
9003 The default is @option{-fno-finite-math-only}.
9005 @item -fno-signed-zeros
9006 @opindex fno-signed-zeros
9007 Allow optimizations for floating-point arithmetic that ignore the
9008 signedness of zero. IEEE arithmetic specifies the behavior of
9009 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9010 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9011 This option implies that the sign of a zero result isn't significant.
9013 The default is @option{-fsigned-zeros}.
9015 @item -fno-trapping-math
9016 @opindex fno-trapping-math
9017 Compile code assuming that floating-point operations cannot generate
9018 user-visible traps. These traps include division by zero, overflow,
9019 underflow, inexact result and invalid operation. This option requires
9020 that @option{-fno-signaling-nans} be in effect. Setting this option may
9021 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9023 This option should never be turned on by any @option{-O} option since
9024 it can result in incorrect output for programs that depend on
9025 an exact implementation of IEEE or ISO rules/specifications for
9028 The default is @option{-ftrapping-math}.
9030 @item -frounding-math
9031 @opindex frounding-math
9032 Disable transformations and optimizations that assume default floating-point
9033 rounding behavior. This is round-to-zero for all floating point
9034 to integer conversions, and round-to-nearest for all other arithmetic
9035 truncations. This option should be specified for programs that change
9036 the FP rounding mode dynamically, or that may be executed with a
9037 non-default rounding mode. This option disables constant folding of
9038 floating-point expressions at compile time (which may be affected by
9039 rounding mode) and arithmetic transformations that are unsafe in the
9040 presence of sign-dependent rounding modes.
9042 The default is @option{-fno-rounding-math}.
9044 This option is experimental and does not currently guarantee to
9045 disable all GCC optimizations that are affected by rounding mode.
9046 Future versions of GCC may provide finer control of this setting
9047 using C99's @code{FENV_ACCESS} pragma. This command-line option
9048 will be used to specify the default state for @code{FENV_ACCESS}.
9050 @item -fsignaling-nans
9051 @opindex fsignaling-nans
9052 Compile code assuming that IEEE signaling NaNs may generate user-visible
9053 traps during floating-point operations. Setting this option disables
9054 optimizations that may change the number of exceptions visible with
9055 signaling NaNs. This option implies @option{-ftrapping-math}.
9057 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9060 The default is @option{-fno-signaling-nans}.
9062 This option is experimental and does not currently guarantee to
9063 disable all GCC optimizations that affect signaling NaN behavior.
9065 @item -fsingle-precision-constant
9066 @opindex fsingle-precision-constant
9067 Treat floating-point constants as single precision instead of
9068 implicitly converting them to double-precision constants.
9070 @item -fcx-limited-range
9071 @opindex fcx-limited-range
9072 When enabled, this option states that a range reduction step is not
9073 needed when performing complex division. Also, there is no checking
9074 whether the result of a complex multiplication or division is @code{NaN
9075 + I*NaN}, with an attempt to rescue the situation in that case. The
9076 default is @option{-fno-cx-limited-range}, but is enabled by
9077 @option{-ffast-math}.
9079 This option controls the default setting of the ISO C99
9080 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9083 @item -fcx-fortran-rules
9084 @opindex fcx-fortran-rules
9085 Complex multiplication and division follow Fortran rules. Range
9086 reduction is done as part of complex division, but there is no checking
9087 whether the result of a complex multiplication or division is @code{NaN
9088 + I*NaN}, with an attempt to rescue the situation in that case.
9090 The default is @option{-fno-cx-fortran-rules}.
9094 The following options control optimizations that may improve
9095 performance, but are not enabled by any @option{-O} options. This
9096 section includes experimental options that may produce broken code.
9099 @item -fbranch-probabilities
9100 @opindex fbranch-probabilities
9101 After running a program compiled with @option{-fprofile-arcs}
9102 (@pxref{Debugging Options,, Options for Debugging Your Program or
9103 @command{gcc}}), you can compile it a second time using
9104 @option{-fbranch-probabilities}, to improve optimizations based on
9105 the number of times each branch was taken. When a program
9106 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9107 counts to a file called @file{@var{sourcename}.gcda} for each source
9108 file. The information in this data file is very dependent on the
9109 structure of the generated code, so you must use the same source code
9110 and the same optimization options for both compilations.
9112 With @option{-fbranch-probabilities}, GCC puts a
9113 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9114 These can be used to improve optimization. Currently, they are only
9115 used in one place: in @file{reorg.c}, instead of guessing which path a
9116 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9117 exactly determine which path is taken more often.
9119 @item -fprofile-values
9120 @opindex fprofile-values
9121 If combined with @option{-fprofile-arcs}, it adds code so that some
9122 data about values of expressions in the program is gathered.
9124 With @option{-fbranch-probabilities}, it reads back the data gathered
9125 from profiling values of expressions for usage in optimizations.
9127 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9129 @item -fprofile-reorder-functions
9130 @opindex fprofile-reorder-functions
9131 Function reordering based on profile instrumentation collects
9132 first time of execution of a function and orders these functions
9135 Enabled with @option{-fprofile-use}.
9139 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9140 to add code to gather information about values of expressions.
9142 With @option{-fbranch-probabilities}, it reads back the data gathered
9143 and actually performs the optimizations based on them.
9144 Currently the optimizations include specialization of division operations
9145 using the knowledge about the value of the denominator.
9147 @item -frename-registers
9148 @opindex frename-registers
9149 Attempt to avoid false dependencies in scheduled code by making use
9150 of registers left over after register allocation. This optimization
9151 most benefits processors with lots of registers. Depending on the
9152 debug information format adopted by the target, however, it can
9153 make debugging impossible, since variables no longer stay in
9154 a ``home register''.
9156 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9160 Perform tail duplication to enlarge superblock size. This transformation
9161 simplifies the control flow of the function allowing other optimizations to do
9164 Enabled with @option{-fprofile-use}.
9166 @item -funroll-loops
9167 @opindex funroll-loops
9168 Unroll loops whose number of iterations can be determined at compile time or
9169 upon entry to the loop. @option{-funroll-loops} implies
9170 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9171 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9172 a small constant number of iterations). This option makes code larger, and may
9173 or may not make it run faster.
9175 Enabled with @option{-fprofile-use}.
9177 @item -funroll-all-loops
9178 @opindex funroll-all-loops
9179 Unroll all loops, even if their number of iterations is uncertain when
9180 the loop is entered. This usually makes programs run more slowly.
9181 @option{-funroll-all-loops} implies the same options as
9182 @option{-funroll-loops}.
9185 @opindex fpeel-loops
9186 Peels loops for which there is enough information that they do not
9187 roll much (from profile feedback). It also turns on complete loop peeling
9188 (i.e.@: complete removal of loops with small constant number of iterations).
9190 Enabled with @option{-fprofile-use}.
9192 @item -fmove-loop-invariants
9193 @opindex fmove-loop-invariants
9194 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9195 at level @option{-O1}
9197 @item -funswitch-loops
9198 @opindex funswitch-loops
9199 Move branches with loop invariant conditions out of the loop, with duplicates
9200 of the loop on both branches (modified according to result of the condition).
9202 @item -ffunction-sections
9203 @itemx -fdata-sections
9204 @opindex ffunction-sections
9205 @opindex fdata-sections
9206 Place each function or data item into its own section in the output
9207 file if the target supports arbitrary sections. The name of the
9208 function or the name of the data item determines the section's name
9211 Use these options on systems where the linker can perform optimizations
9212 to improve locality of reference in the instruction space. Most systems
9213 using the ELF object format and SPARC processors running Solaris 2 have
9214 linkers with such optimizations. AIX may have these optimizations in
9217 Only use these options when there are significant benefits from doing
9218 so. When you specify these options, the assembler and linker
9219 create larger object and executable files and are also slower.
9220 You cannot use @code{gprof} on all systems if you
9221 specify this option, and you may have problems with debugging if
9222 you specify both this option and @option{-g}.
9224 @item -fbranch-target-load-optimize
9225 @opindex fbranch-target-load-optimize
9226 Perform branch target register load optimization before prologue / epilogue
9228 The use of target registers can typically be exposed only during reload,
9229 thus hoisting loads out of loops and doing inter-block scheduling needs
9230 a separate optimization pass.
9232 @item -fbranch-target-load-optimize2
9233 @opindex fbranch-target-load-optimize2
9234 Perform branch target register load optimization after prologue / epilogue
9237 @item -fbtr-bb-exclusive
9238 @opindex fbtr-bb-exclusive
9239 When performing branch target register load optimization, don't reuse
9240 branch target registers within any basic block.
9242 @item -fstack-protector
9243 @opindex fstack-protector
9244 Emit extra code to check for buffer overflows, such as stack smashing
9245 attacks. This is done by adding a guard variable to functions with
9246 vulnerable objects. This includes functions that call @code{alloca}, and
9247 functions with buffers larger than 8 bytes. The guards are initialized
9248 when a function is entered and then checked when the function exits.
9249 If a guard check fails, an error message is printed and the program exits.
9251 @item -fstack-protector-all
9252 @opindex fstack-protector-all
9253 Like @option{-fstack-protector} except that all functions are protected.
9255 @item -fstack-protector-strong
9256 @opindex fstack-protector-strong
9257 Like @option{-fstack-protector} but includes additional functions to
9258 be protected --- those that have local array definitions, or have
9259 references to local frame addresses.
9261 @item -fsection-anchors
9262 @opindex fsection-anchors
9263 Try to reduce the number of symbolic address calculations by using
9264 shared ``anchor'' symbols to address nearby objects. This transformation
9265 can help to reduce the number of GOT entries and GOT accesses on some
9268 For example, the implementation of the following function @code{foo}:
9272 int foo (void) @{ return a + b + c; @}
9276 usually calculates the addresses of all three variables, but if you
9277 compile it with @option{-fsection-anchors}, it accesses the variables
9278 from a common anchor point instead. The effect is similar to the
9279 following pseudocode (which isn't valid C):
9284 register int *xr = &x;
9285 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9289 Not all targets support this option.
9291 @item --param @var{name}=@var{value}
9293 In some places, GCC uses various constants to control the amount of
9294 optimization that is done. For example, GCC does not inline functions
9295 that contain more than a certain number of instructions. You can
9296 control some of these constants on the command line using the
9297 @option{--param} option.
9299 The names of specific parameters, and the meaning of the values, are
9300 tied to the internals of the compiler, and are subject to change
9301 without notice in future releases.
9303 In each case, the @var{value} is an integer. The allowable choices for
9307 @item predictable-branch-outcome
9308 When branch is predicted to be taken with probability lower than this threshold
9309 (in percent), then it is considered well predictable. The default is 10.
9311 @item max-crossjump-edges
9312 The maximum number of incoming edges to consider for cross-jumping.
9313 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9314 the number of edges incoming to each block. Increasing values mean
9315 more aggressive optimization, making the compilation time increase with
9316 probably small improvement in executable size.
9318 @item min-crossjump-insns
9319 The minimum number of instructions that must be matched at the end
9320 of two blocks before cross-jumping is performed on them. This
9321 value is ignored in the case where all instructions in the block being
9322 cross-jumped from are matched. The default value is 5.
9324 @item max-grow-copy-bb-insns
9325 The maximum code size expansion factor when copying basic blocks
9326 instead of jumping. The expansion is relative to a jump instruction.
9327 The default value is 8.
9329 @item max-goto-duplication-insns
9330 The maximum number of instructions to duplicate to a block that jumps
9331 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9332 passes, GCC factors computed gotos early in the compilation process,
9333 and unfactors them as late as possible. Only computed jumps at the
9334 end of a basic blocks with no more than max-goto-duplication-insns are
9335 unfactored. The default value is 8.
9337 @item max-delay-slot-insn-search
9338 The maximum number of instructions to consider when looking for an
9339 instruction to fill a delay slot. If more than this arbitrary number of
9340 instructions are searched, the time savings from filling the delay slot
9341 are minimal, so stop searching. Increasing values mean more
9342 aggressive optimization, making the compilation time increase with probably
9343 small improvement in execution time.
9345 @item max-delay-slot-live-search
9346 When trying to fill delay slots, the maximum number of instructions to
9347 consider when searching for a block with valid live register
9348 information. Increasing this arbitrarily chosen value means more
9349 aggressive optimization, increasing the compilation time. This parameter
9350 should be removed when the delay slot code is rewritten to maintain the
9353 @item max-gcse-memory
9354 The approximate maximum amount of memory that can be allocated in
9355 order to perform the global common subexpression elimination
9356 optimization. If more memory than specified is required, the
9357 optimization is not done.
9359 @item max-gcse-insertion-ratio
9360 If the ratio of expression insertions to deletions is larger than this value
9361 for any expression, then RTL PRE inserts or removes the expression and thus
9362 leaves partially redundant computations in the instruction stream. The default value is 20.
9364 @item max-pending-list-length
9365 The maximum number of pending dependencies scheduling allows
9366 before flushing the current state and starting over. Large functions
9367 with few branches or calls can create excessively large lists which
9368 needlessly consume memory and resources.
9370 @item max-modulo-backtrack-attempts
9371 The maximum number of backtrack attempts the scheduler should make
9372 when modulo scheduling a loop. Larger values can exponentially increase
9375 @item max-inline-insns-single
9376 Several parameters control the tree inliner used in GCC@.
9377 This number sets the maximum number of instructions (counted in GCC's
9378 internal representation) in a single function that the tree inliner
9379 considers for inlining. This only affects functions declared
9380 inline and methods implemented in a class declaration (C++).
9381 The default value is 400.
9383 @item max-inline-insns-auto
9384 When you use @option{-finline-functions} (included in @option{-O3}),
9385 a lot of functions that would otherwise not be considered for inlining
9386 by the compiler are investigated. To those functions, a different
9387 (more restrictive) limit compared to functions declared inline can
9389 The default value is 40.
9391 @item inline-min-speedup
9392 When estimated performance improvement of caller + callee runtime exceeds this
9393 threshold (in precent), the function can be inlined regardless the limit on
9394 @option{--param max-inline-insns-single} and @option{--param
9395 max-inline-insns-auto}.
9397 @item large-function-insns
9398 The limit specifying really large functions. For functions larger than this
9399 limit after inlining, inlining is constrained by
9400 @option{--param large-function-growth}. This parameter is useful primarily
9401 to avoid extreme compilation time caused by non-linear algorithms used by the
9403 The default value is 2700.
9405 @item large-function-growth
9406 Specifies maximal growth of large function caused by inlining in percents.
9407 The default value is 100 which limits large function growth to 2.0 times
9410 @item large-unit-insns
9411 The limit specifying large translation unit. Growth caused by inlining of
9412 units larger than this limit is limited by @option{--param inline-unit-growth}.
9413 For small units this might be too tight.
9414 For example, consider a unit consisting of function A
9415 that is inline and B that just calls A three times. If B is small relative to
9416 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9417 large units consisting of small inlineable functions, however, the overall unit
9418 growth limit is needed to avoid exponential explosion of code size. Thus for
9419 smaller units, the size is increased to @option{--param large-unit-insns}
9420 before applying @option{--param inline-unit-growth}. The default is 10000.
9422 @item inline-unit-growth
9423 Specifies maximal overall growth of the compilation unit caused by inlining.
9424 The default value is 30 which limits unit growth to 1.3 times the original
9427 @item ipcp-unit-growth
9428 Specifies maximal overall growth of the compilation unit caused by
9429 interprocedural constant propagation. The default value is 10 which limits
9430 unit growth to 1.1 times the original size.
9432 @item large-stack-frame
9433 The limit specifying large stack frames. While inlining the algorithm is trying
9434 to not grow past this limit too much. The default value is 256 bytes.
9436 @item large-stack-frame-growth
9437 Specifies maximal growth of large stack frames caused by inlining in percents.
9438 The default value is 1000 which limits large stack frame growth to 11 times
9441 @item max-inline-insns-recursive
9442 @itemx max-inline-insns-recursive-auto
9443 Specifies the maximum number of instructions an out-of-line copy of a
9444 self-recursive inline
9445 function can grow into by performing recursive inlining.
9447 For functions declared inline, @option{--param max-inline-insns-recursive} is
9448 taken into account. For functions not declared inline, recursive inlining
9449 happens only when @option{-finline-functions} (included in @option{-O3}) is
9450 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
9451 default value is 450.
9453 @item max-inline-recursive-depth
9454 @itemx max-inline-recursive-depth-auto
9455 Specifies the maximum recursion depth used for recursive inlining.
9457 For functions declared inline, @option{--param max-inline-recursive-depth} is
9458 taken into account. For functions not declared inline, recursive inlining
9459 happens only when @option{-finline-functions} (included in @option{-O3}) is
9460 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
9463 @item min-inline-recursive-probability
9464 Recursive inlining is profitable only for function having deep recursion
9465 in average and can hurt for function having little recursion depth by
9466 increasing the prologue size or complexity of function body to other
9469 When profile feedback is available (see @option{-fprofile-generate}) the actual
9470 recursion depth can be guessed from probability that function recurses via a
9471 given call expression. This parameter limits inlining only to call expressions
9472 whose probability exceeds the given threshold (in percents).
9473 The default value is 10.
9475 @item early-inlining-insns
9476 Specify growth that the early inliner can make. In effect it increases
9477 the amount of inlining for code having a large abstraction penalty.
9478 The default value is 10.
9480 @item max-early-inliner-iterations
9481 @itemx max-early-inliner-iterations
9482 Limit of iterations of the early inliner. This basically bounds
9483 the number of nested indirect calls the early inliner can resolve.
9484 Deeper chains are still handled by late inlining.
9486 @item comdat-sharing-probability
9487 @itemx comdat-sharing-probability
9488 Probability (in percent) that C++ inline function with comdat visibility
9489 are shared across multiple compilation units. The default value is 20.
9491 @item min-vect-loop-bound
9492 The minimum number of iterations under which loops are not vectorized
9493 when @option{-ftree-vectorize} is used. The number of iterations after
9494 vectorization needs to be greater than the value specified by this option
9495 to allow vectorization. The default value is 0.
9497 @item gcse-cost-distance-ratio
9498 Scaling factor in calculation of maximum distance an expression
9499 can be moved by GCSE optimizations. This is currently supported only in the
9500 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9501 is with simple expressions, i.e., the expressions that have cost
9502 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9503 hoisting of simple expressions. The default value is 10.
9505 @item gcse-unrestricted-cost
9506 Cost, roughly measured as the cost of a single typical machine
9507 instruction, at which GCSE optimizations do not constrain
9508 the distance an expression can travel. This is currently
9509 supported only in the code hoisting pass. The lesser the cost,
9510 the more aggressive code hoisting is. Specifying 0
9511 allows all expressions to travel unrestricted distances.
9512 The default value is 3.
9514 @item max-hoist-depth
9515 The depth of search in the dominator tree for expressions to hoist.
9516 This is used to avoid quadratic behavior in hoisting algorithm.
9517 The value of 0 does not limit on the search, but may slow down compilation
9518 of huge functions. The default value is 30.
9520 @item max-tail-merge-comparisons
9521 The maximum amount of similar bbs to compare a bb with. This is used to
9522 avoid quadratic behavior in tree tail merging. The default value is 10.
9524 @item max-tail-merge-iterations
9525 The maximum amount of iterations of the pass over the function. This is used to
9526 limit compilation time in tree tail merging. The default value is 2.
9528 @item max-unrolled-insns
9529 The maximum number of instructions that a loop may have to be unrolled.
9530 If a loop is unrolled, this parameter also determines how many times
9531 the loop code is unrolled.
9533 @item max-average-unrolled-insns
9534 The maximum number of instructions biased by probabilities of their execution
9535 that a loop may have to be unrolled. If a loop is unrolled,
9536 this parameter also determines how many times the loop code is unrolled.
9538 @item max-unroll-times
9539 The maximum number of unrollings of a single loop.
9541 @item max-peeled-insns
9542 The maximum number of instructions that a loop may have to be peeled.
9543 If a loop is peeled, this parameter also determines how many times
9544 the loop code is peeled.
9546 @item max-peel-times
9547 The maximum number of peelings of a single loop.
9549 @item max-peel-branches
9550 The maximum number of branches on the hot path through the peeled sequence.
9552 @item max-completely-peeled-insns
9553 The maximum number of insns of a completely peeled loop.
9555 @item max-completely-peel-times
9556 The maximum number of iterations of a loop to be suitable for complete peeling.
9558 @item max-completely-peel-loop-nest-depth
9559 The maximum depth of a loop nest suitable for complete peeling.
9561 @item max-unswitch-insns
9562 The maximum number of insns of an unswitched loop.
9564 @item max-unswitch-level
9565 The maximum number of branches unswitched in a single loop.
9568 The minimum cost of an expensive expression in the loop invariant motion.
9570 @item iv-consider-all-candidates-bound
9571 Bound on number of candidates for induction variables, below which
9572 all candidates are considered for each use in induction variable
9573 optimizations. If there are more candidates than this,
9574 only the most relevant ones are considered to avoid quadratic time complexity.
9576 @item iv-max-considered-uses
9577 The induction variable optimizations give up on loops that contain more
9578 induction variable uses.
9580 @item iv-always-prune-cand-set-bound
9581 If the number of candidates in the set is smaller than this value,
9582 always try to remove unnecessary ivs from the set
9583 when adding a new one.
9585 @item scev-max-expr-size
9586 Bound on size of expressions used in the scalar evolutions analyzer.
9587 Large expressions slow the analyzer.
9589 @item scev-max-expr-complexity
9590 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9591 Complex expressions slow the analyzer.
9593 @item omega-max-vars
9594 The maximum number of variables in an Omega constraint system.
9595 The default value is 128.
9597 @item omega-max-geqs
9598 The maximum number of inequalities in an Omega constraint system.
9599 The default value is 256.
9602 The maximum number of equalities in an Omega constraint system.
9603 The default value is 128.
9605 @item omega-max-wild-cards
9606 The maximum number of wildcard variables that the Omega solver is
9607 able to insert. The default value is 18.
9609 @item omega-hash-table-size
9610 The size of the hash table in the Omega solver. The default value is
9613 @item omega-max-keys
9614 The maximal number of keys used by the Omega solver. The default
9617 @item omega-eliminate-redundant-constraints
9618 When set to 1, use expensive methods to eliminate all redundant
9619 constraints. The default value is 0.
9621 @item vect-max-version-for-alignment-checks
9622 The maximum number of run-time checks that can be performed when
9623 doing loop versioning for alignment in the vectorizer.
9625 @item vect-max-version-for-alias-checks
9626 The maximum number of run-time checks that can be performed when
9627 doing loop versioning for alias in the vectorizer.
9629 @item vect-max-peeling-for-alignment
9630 The maximum number of loop peels to enhance access alignment
9631 for vectorizer. Value -1 means 'no limit'.
9633 @item max-iterations-to-track
9634 The maximum number of iterations of a loop the brute-force algorithm
9635 for analysis of the number of iterations of the loop tries to evaluate.
9637 @item hot-bb-count-ws-permille
9638 A basic block profile count is considered hot if it contributes to
9639 the given permillage (i.e. 0...1000) of the entire profiled execution.
9641 @item hot-bb-frequency-fraction
9642 Select fraction of the entry block frequency of executions of basic block in
9643 function given basic block needs to have to be considered hot.
9645 @item max-predicted-iterations
9646 The maximum number of loop iterations we predict statically. This is useful
9647 in cases where a function contains a single loop with known bound and
9648 another loop with unknown bound.
9649 The known number of iterations is predicted correctly, while
9650 the unknown number of iterations average to roughly 10. This means that the
9651 loop without bounds appears artificially cold relative to the other one.
9653 @item builtin-expect-probability
9654 Control the probability of the expression having the specified value. This
9655 parameter takes a percentage (i.e. 0 ... 100) as input.
9656 The default probability of 90 is obtained empirically.
9658 @item align-threshold
9660 Select fraction of the maximal frequency of executions of a basic block in
9661 a function to align the basic block.
9663 @item align-loop-iterations
9665 A loop expected to iterate at least the selected number of iterations is
9668 @item tracer-dynamic-coverage
9669 @itemx tracer-dynamic-coverage-feedback
9671 This value is used to limit superblock formation once the given percentage of
9672 executed instructions is covered. This limits unnecessary code size
9675 The @option{tracer-dynamic-coverage-feedback} is used only when profile
9676 feedback is available. The real profiles (as opposed to statically estimated
9677 ones) are much less balanced allowing the threshold to be larger value.
9679 @item tracer-max-code-growth
9680 Stop tail duplication once code growth has reached given percentage. This is
9681 a rather artificial limit, as most of the duplicates are eliminated later in
9682 cross jumping, so it may be set to much higher values than is the desired code
9685 @item tracer-min-branch-ratio
9687 Stop reverse growth when the reverse probability of best edge is less than this
9688 threshold (in percent).
9690 @item tracer-min-branch-ratio
9691 @itemx tracer-min-branch-ratio-feedback
9693 Stop forward growth if the best edge has probability lower than this
9696 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
9697 compilation for profile feedback and one for compilation without. The value
9698 for compilation with profile feedback needs to be more conservative (higher) in
9699 order to make tracer effective.
9701 @item max-cse-path-length
9703 The maximum number of basic blocks on path that CSE considers.
9707 The maximum number of instructions CSE processes before flushing.
9708 The default is 1000.
9710 @item ggc-min-expand
9712 GCC uses a garbage collector to manage its own memory allocation. This
9713 parameter specifies the minimum percentage by which the garbage
9714 collector's heap should be allowed to expand between collections.
9715 Tuning this may improve compilation speed; it has no effect on code
9718 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9719 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9720 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9721 GCC is not able to calculate RAM on a particular platform, the lower
9722 bound of 30% is used. Setting this parameter and
9723 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9724 every opportunity. This is extremely slow, but can be useful for
9727 @item ggc-min-heapsize
9729 Minimum size of the garbage collector's heap before it begins bothering
9730 to collect garbage. The first collection occurs after the heap expands
9731 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9732 tuning this may improve compilation speed, and has no effect on code
9735 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9736 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9737 with a lower bound of 4096 (four megabytes) and an upper bound of
9738 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9739 particular platform, the lower bound is used. Setting this parameter
9740 very large effectively disables garbage collection. Setting this
9741 parameter and @option{ggc-min-expand} to zero causes a full collection
9742 to occur at every opportunity.
9744 @item max-reload-search-insns
9745 The maximum number of instruction reload should look backward for equivalent
9746 register. Increasing values mean more aggressive optimization, making the
9747 compilation time increase with probably slightly better performance.
9748 The default value is 100.
9750 @item max-cselib-memory-locations
9751 The maximum number of memory locations cselib should take into account.
9752 Increasing values mean more aggressive optimization, making the compilation time
9753 increase with probably slightly better performance. The default value is 500.
9755 @item reorder-blocks-duplicate
9756 @itemx reorder-blocks-duplicate-feedback
9758 Used by the basic block reordering pass to decide whether to use unconditional
9759 branch or duplicate the code on its destination. Code is duplicated when its
9760 estimated size is smaller than this value multiplied by the estimated size of
9761 unconditional jump in the hot spots of the program.
9763 The @option{reorder-block-duplicate-feedback} is used only when profile
9764 feedback is available. It may be set to higher values than
9765 @option{reorder-block-duplicate} since information about the hot spots is more
9768 @item max-sched-ready-insns
9769 The maximum number of instructions ready to be issued the scheduler should
9770 consider at any given time during the first scheduling pass. Increasing
9771 values mean more thorough searches, making the compilation time increase
9772 with probably little benefit. The default value is 100.
9774 @item max-sched-region-blocks
9775 The maximum number of blocks in a region to be considered for
9776 interblock scheduling. The default value is 10.
9778 @item max-pipeline-region-blocks
9779 The maximum number of blocks in a region to be considered for
9780 pipelining in the selective scheduler. The default value is 15.
9782 @item max-sched-region-insns
9783 The maximum number of insns in a region to be considered for
9784 interblock scheduling. The default value is 100.
9786 @item max-pipeline-region-insns
9787 The maximum number of insns in a region to be considered for
9788 pipelining in the selective scheduler. The default value is 200.
9791 The minimum probability (in percents) of reaching a source block
9792 for interblock speculative scheduling. The default value is 40.
9794 @item max-sched-extend-regions-iters
9795 The maximum number of iterations through CFG to extend regions.
9796 A value of 0 (the default) disables region extensions.
9798 @item max-sched-insn-conflict-delay
9799 The maximum conflict delay for an insn to be considered for speculative motion.
9800 The default value is 3.
9802 @item sched-spec-prob-cutoff
9803 The minimal probability of speculation success (in percents), so that
9804 speculative insns are scheduled.
9805 The default value is 40.
9807 @item sched-spec-state-edge-prob-cutoff
9808 The minimum probability an edge must have for the scheduler to save its
9810 The default value is 10.
9812 @item sched-mem-true-dep-cost
9813 Minimal distance (in CPU cycles) between store and load targeting same
9814 memory locations. The default value is 1.
9816 @item selsched-max-lookahead
9817 The maximum size of the lookahead window of selective scheduling. It is a
9818 depth of search for available instructions.
9819 The default value is 50.
9821 @item selsched-max-sched-times
9822 The maximum number of times that an instruction is scheduled during
9823 selective scheduling. This is the limit on the number of iterations
9824 through which the instruction may be pipelined. The default value is 2.
9826 @item selsched-max-insns-to-rename
9827 The maximum number of best instructions in the ready list that are considered
9828 for renaming in the selective scheduler. The default value is 2.
9831 The minimum value of stage count that swing modulo scheduler
9832 generates. The default value is 2.
9834 @item max-last-value-rtl
9835 The maximum size measured as number of RTLs that can be recorded in an expression
9836 in combiner for a pseudo register as last known value of that register. The default
9839 @item integer-share-limit
9840 Small integer constants can use a shared data structure, reducing the
9841 compiler's memory usage and increasing its speed. This sets the maximum
9842 value of a shared integer constant. The default value is 256.
9844 @item ssp-buffer-size
9845 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
9846 protection when @option{-fstack-protection} is used.
9848 @item min-size-for-stack-sharing
9849 The minimum size of variables taking part in stack slot sharing when not
9850 optimizing. The default value is 32.
9852 @item max-jump-thread-duplication-stmts
9853 Maximum number of statements allowed in a block that needs to be
9854 duplicated when threading jumps.
9856 @item max-fields-for-field-sensitive
9857 Maximum number of fields in a structure treated in
9858 a field sensitive manner during pointer analysis. The default is zero
9859 for @option{-O0} and @option{-O1},
9860 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
9862 @item prefetch-latency
9863 Estimate on average number of instructions that are executed before
9864 prefetch finishes. The distance prefetched ahead is proportional
9865 to this constant. Increasing this number may also lead to less
9866 streams being prefetched (see @option{simultaneous-prefetches}).
9868 @item simultaneous-prefetches
9869 Maximum number of prefetches that can run at the same time.
9871 @item l1-cache-line-size
9872 The size of cache line in L1 cache, in bytes.
9875 The size of L1 cache, in kilobytes.
9878 The size of L2 cache, in kilobytes.
9880 @item min-insn-to-prefetch-ratio
9881 The minimum ratio between the number of instructions and the
9882 number of prefetches to enable prefetching in a loop.
9884 @item prefetch-min-insn-to-mem-ratio
9885 The minimum ratio between the number of instructions and the
9886 number of memory references to enable prefetching in a loop.
9888 @item use-canonical-types
9889 Whether the compiler should use the ``canonical'' type system. By
9890 default, this should always be 1, which uses a more efficient internal
9891 mechanism for comparing types in C++ and Objective-C++. However, if
9892 bugs in the canonical type system are causing compilation failures,
9893 set this value to 0 to disable canonical types.
9895 @item switch-conversion-max-branch-ratio
9896 Switch initialization conversion refuses to create arrays that are
9897 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9898 branches in the switch.
9900 @item max-partial-antic-length
9901 Maximum length of the partial antic set computed during the tree
9902 partial redundancy elimination optimization (@option{-ftree-pre}) when
9903 optimizing at @option{-O3} and above. For some sorts of source code
9904 the enhanced partial redundancy elimination optimization can run away,
9905 consuming all of the memory available on the host machine. This
9906 parameter sets a limit on the length of the sets that are computed,
9907 which prevents the runaway behavior. Setting a value of 0 for
9908 this parameter allows an unlimited set length.
9910 @item sccvn-max-scc-size
9911 Maximum size of a strongly connected component (SCC) during SCCVN
9912 processing. If this limit is hit, SCCVN processing for the whole
9913 function is not done and optimizations depending on it are
9914 disabled. The default maximum SCC size is 10000.
9916 @item sccvn-max-alias-queries-per-access
9917 Maximum number of alias-oracle queries we perform when looking for
9918 redundancies for loads and stores. If this limit is hit the search
9919 is aborted and the load or store is not considered redundant. The
9920 number of queries is algorithmically limited to the number of
9921 stores on all paths from the load to the function entry.
9922 The default maxmimum number of queries is 1000.
9924 @item ira-max-loops-num
9925 IRA uses regional register allocation by default. If a function
9926 contains more loops than the number given by this parameter, only at most
9927 the given number of the most frequently-executed loops form regions
9928 for regional register allocation. The default value of the
9931 @item ira-max-conflict-table-size
9932 Although IRA uses a sophisticated algorithm to compress the conflict
9933 table, the table can still require excessive amounts of memory for
9934 huge functions. If the conflict table for a function could be more
9935 than the size in MB given by this parameter, the register allocator
9936 instead uses a faster, simpler, and lower-quality
9937 algorithm that does not require building a pseudo-register conflict table.
9938 The default value of the parameter is 2000.
9940 @item ira-loop-reserved-regs
9941 IRA can be used to evaluate more accurate register pressure in loops
9942 for decisions to move loop invariants (see @option{-O3}). The number
9943 of available registers reserved for some other purposes is given
9944 by this parameter. The default value of the parameter is 2, which is
9945 the minimal number of registers needed by typical instructions.
9946 This value is the best found from numerous experiments.
9948 @item loop-invariant-max-bbs-in-loop
9949 Loop invariant motion can be very expensive, both in compilation time and
9950 in amount of needed compile-time memory, with very large loops. Loops
9951 with more basic blocks than this parameter won't have loop invariant
9952 motion optimization performed on them. The default value of the
9953 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
9955 @item loop-max-datarefs-for-datadeps
9956 Building data dapendencies is expensive for very large loops. This
9957 parameter limits the number of data references in loops that are
9958 considered for data dependence analysis. These large loops are no
9959 handled by the optimizations using loop data dependencies.
9960 The default value is 1000.
9962 @item max-vartrack-size
9963 Sets a maximum number of hash table slots to use during variable
9964 tracking dataflow analysis of any function. If this limit is exceeded
9965 with variable tracking at assignments enabled, analysis for that
9966 function is retried without it, after removing all debug insns from
9967 the function. If the limit is exceeded even without debug insns, var
9968 tracking analysis is completely disabled for the function. Setting
9969 the parameter to zero makes it unlimited.
9971 @item max-vartrack-expr-depth
9972 Sets a maximum number of recursion levels when attempting to map
9973 variable names or debug temporaries to value expressions. This trades
9974 compilation time for more complete debug information. If this is set too
9975 low, value expressions that are available and could be represented in
9976 debug information may end up not being used; setting this higher may
9977 enable the compiler to find more complex debug expressions, but compile
9978 time and memory use may grow. The default is 12.
9980 @item min-nondebug-insn-uid
9981 Use uids starting at this parameter for nondebug insns. The range below
9982 the parameter is reserved exclusively for debug insns created by
9983 @option{-fvar-tracking-assignments}, but debug insns may get
9984 (non-overlapping) uids above it if the reserved range is exhausted.
9986 @item ipa-sra-ptr-growth-factor
9987 IPA-SRA replaces a pointer to an aggregate with one or more new
9988 parameters only when their cumulative size is less or equal to
9989 @option{ipa-sra-ptr-growth-factor} times the size of the original
9992 @item tm-max-aggregate-size
9993 When making copies of thread-local variables in a transaction, this
9994 parameter specifies the size in bytes after which variables are
9995 saved with the logging functions as opposed to save/restore code
9996 sequence pairs. This option only applies when using
9999 @item graphite-max-nb-scop-params
10000 To avoid exponential effects in the Graphite loop transforms, the
10001 number of parameters in a Static Control Part (SCoP) is bounded. The
10002 default value is 10 parameters. A variable whose value is unknown at
10003 compilation time and defined outside a SCoP is a parameter of the SCoP.
10005 @item graphite-max-bbs-per-function
10006 To avoid exponential effects in the detection of SCoPs, the size of
10007 the functions analyzed by Graphite is bounded. The default value is
10010 @item loop-block-tile-size
10011 Loop blocking or strip mining transforms, enabled with
10012 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10013 loop in the loop nest by a given number of iterations. The strip
10014 length can be changed using the @option{loop-block-tile-size}
10015 parameter. The default value is 51 iterations.
10017 @item ipa-cp-value-list-size
10018 IPA-CP attempts to track all possible values and types passed to a function's
10019 parameter in order to propagate them and perform devirtualization.
10020 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10021 stores per one formal parameter of a function.
10023 @item lto-partitions
10024 Specify desired number of partitions produced during WHOPR compilation.
10025 The number of partitions should exceed the number of CPUs used for compilation.
10026 The default value is 32.
10028 @item lto-minpartition
10029 Size of minimal partition for WHOPR (in estimated instructions).
10030 This prevents expenses of splitting very small programs into too many
10033 @item cxx-max-namespaces-for-diagnostic-help
10034 The maximum number of namespaces to consult for suggestions when C++
10035 name lookup fails for an identifier. The default is 1000.
10037 @item sink-frequency-threshold
10038 The maximum relative execution frequency (in percents) of the target block
10039 relative to a statement's original block to allow statement sinking of a
10040 statement. Larger numbers result in more aggressive statement sinking.
10041 The default value is 75. A small positive adjustment is applied for
10042 statements with memory operands as those are even more profitable so sink.
10044 @item max-stores-to-sink
10045 The maximum number of conditional stores paires that can be sunk. Set to 0
10046 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10047 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10049 @item allow-load-data-races
10050 Allow optimizers to introduce new data races on loads.
10051 Set to 1 to allow, otherwise to 0. This option is enabled by default
10052 unless implicitly set by the @option{-fmemory-model=} option.
10054 @item allow-store-data-races
10055 Allow optimizers to introduce new data races on stores.
10056 Set to 1 to allow, otherwise to 0. This option is enabled by default
10057 unless implicitly set by the @option{-fmemory-model=} option.
10059 @item allow-packed-load-data-races
10060 Allow optimizers to introduce new data races on packed data loads.
10061 Set to 1 to allow, otherwise to 0. This option is enabled by default
10062 unless implicitly set by the @option{-fmemory-model=} option.
10064 @item allow-packed-store-data-races
10065 Allow optimizers to introduce new data races on packed data stores.
10066 Set to 1 to allow, otherwise to 0. This option is enabled by default
10067 unless implicitly set by the @option{-fmemory-model=} option.
10069 @item case-values-threshold
10070 The smallest number of different values for which it is best to use a
10071 jump-table instead of a tree of conditional branches. If the value is
10072 0, use the default for the machine. The default is 0.
10074 @item tree-reassoc-width
10075 Set the maximum number of instructions executed in parallel in
10076 reassociated tree. This parameter overrides target dependent
10077 heuristics used by default if has non zero value.
10079 @item sched-pressure-algorithm
10080 Choose between the two available implementations of
10081 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10082 and is the more likely to prevent instructions from being reordered.
10083 Algorithm 2 was designed to be a compromise between the relatively
10084 conservative approach taken by algorithm 1 and the rather aggressive
10085 approach taken by the default scheduler. It relies more heavily on
10086 having a regular register file and accurate register pressure classes.
10087 See @file{haifa-sched.c} in the GCC sources for more details.
10089 The default choice depends on the target.
10091 @item max-slsr-cand-scan
10092 Set the maximum number of existing candidates that will be considered when
10093 seeking a basis for a new straight-line strength reduction candidate.
10096 Enable buffer overflow detection for global objects. This kind
10097 of protection is enabled by default if you are using
10098 @option{-fsanitize=address} option.
10099 To disable global objects protection use @option{--param asan-globals=0}.
10102 Enable buffer overflow detection for stack objects. This kind of
10103 protection is enabled by default when using@option{-fsanitize=address}.
10104 To disable stack protection use @option{--param asan-stack=0} option.
10106 @item asan-instrument-reads
10107 Enable buffer overflow detection for memory reads. This kind of
10108 protection is enabled by default when using @option{-fsanitize=address}.
10109 To disable memory reads protection use
10110 @option{--param asan-instrument-reads=0}.
10112 @item asan-instrument-writes
10113 Enable buffer overflow detection for memory writes. This kind of
10114 protection is enabled by default when using @option{-fsanitize=address}.
10115 To disable memory writes protection use
10116 @option{--param asan-instrument-writes=0} option.
10118 @item asan-memintrin
10119 Enable detection for built-in functions. This kind of protection
10120 is enabled by default when using @option{-fsanitize=address}.
10121 To disable built-in functions protection use
10122 @option{--param asan-memintrin=0}.
10124 @item asan-use-after-return
10125 Enable detection of use-after-return. This kind of protection
10126 is enabled by default when using @option{-fsanitize=address} option.
10127 To disable use-after-return detection use
10128 @option{--param asan-use-after-return=0}.
10133 @node Preprocessor Options
10134 @section Options Controlling the Preprocessor
10135 @cindex preprocessor options
10136 @cindex options, preprocessor
10138 These options control the C preprocessor, which is run on each C source
10139 file before actual compilation.
10141 If you use the @option{-E} option, nothing is done except preprocessing.
10142 Some of these options make sense only together with @option{-E} because
10143 they cause the preprocessor output to be unsuitable for actual
10147 @item -Wp,@var{option}
10149 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10150 and pass @var{option} directly through to the preprocessor. If
10151 @var{option} contains commas, it is split into multiple options at the
10152 commas. However, many options are modified, translated or interpreted
10153 by the compiler driver before being passed to the preprocessor, and
10154 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10155 interface is undocumented and subject to change, so whenever possible
10156 you should avoid using @option{-Wp} and let the driver handle the
10159 @item -Xpreprocessor @var{option}
10160 @opindex Xpreprocessor
10161 Pass @var{option} as an option to the preprocessor. You can use this to
10162 supply system-specific preprocessor options that GCC does not
10165 If you want to pass an option that takes an argument, you must use
10166 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10168 @item -no-integrated-cpp
10169 @opindex no-integrated-cpp
10170 Perform preprocessing as a separate pass before compilation.
10171 By default, GCC performs preprocessing as an integrated part of
10172 input tokenization and parsing.
10173 If this option is provided, the appropriate language front end
10174 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10175 and Objective-C, respectively) is instead invoked twice,
10176 once for preprocessing only and once for actual compilation
10177 of the preprocessed input.
10178 This option may be useful in conjunction with the @option{-B} or
10179 @option{-wrapper} options to specify an alternate preprocessor or
10180 perform additional processing of the program source between
10181 normal preprocessing and compilation.
10184 @include cppopts.texi
10186 @node Assembler Options
10187 @section Passing Options to the Assembler
10189 @c prevent bad page break with this line
10190 You can pass options to the assembler.
10193 @item -Wa,@var{option}
10195 Pass @var{option} as an option to the assembler. If @var{option}
10196 contains commas, it is split into multiple options at the commas.
10198 @item -Xassembler @var{option}
10199 @opindex Xassembler
10200 Pass @var{option} as an option to the assembler. You can use this to
10201 supply system-specific assembler options that GCC does not
10204 If you want to pass an option that takes an argument, you must use
10205 @option{-Xassembler} twice, once for the option and once for the argument.
10210 @section Options for Linking
10211 @cindex link options
10212 @cindex options, linking
10214 These options come into play when the compiler links object files into
10215 an executable output file. They are meaningless if the compiler is
10216 not doing a link step.
10220 @item @var{object-file-name}
10221 A file name that does not end in a special recognized suffix is
10222 considered to name an object file or library. (Object files are
10223 distinguished from libraries by the linker according to the file
10224 contents.) If linking is done, these object files are used as input
10233 If any of these options is used, then the linker is not run, and
10234 object file names should not be used as arguments. @xref{Overall
10238 @item -l@var{library}
10239 @itemx -l @var{library}
10241 Search the library named @var{library} when linking. (The second
10242 alternative with the library as a separate argument is only for
10243 POSIX compliance and is not recommended.)
10245 It makes a difference where in the command you write this option; the
10246 linker searches and processes libraries and object files in the order they
10247 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10248 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10249 to functions in @samp{z}, those functions may not be loaded.
10251 The linker searches a standard list of directories for the library,
10252 which is actually a file named @file{lib@var{library}.a}. The linker
10253 then uses this file as if it had been specified precisely by name.
10255 The directories searched include several standard system directories
10256 plus any that you specify with @option{-L}.
10258 Normally the files found this way are library files---archive files
10259 whose members are object files. The linker handles an archive file by
10260 scanning through it for members which define symbols that have so far
10261 been referenced but not defined. But if the file that is found is an
10262 ordinary object file, it is linked in the usual fashion. The only
10263 difference between using an @option{-l} option and specifying a file name
10264 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10265 and searches several directories.
10269 You need this special case of the @option{-l} option in order to
10270 link an Objective-C or Objective-C++ program.
10272 @item -nostartfiles
10273 @opindex nostartfiles
10274 Do not use the standard system startup files when linking.
10275 The standard system libraries are used normally, unless @option{-nostdlib}
10276 or @option{-nodefaultlibs} is used.
10278 @item -nodefaultlibs
10279 @opindex nodefaultlibs
10280 Do not use the standard system libraries when linking.
10281 Only the libraries you specify are passed to the linker, and options
10282 specifying linkage of the system libraries, such as @code{-static-libgcc}
10283 or @code{-shared-libgcc}, are ignored.
10284 The standard startup files are used normally, unless @option{-nostartfiles}
10287 The compiler may generate calls to @code{memcmp},
10288 @code{memset}, @code{memcpy} and @code{memmove}.
10289 These entries are usually resolved by entries in
10290 libc. These entry points should be supplied through some other
10291 mechanism when this option is specified.
10295 Do not use the standard system startup files or libraries when linking.
10296 No startup files and only the libraries you specify are passed to
10297 the linker, and options specifying linkage of the system libraries, such as
10298 @code{-static-libgcc} or @code{-shared-libgcc}, are ignored.
10300 The compiler may generate calls to @code{memcmp}, @code{memset},
10301 @code{memcpy} and @code{memmove}.
10302 These entries are usually resolved by entries in
10303 libc. These entry points should be supplied through some other
10304 mechanism when this option is specified.
10306 @cindex @option{-lgcc}, use with @option{-nostdlib}
10307 @cindex @option{-nostdlib} and unresolved references
10308 @cindex unresolved references and @option{-nostdlib}
10309 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10310 @cindex @option{-nodefaultlibs} and unresolved references
10311 @cindex unresolved references and @option{-nodefaultlibs}
10312 One of the standard libraries bypassed by @option{-nostdlib} and
10313 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10314 which GCC uses to overcome shortcomings of particular machines, or special
10315 needs for some languages.
10316 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10317 Collection (GCC) Internals},
10318 for more discussion of @file{libgcc.a}.)
10319 In most cases, you need @file{libgcc.a} even when you want to avoid
10320 other standard libraries. In other words, when you specify @option{-nostdlib}
10321 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10322 This ensures that you have no unresolved references to internal GCC
10323 library subroutines.
10324 (An example of such an internal subroutine is @samp{__main}, used to ensure C++
10325 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10326 GNU Compiler Collection (GCC) Internals}.)
10330 Produce a position independent executable on targets that support it.
10331 For predictable results, you must also specify the same set of options
10332 used for compilation (@option{-fpie}, @option{-fPIE},
10333 or model suboptions) when you specify this linker option.
10337 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10338 that support it. This instructs the linker to add all symbols, not
10339 only used ones, to the dynamic symbol table. This option is needed
10340 for some uses of @code{dlopen} or to allow obtaining backtraces
10341 from within a program.
10345 Remove all symbol table and relocation information from the executable.
10349 On systems that support dynamic linking, this prevents linking with the shared
10350 libraries. On other systems, this option has no effect.
10354 Produce a shared object which can then be linked with other objects to
10355 form an executable. Not all systems support this option. For predictable
10356 results, you must also specify the same set of options used for compilation
10357 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
10358 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
10359 needs to build supplementary stub code for constructors to work. On
10360 multi-libbed systems, @samp{gcc -shared} must select the correct support
10361 libraries to link against. Failing to supply the correct flags may lead
10362 to subtle defects. Supplying them in cases where they are not necessary
10365 @item -shared-libgcc
10366 @itemx -static-libgcc
10367 @opindex shared-libgcc
10368 @opindex static-libgcc
10369 On systems that provide @file{libgcc} as a shared library, these options
10370 force the use of either the shared or static version, respectively.
10371 If no shared version of @file{libgcc} was built when the compiler was
10372 configured, these options have no effect.
10374 There are several situations in which an application should use the
10375 shared @file{libgcc} instead of the static version. The most common
10376 of these is when the application wishes to throw and catch exceptions
10377 across different shared libraries. In that case, each of the libraries
10378 as well as the application itself should use the shared @file{libgcc}.
10380 Therefore, the G++ and GCJ drivers automatically add
10381 @option{-shared-libgcc} whenever you build a shared library or a main
10382 executable, because C++ and Java programs typically use exceptions, so
10383 this is the right thing to do.
10385 If, instead, you use the GCC driver to create shared libraries, you may
10386 find that they are not always linked with the shared @file{libgcc}.
10387 If GCC finds, at its configuration time, that you have a non-GNU linker
10388 or a GNU linker that does not support option @option{--eh-frame-hdr},
10389 it links the shared version of @file{libgcc} into shared libraries
10390 by default. Otherwise, it takes advantage of the linker and optimizes
10391 away the linking with the shared version of @file{libgcc}, linking with
10392 the static version of libgcc by default. This allows exceptions to
10393 propagate through such shared libraries, without incurring relocation
10394 costs at library load time.
10396 However, if a library or main executable is supposed to throw or catch
10397 exceptions, you must link it using the G++ or GCJ driver, as appropriate
10398 for the languages used in the program, or using the option
10399 @option{-shared-libgcc}, such that it is linked with the shared
10402 @item -static-libasan
10403 @opindex static-libasan
10404 When the @option{-fsanitize=address} option is used to link a program,
10405 the GCC driver automatically links against @option{libasan}. If
10406 @file{libasan} is available as a shared library, and the @option{-static}
10407 option is not used, then this links against the shared version of
10408 @file{libasan}. The @option{-static-libasan} option directs the GCC
10409 driver to link @file{libasan} statically, without necessarily linking
10410 other libraries statically.
10412 @item -static-libtsan
10413 @opindex static-libtsan
10414 When the @option{-fsanitize=thread} option is used to link a program,
10415 the GCC driver automatically links against @option{libtsan}. If
10416 @file{libtsan} is available as a shared library, and the @option{-static}
10417 option is not used, then this links against the shared version of
10418 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
10419 driver to link @file{libtsan} statically, without necessarily linking
10420 other libraries statically.
10422 @item -static-liblsan
10423 @opindex static-liblsan
10424 When the @option{-fsanitize=leak} option is used to link a program,
10425 the GCC driver automatically links against @option{liblsan}. If
10426 @file{liblsan} is available as a shared library, and the @option{-static}
10427 option is not used, then this links against the shared version of
10428 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
10429 driver to link @file{liblsan} statically, without necessarily linking
10430 other libraries statically.
10432 @item -static-libubsan
10433 @opindex static-libubsan
10434 When the @option{-fsanitize=undefined} option is used to link a program,
10435 the GCC driver automatically links against @option{libubsan}. If
10436 @file{libubsan} is available as a shared library, and the @option{-static}
10437 option is not used, then this links against the shared version of
10438 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
10439 driver to link @file{libubsan} statically, without necessarily linking
10440 other libraries statically.
10442 @item -static-libstdc++
10443 @opindex static-libstdc++
10444 When the @command{g++} program is used to link a C++ program, it
10445 normally automatically links against @option{libstdc++}. If
10446 @file{libstdc++} is available as a shared library, and the
10447 @option{-static} option is not used, then this links against the
10448 shared version of @file{libstdc++}. That is normally fine. However, it
10449 is sometimes useful to freeze the version of @file{libstdc++} used by
10450 the program without going all the way to a fully static link. The
10451 @option{-static-libstdc++} option directs the @command{g++} driver to
10452 link @file{libstdc++} statically, without necessarily linking other
10453 libraries statically.
10457 Bind references to global symbols when building a shared object. Warn
10458 about any unresolved references (unless overridden by the link editor
10459 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
10462 @item -T @var{script}
10464 @cindex linker script
10465 Use @var{script} as the linker script. This option is supported by most
10466 systems using the GNU linker. On some targets, such as bare-board
10467 targets without an operating system, the @option{-T} option may be required
10468 when linking to avoid references to undefined symbols.
10470 @item -Xlinker @var{option}
10472 Pass @var{option} as an option to the linker. You can use this to
10473 supply system-specific linker options that GCC does not recognize.
10475 If you want to pass an option that takes a separate argument, you must use
10476 @option{-Xlinker} twice, once for the option and once for the argument.
10477 For example, to pass @option{-assert definitions}, you must write
10478 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
10479 @option{-Xlinker "-assert definitions"}, because this passes the entire
10480 string as a single argument, which is not what the linker expects.
10482 When using the GNU linker, it is usually more convenient to pass
10483 arguments to linker options using the @option{@var{option}=@var{value}}
10484 syntax than as separate arguments. For example, you can specify
10485 @option{-Xlinker -Map=output.map} rather than
10486 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
10487 this syntax for command-line options.
10489 @item -Wl,@var{option}
10491 Pass @var{option} as an option to the linker. If @var{option} contains
10492 commas, it is split into multiple options at the commas. You can use this
10493 syntax to pass an argument to the option.
10494 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
10495 linker. When using the GNU linker, you can also get the same effect with
10496 @option{-Wl,-Map=output.map}.
10498 @item -u @var{symbol}
10500 Pretend the symbol @var{symbol} is undefined, to force linking of
10501 library modules to define it. You can use @option{-u} multiple times with
10502 different symbols to force loading of additional library modules.
10505 @node Directory Options
10506 @section Options for Directory Search
10507 @cindex directory options
10508 @cindex options, directory search
10509 @cindex search path
10511 These options specify directories to search for header files, for
10512 libraries and for parts of the compiler:
10517 Add the directory @var{dir} to the head of the list of directories to be
10518 searched for header files. This can be used to override a system header
10519 file, substituting your own version, since these directories are
10520 searched before the system header file directories. However, you should
10521 not use this option to add directories that contain vendor-supplied
10522 system header files (use @option{-isystem} for that). If you use more than
10523 one @option{-I} option, the directories are scanned in left-to-right
10524 order; the standard system directories come after.
10526 If a standard system include directory, or a directory specified with
10527 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
10528 option is ignored. The directory is still searched but as a
10529 system directory at its normal position in the system include chain.
10530 This is to ensure that GCC's procedure to fix buggy system headers and
10531 the ordering for the @code{include_next} directive are not inadvertently changed.
10532 If you really need to change the search order for system directories,
10533 use the @option{-nostdinc} and/or @option{-isystem} options.
10535 @item -iplugindir=@var{dir}
10536 @opindex iplugindir=
10537 Set the directory to search for plugins that are passed
10538 by @option{-fplugin=@var{name}} instead of
10539 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
10540 to be used by the user, but only passed by the driver.
10542 @item -iquote@var{dir}
10544 Add the directory @var{dir} to the head of the list of directories to
10545 be searched for header files only for the case of @samp{#include
10546 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
10547 otherwise just like @option{-I}.
10551 Add directory @var{dir} to the list of directories to be searched
10554 @item -B@var{prefix}
10556 This option specifies where to find the executables, libraries,
10557 include files, and data files of the compiler itself.
10559 The compiler driver program runs one or more of the subprograms
10560 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
10561 @var{prefix} as a prefix for each program it tries to run, both with and
10562 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
10564 For each subprogram to be run, the compiler driver first tries the
10565 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
10566 is not specified, the driver tries two standard prefixes,
10567 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
10568 those results in a file name that is found, the unmodified program
10569 name is searched for using the directories specified in your
10570 @env{PATH} environment variable.
10572 The compiler checks to see if the path provided by the @option{-B}
10573 refers to a directory, and if necessary it adds a directory
10574 separator character at the end of the path.
10576 @option{-B} prefixes that effectively specify directory names also apply
10577 to libraries in the linker, because the compiler translates these
10578 options into @option{-L} options for the linker. They also apply to
10579 include files in the preprocessor, because the compiler translates these
10580 options into @option{-isystem} options for the preprocessor. In this case,
10581 the compiler appends @samp{include} to the prefix.
10583 The runtime support file @file{libgcc.a} can also be searched for using
10584 the @option{-B} prefix, if needed. If it is not found there, the two
10585 standard prefixes above are tried, and that is all. The file is left
10586 out of the link if it is not found by those means.
10588 Another way to specify a prefix much like the @option{-B} prefix is to use
10589 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
10592 As a special kludge, if the path provided by @option{-B} is
10593 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
10594 9, then it is replaced by @file{[dir/]include}. This is to help
10595 with boot-strapping the compiler.
10597 @item -specs=@var{file}
10599 Process @var{file} after the compiler reads in the standard @file{specs}
10600 file, in order to override the defaults which the @command{gcc} driver
10601 program uses when determining what switches to pass to @command{cc1},
10602 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
10603 @option{-specs=@var{file}} can be specified on the command line, and they
10604 are processed in order, from left to right.
10606 @item --sysroot=@var{dir}
10608 Use @var{dir} as the logical root directory for headers and libraries.
10609 For example, if the compiler normally searches for headers in
10610 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
10611 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
10613 If you use both this option and the @option{-isysroot} option, then
10614 the @option{--sysroot} option applies to libraries, but the
10615 @option{-isysroot} option applies to header files.
10617 The GNU linker (beginning with version 2.16) has the necessary support
10618 for this option. If your linker does not support this option, the
10619 header file aspect of @option{--sysroot} still works, but the
10620 library aspect does not.
10622 @item --no-sysroot-suffix
10623 @opindex no-sysroot-suffix
10624 For some targets, a suffix is added to the root directory specified
10625 with @option{--sysroot}, depending on the other options used, so that
10626 headers may for example be found in
10627 @file{@var{dir}/@var{suffix}/usr/include} instead of
10628 @file{@var{dir}/usr/include}. This option disables the addition of
10633 This option has been deprecated. Please use @option{-iquote} instead for
10634 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
10635 Any directories you specify with @option{-I} options before the @option{-I-}
10636 option are searched only for the case of @samp{#include "@var{file}"};
10637 they are not searched for @samp{#include <@var{file}>}.
10639 If additional directories are specified with @option{-I} options after
10640 the @option{-I-}, these directories are searched for all @samp{#include}
10641 directives. (Ordinarily @emph{all} @option{-I} directories are used
10644 In addition, the @option{-I-} option inhibits the use of the current
10645 directory (where the current input file came from) as the first search
10646 directory for @samp{#include "@var{file}"}. There is no way to
10647 override this effect of @option{-I-}. With @option{-I.} you can specify
10648 searching the directory that is current when the compiler is
10649 invoked. That is not exactly the same as what the preprocessor does
10650 by default, but it is often satisfactory.
10652 @option{-I-} does not inhibit the use of the standard system directories
10653 for header files. Thus, @option{-I-} and @option{-nostdinc} are
10660 @section Specifying subprocesses and the switches to pass to them
10663 @command{gcc} is a driver program. It performs its job by invoking a
10664 sequence of other programs to do the work of compiling, assembling and
10665 linking. GCC interprets its command-line parameters and uses these to
10666 deduce which programs it should invoke, and which command-line options
10667 it ought to place on their command lines. This behavior is controlled
10668 by @dfn{spec strings}. In most cases there is one spec string for each
10669 program that GCC can invoke, but a few programs have multiple spec
10670 strings to control their behavior. The spec strings built into GCC can
10671 be overridden by using the @option{-specs=} command-line switch to specify
10674 @dfn{Spec files} are plaintext files that are used to construct spec
10675 strings. They consist of a sequence of directives separated by blank
10676 lines. The type of directive is determined by the first non-whitespace
10677 character on the line, which can be one of the following:
10680 @item %@var{command}
10681 Issues a @var{command} to the spec file processor. The commands that can
10685 @item %include <@var{file}>
10686 @cindex @code{%include}
10687 Search for @var{file} and insert its text at the current point in the
10690 @item %include_noerr <@var{file}>
10691 @cindex @code{%include_noerr}
10692 Just like @samp{%include}, but do not generate an error message if the include
10693 file cannot be found.
10695 @item %rename @var{old_name} @var{new_name}
10696 @cindex @code{%rename}
10697 Rename the spec string @var{old_name} to @var{new_name}.
10701 @item *[@var{spec_name}]:
10702 This tells the compiler to create, override or delete the named spec
10703 string. All lines after this directive up to the next directive or
10704 blank line are considered to be the text for the spec string. If this
10705 results in an empty string then the spec is deleted. (Or, if the
10706 spec did not exist, then nothing happens.) Otherwise, if the spec
10707 does not currently exist a new spec is created. If the spec does
10708 exist then its contents are overridden by the text of this
10709 directive, unless the first character of that text is the @samp{+}
10710 character, in which case the text is appended to the spec.
10712 @item [@var{suffix}]:
10713 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
10714 and up to the next directive or blank line are considered to make up the
10715 spec string for the indicated suffix. When the compiler encounters an
10716 input file with the named suffix, it processes the spec string in
10717 order to work out how to compile that file. For example:
10721 z-compile -input %i
10724 This says that any input file whose name ends in @samp{.ZZ} should be
10725 passed to the program @samp{z-compile}, which should be invoked with the
10726 command-line switch @option{-input} and with the result of performing the
10727 @samp{%i} substitution. (See below.)
10729 As an alternative to providing a spec string, the text following a
10730 suffix directive can be one of the following:
10733 @item @@@var{language}
10734 This says that the suffix is an alias for a known @var{language}. This is
10735 similar to using the @option{-x} command-line switch to GCC to specify a
10736 language explicitly. For example:
10743 Says that .ZZ files are, in fact, C++ source files.
10746 This causes an error messages saying:
10749 @var{name} compiler not installed on this system.
10753 GCC already has an extensive list of suffixes built into it.
10754 This directive adds an entry to the end of the list of suffixes, but
10755 since the list is searched from the end backwards, it is effectively
10756 possible to override earlier entries using this technique.
10760 GCC has the following spec strings built into it. Spec files can
10761 override these strings or create their own. Note that individual
10762 targets can also add their own spec strings to this list.
10765 asm Options to pass to the assembler
10766 asm_final Options to pass to the assembler post-processor
10767 cpp Options to pass to the C preprocessor
10768 cc1 Options to pass to the C compiler
10769 cc1plus Options to pass to the C++ compiler
10770 endfile Object files to include at the end of the link
10771 link Options to pass to the linker
10772 lib Libraries to include on the command line to the linker
10773 libgcc Decides which GCC support library to pass to the linker
10774 linker Sets the name of the linker
10775 predefines Defines to be passed to the C preprocessor
10776 signed_char Defines to pass to CPP to say whether @code{char} is signed
10778 startfile Object files to include at the start of the link
10781 Here is a small example of a spec file:
10784 %rename lib old_lib
10787 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
10790 This example renames the spec called @samp{lib} to @samp{old_lib} and
10791 then overrides the previous definition of @samp{lib} with a new one.
10792 The new definition adds in some extra command-line options before
10793 including the text of the old definition.
10795 @dfn{Spec strings} are a list of command-line options to be passed to their
10796 corresponding program. In addition, the spec strings can contain
10797 @samp{%}-prefixed sequences to substitute variable text or to
10798 conditionally insert text into the command line. Using these constructs
10799 it is possible to generate quite complex command lines.
10801 Here is a table of all defined @samp{%}-sequences for spec
10802 strings. Note that spaces are not generated automatically around the
10803 results of expanding these sequences. Therefore you can concatenate them
10804 together or combine them with constant text in a single argument.
10808 Substitute one @samp{%} into the program name or argument.
10811 Substitute the name of the input file being processed.
10814 Substitute the basename of the input file being processed.
10815 This is the substring up to (and not including) the last period
10816 and not including the directory.
10819 This is the same as @samp{%b}, but include the file suffix (text after
10823 Marks the argument containing or following the @samp{%d} as a
10824 temporary file name, so that that file is deleted if GCC exits
10825 successfully. Unlike @samp{%g}, this contributes no text to the
10828 @item %g@var{suffix}
10829 Substitute a file name that has suffix @var{suffix} and is chosen
10830 once per compilation, and mark the argument in the same way as
10831 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
10832 name is now chosen in a way that is hard to predict even when previously
10833 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
10834 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
10835 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
10836 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
10837 was simply substituted with a file name chosen once per compilation,
10838 without regard to any appended suffix (which was therefore treated
10839 just like ordinary text), making such attacks more likely to succeed.
10841 @item %u@var{suffix}
10842 Like @samp{%g}, but generates a new temporary file name
10843 each time it appears instead of once per compilation.
10845 @item %U@var{suffix}
10846 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
10847 new one if there is no such last file name. In the absence of any
10848 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
10849 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
10850 involves the generation of two distinct file names, one
10851 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
10852 simply substituted with a file name chosen for the previous @samp{%u},
10853 without regard to any appended suffix.
10855 @item %j@var{suffix}
10856 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
10857 writable, and if @option{-save-temps} is not used;
10858 otherwise, substitute the name
10859 of a temporary file, just like @samp{%u}. This temporary file is not
10860 meant for communication between processes, but rather as a junk
10861 disposal mechanism.
10863 @item %|@var{suffix}
10864 @itemx %m@var{suffix}
10865 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
10866 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
10867 all. These are the two most common ways to instruct a program that it
10868 should read from standard input or write to standard output. If you
10869 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
10870 construct: see for example @file{f/lang-specs.h}.
10872 @item %.@var{SUFFIX}
10873 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
10874 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
10875 terminated by the next space or %.
10878 Marks the argument containing or following the @samp{%w} as the
10879 designated output file of this compilation. This puts the argument
10880 into the sequence of arguments that @samp{%o} substitutes.
10883 Substitutes the names of all the output files, with spaces
10884 automatically placed around them. You should write spaces
10885 around the @samp{%o} as well or the results are undefined.
10886 @samp{%o} is for use in the specs for running the linker.
10887 Input files whose names have no recognized suffix are not compiled
10888 at all, but they are included among the output files, so they are
10892 Substitutes the suffix for object files. Note that this is
10893 handled specially when it immediately follows @samp{%g, %u, or %U},
10894 because of the need for those to form complete file names. The
10895 handling is such that @samp{%O} is treated exactly as if it had already
10896 been substituted, except that @samp{%g, %u, and %U} do not currently
10897 support additional @var{suffix} characters following @samp{%O} as they do
10898 following, for example, @samp{.o}.
10901 Substitutes the standard macro predefinitions for the
10902 current target machine. Use this when running @code{cpp}.
10905 Like @samp{%p}, but puts @samp{__} before and after the name of each
10906 predefined macro, except for macros that start with @samp{__} or with
10907 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
10911 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
10912 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
10913 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
10914 and @option{-imultilib} as necessary.
10917 Current argument is the name of a library or startup file of some sort.
10918 Search for that file in a standard list of directories and substitute
10919 the full name found. The current working directory is included in the
10920 list of directories scanned.
10923 Current argument is the name of a linker script. Search for that file
10924 in the current list of directories to scan for libraries. If the file
10925 is located insert a @option{--script} option into the command line
10926 followed by the full path name found. If the file is not found then
10927 generate an error message. Note: the current working directory is not
10931 Print @var{str} as an error message. @var{str} is terminated by a newline.
10932 Use this when inconsistent options are detected.
10934 @item %(@var{name})
10935 Substitute the contents of spec string @var{name} at this point.
10937 @item %x@{@var{option}@}
10938 Accumulate an option for @samp{%X}.
10941 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
10945 Output the accumulated assembler options specified by @option{-Wa}.
10948 Output the accumulated preprocessor options specified by @option{-Wp}.
10951 Process the @code{asm} spec. This is used to compute the
10952 switches to be passed to the assembler.
10955 Process the @code{asm_final} spec. This is a spec string for
10956 passing switches to an assembler post-processor, if such a program is
10960 Process the @code{link} spec. This is the spec for computing the
10961 command line passed to the linker. Typically it makes use of the
10962 @samp{%L %G %S %D and %E} sequences.
10965 Dump out a @option{-L} option for each directory that GCC believes might
10966 contain startup files. If the target supports multilibs then the
10967 current multilib directory is prepended to each of these paths.
10970 Process the @code{lib} spec. This is a spec string for deciding which
10971 libraries are included on the command line to the linker.
10974 Process the @code{libgcc} spec. This is a spec string for deciding
10975 which GCC support library is included on the command line to the linker.
10978 Process the @code{startfile} spec. This is a spec for deciding which
10979 object files are the first ones passed to the linker. Typically
10980 this might be a file named @file{crt0.o}.
10983 Process the @code{endfile} spec. This is a spec string that specifies
10984 the last object files that are passed to the linker.
10987 Process the @code{cpp} spec. This is used to construct the arguments
10988 to be passed to the C preprocessor.
10991 Process the @code{cc1} spec. This is used to construct the options to be
10992 passed to the actual C compiler (@samp{cc1}).
10995 Process the @code{cc1plus} spec. This is used to construct the options to be
10996 passed to the actual C++ compiler (@samp{cc1plus}).
10999 Substitute the variable part of a matched option. See below.
11000 Note that each comma in the substituted string is replaced by
11004 Remove all occurrences of @code{-S} from the command line. Note---this
11005 command is position dependent. @samp{%} commands in the spec string
11006 before this one see @code{-S}, @samp{%} commands in the spec string
11007 after this one do not.
11009 @item %:@var{function}(@var{args})
11010 Call the named function @var{function}, passing it @var{args}.
11011 @var{args} is first processed as a nested spec string, then split
11012 into an argument vector in the usual fashion. The function returns
11013 a string which is processed as if it had appeared literally as part
11014 of the current spec.
11016 The following built-in spec functions are provided:
11019 @item @code{getenv}
11020 The @code{getenv} spec function takes two arguments: an environment
11021 variable name and a string. If the environment variable is not
11022 defined, a fatal error is issued. Otherwise, the return value is the
11023 value of the environment variable concatenated with the string. For
11024 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11027 %:getenv(TOPDIR /include)
11030 expands to @file{/path/to/top/include}.
11032 @item @code{if-exists}
11033 The @code{if-exists} spec function takes one argument, an absolute
11034 pathname to a file. If the file exists, @code{if-exists} returns the
11035 pathname. Here is a small example of its usage:
11039 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11042 @item @code{if-exists-else}
11043 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11044 spec function, except that it takes two arguments. The first argument is
11045 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11046 returns the pathname. If it does not exist, it returns the second argument.
11047 This way, @code{if-exists-else} can be used to select one file or another,
11048 based on the existence of the first. Here is a small example of its usage:
11052 crt0%O%s %:if-exists(crti%O%s) \
11053 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11056 @item @code{replace-outfile}
11057 The @code{replace-outfile} spec function takes two arguments. It looks for the
11058 first argument in the outfiles array and replaces it with the second argument. Here
11059 is a small example of its usage:
11062 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11065 @item @code{remove-outfile}
11066 The @code{remove-outfile} spec function takes one argument. It looks for the
11067 first argument in the outfiles array and removes it. Here is a small example
11071 %:remove-outfile(-lm)
11074 @item @code{pass-through-libs}
11075 The @code{pass-through-libs} spec function takes any number of arguments. It
11076 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11077 assumes are the names of linker input library archive files) and returns a
11078 result containing all the found arguments each prepended by
11079 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11080 intended to be passed to the LTO linker plugin.
11083 %:pass-through-libs(%G %L %G)
11086 @item @code{print-asm-header}
11087 The @code{print-asm-header} function takes no arguments and simply
11088 prints a banner like:
11094 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11097 It is used to separate compiler options from assembler options
11098 in the @option{--target-help} output.
11101 @item %@{@code{S}@}
11102 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11103 If that switch is not specified, this substitutes nothing. Note that
11104 the leading dash is omitted when specifying this option, and it is
11105 automatically inserted if the substitution is performed. Thus the spec
11106 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11107 and outputs the command-line option @option{-foo}.
11109 @item %W@{@code{S}@}
11110 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11111 deleted on failure.
11113 @item %@{@code{S}*@}
11114 Substitutes all the switches specified to GCC whose names start
11115 with @code{-S}, but which also take an argument. This is used for
11116 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11117 GCC considers @option{-o foo} as being
11118 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11119 text, including the space. Thus two arguments are generated.
11121 @item %@{@code{S}*&@code{T}*@}
11122 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11123 (the order of @code{S} and @code{T} in the spec is not significant).
11124 There can be any number of ampersand-separated variables; for each the
11125 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11127 @item %@{@code{S}:@code{X}@}
11128 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11130 @item %@{!@code{S}:@code{X}@}
11131 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11133 @item %@{@code{S}*:@code{X}@}
11134 Substitutes @code{X} if one or more switches whose names start with
11135 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11136 once, no matter how many such switches appeared. However, if @code{%*}
11137 appears somewhere in @code{X}, then @code{X} is substituted once
11138 for each matching switch, with the @code{%*} replaced by the part of
11139 that switch matching the @code{*}.
11141 If @code{%*} appears as the last part of a spec sequence then a space
11142 will be added after the end of the last substitution. If there is more
11143 text in the sequence however then a space will not be generated. This
11144 allows the @code{%*} substitution to be used as part of a larger
11145 string. For example, a spec string like this:
11148 %@{mcu=*:--script=%*/memory.ld@}
11151 when matching an option like @code{-mcu=newchip} will produce:
11154 --script=newchip/memory.ld
11157 @item %@{.@code{S}:@code{X}@}
11158 Substitutes @code{X}, if processing a file with suffix @code{S}.
11160 @item %@{!.@code{S}:@code{X}@}
11161 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
11163 @item %@{,@code{S}:@code{X}@}
11164 Substitutes @code{X}, if processing a file for language @code{S}.
11166 @item %@{!,@code{S}:@code{X}@}
11167 Substitutes @code{X}, if not processing a file for language @code{S}.
11169 @item %@{@code{S}|@code{P}:@code{X}@}
11170 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
11171 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
11172 @code{*} sequences as well, although they have a stronger binding than
11173 the @samp{|}. If @code{%*} appears in @code{X}, all of the
11174 alternatives must be starred, and only the first matching alternative
11177 For example, a spec string like this:
11180 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
11184 outputs the following command-line options from the following input
11185 command-line options:
11190 -d fred.c -foo -baz -boggle
11191 -d jim.d -bar -baz -boggle
11194 @item %@{S:X; T:Y; :D@}
11196 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
11197 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
11198 be as many clauses as you need. This may be combined with @code{.},
11199 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
11204 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
11205 construct may contain other nested @samp{%} constructs or spaces, or
11206 even newlines. They are processed as usual, as described above.
11207 Trailing white space in @code{X} is ignored. White space may also
11208 appear anywhere on the left side of the colon in these constructs,
11209 except between @code{.} or @code{*} and the corresponding word.
11211 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
11212 handled specifically in these constructs. If another value of
11213 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
11214 @option{-W} switch is found later in the command line, the earlier
11215 switch value is ignored, except with @{@code{S}*@} where @code{S} is
11216 just one letter, which passes all matching options.
11218 The character @samp{|} at the beginning of the predicate text is used to
11219 indicate that a command should be piped to the following command, but
11220 only if @option{-pipe} is specified.
11222 It is built into GCC which switches take arguments and which do not.
11223 (You might think it would be useful to generalize this to allow each
11224 compiler's spec to say which switches take arguments. But this cannot
11225 be done in a consistent fashion. GCC cannot even decide which input
11226 files have been specified without knowing which switches take arguments,
11227 and it must know which input files to compile in order to tell which
11230 GCC also knows implicitly that arguments starting in @option{-l} are to be
11231 treated as compiler output files, and passed to the linker in their
11232 proper position among the other output files.
11234 @c man begin OPTIONS
11236 @node Target Options
11237 @section Specifying Target Machine and Compiler Version
11238 @cindex target options
11239 @cindex cross compiling
11240 @cindex specifying machine version
11241 @cindex specifying compiler version and target machine
11242 @cindex compiler version, specifying
11243 @cindex target machine, specifying
11245 The usual way to run GCC is to run the executable called @command{gcc}, or
11246 @command{@var{machine}-gcc} when cross-compiling, or
11247 @command{@var{machine}-gcc-@var{version}} to run a version other than the
11248 one that was installed last.
11250 @node Submodel Options
11251 @section Hardware Models and Configurations
11252 @cindex submodel options
11253 @cindex specifying hardware config
11254 @cindex hardware models and configurations, specifying
11255 @cindex machine dependent options
11257 Each target machine types can have its own
11258 special options, starting with @samp{-m}, to choose among various
11259 hardware models or configurations---for example, 68010 vs 68020,
11260 floating coprocessor or none. A single installed version of the
11261 compiler can compile for any model or configuration, according to the
11264 Some configurations of the compiler also support additional special
11265 options, usually for compatibility with other compilers on the same
11268 @c This list is ordered alphanumerically by subsection name.
11269 @c It should be the same order and spelling as these options are listed
11270 @c in Machine Dependent Options
11273 * AArch64 Options::
11274 * Adapteva Epiphany Options::
11278 * Blackfin Options::
11283 * DEC Alpha Options::
11286 * GNU/Linux Options::
11289 * i386 and x86-64 Options::
11290 * i386 and x86-64 Windows Options::
11298 * MicroBlaze Options::
11301 * MN10300 Options::
11305 * Nios II Options::
11307 * picoChip Options::
11308 * PowerPC Options::
11310 * RS/6000 and PowerPC Options::
11312 * S/390 and zSeries Options::
11315 * Solaris 2 Options::
11318 * System V Options::
11319 * TILE-Gx Options::
11320 * TILEPro Options::
11324 * VxWorks Options::
11326 * Xstormy16 Options::
11328 * zSeries Options::
11331 @node AArch64 Options
11332 @subsection AArch64 Options
11333 @cindex AArch64 Options
11335 These options are defined for AArch64 implementations:
11339 @item -mabi=@var{name}
11341 Generate code for the specified data model. Permissible values
11342 are @samp{ilp32} for SysV-like data model where int, long int and pointer
11343 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
11344 but long int and pointer are 64-bit.
11346 The default depends on the specific target configuration. Note that
11347 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
11348 entire program with the same ABI, and link with a compatible set of libraries.
11351 @opindex mbig-endian
11352 Generate big-endian code. This is the default when GCC is configured for an
11353 @samp{aarch64_be-*-*} target.
11355 @item -mgeneral-regs-only
11356 @opindex mgeneral-regs-only
11357 Generate code which uses only the general registers.
11359 @item -mlittle-endian
11360 @opindex mlittle-endian
11361 Generate little-endian code. This is the default when GCC is configured for an
11362 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
11364 @item -mcmodel=tiny
11365 @opindex mcmodel=tiny
11366 Generate code for the tiny code model. The program and its statically defined
11367 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
11368 be statically or dynamically linked. This model is not fully implemented and
11369 mostly treated as @samp{small}.
11371 @item -mcmodel=small
11372 @opindex mcmodel=small
11373 Generate code for the small code model. The program and its statically defined
11374 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
11375 be statically or dynamically linked. This is the default code model.
11377 @item -mcmodel=large
11378 @opindex mcmodel=large
11379 Generate code for the large code model. This makes no assumptions about
11380 addresses and sizes of sections. Pointers are 64 bits. Programs can be
11381 statically linked only.
11383 @item -mstrict-align
11384 @opindex mstrict-align
11385 Do not assume that unaligned memory references will be handled by the system.
11387 @item -momit-leaf-frame-pointer
11388 @itemx -mno-omit-leaf-frame-pointer
11389 @opindex momit-leaf-frame-pointer
11390 @opindex mno-omit-leaf-frame-pointer
11391 Omit or keep the frame pointer in leaf functions. The former behaviour is the
11394 @item -mtls-dialect=desc
11395 @opindex mtls-dialect=desc
11396 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
11397 of TLS variables. This is the default.
11399 @item -mtls-dialect=traditional
11400 @opindex mtls-dialect=traditional
11401 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
11404 @item -march=@var{name}
11406 Specify the name of the target architecture, optionally suffixed by one or
11407 more feature modifiers. This option has the form
11408 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11409 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
11410 values for @var{feature} are documented in the sub-section below.
11412 Where conflicting feature modifiers are specified, the right-most feature is
11415 GCC uses this name to determine what kind of instructions it can emit when
11416 generating assembly code.
11418 Where @option{-march} is specified without either of @option{-mtune}
11419 or @option{-mcpu} also being specified, the code will be tuned to perform
11420 well across a range of target processors implementing the target
11423 @item -mtune=@var{name}
11425 Specify the name of the target processor for which GCC should tune the
11426 performance of the code. Permissible values for this option are:
11427 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}.
11429 Additionally, this option can specify that GCC should tune the performance
11430 of the code for a big.LITTLE system. The only permissible value is
11431 @samp{cortex-a57.cortex-a53}.
11433 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
11434 are specified, the code will be tuned to perform well across a range
11435 of target processors.
11437 This option cannot be suffixed by feature modifiers.
11439 @item -mcpu=@var{name}
11441 Specify the name of the target processor, optionally suffixed by one or more
11442 feature modifiers. This option has the form
11443 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11444 permissible values for @var{cpu} are the same as those available for
11447 The permissible values for @var{feature} are documented in the sub-section
11450 Where conflicting feature modifiers are specified, the right-most feature is
11453 GCC uses this name to determine what kind of instructions it can emit when
11454 generating assembly code (as if by @option{-march}) and to determine
11455 the target processor for which to tune for performance (as if
11456 by @option{-mtune}). Where this option is used in conjunction
11457 with @option{-march} or @option{-mtune}, those options take precedence
11458 over the appropriate part of this option.
11461 @subsubsection @option{-march} and @option{-mcpu} feature modifiers
11462 @cindex @option{-march} feature modifiers
11463 @cindex @option{-mcpu} feature modifiers
11464 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
11469 Enable CRC extension.
11471 Enable Crypto extension. This implies Advanced SIMD is enabled.
11473 Enable floating-point instructions.
11475 Enable Advanced SIMD instructions. This implies floating-point instructions
11476 are enabled. This is the default for all current possible values for options
11477 @option{-march} and @option{-mcpu=}.
11480 @node Adapteva Epiphany Options
11481 @subsection Adapteva Epiphany Options
11483 These @samp{-m} options are defined for Adapteva Epiphany:
11486 @item -mhalf-reg-file
11487 @opindex mhalf-reg-file
11488 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
11489 That allows code to run on hardware variants that lack these registers.
11491 @item -mprefer-short-insn-regs
11492 @opindex mprefer-short-insn-regs
11493 Preferrentially allocate registers that allow short instruction generation.
11494 This can result in increased instruction count, so this may either reduce or
11495 increase overall code size.
11497 @item -mbranch-cost=@var{num}
11498 @opindex mbranch-cost
11499 Set the cost of branches to roughly @var{num} ``simple'' instructions.
11500 This cost is only a heuristic and is not guaranteed to produce
11501 consistent results across releases.
11505 Enable the generation of conditional moves.
11507 @item -mnops=@var{num}
11509 Emit @var{num} NOPs before every other generated instruction.
11511 @item -mno-soft-cmpsf
11512 @opindex mno-soft-cmpsf
11513 For single-precision floating-point comparisons, emit an @code{fsub} instruction
11514 and test the flags. This is faster than a software comparison, but can
11515 get incorrect results in the presence of NaNs, or when two different small
11516 numbers are compared such that their difference is calculated as zero.
11517 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
11518 software comparisons.
11520 @item -mstack-offset=@var{num}
11521 @opindex mstack-offset
11522 Set the offset between the top of the stack and the stack pointer.
11523 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
11524 can be used by leaf functions without stack allocation.
11525 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
11526 Note also that this option changes the ABI; compiling a program with a
11527 different stack offset than the libraries have been compiled with
11528 generally does not work.
11529 This option can be useful if you want to evaluate if a different stack
11530 offset would give you better code, but to actually use a different stack
11531 offset to build working programs, it is recommended to configure the
11532 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
11534 @item -mno-round-nearest
11535 @opindex mno-round-nearest
11536 Make the scheduler assume that the rounding mode has been set to
11537 truncating. The default is @option{-mround-nearest}.
11540 @opindex mlong-calls
11541 If not otherwise specified by an attribute, assume all calls might be beyond
11542 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
11543 function address into a register before performing a (otherwise direct) call.
11544 This is the default.
11546 @item -mshort-calls
11547 @opindex short-calls
11548 If not otherwise specified by an attribute, assume all direct calls are
11549 in the range of the @code{b} / @code{bl} instructions, so use these instructions
11550 for direct calls. The default is @option{-mlong-calls}.
11554 Assume addresses can be loaded as 16-bit unsigned values. This does not
11555 apply to function addresses for which @option{-mlong-calls} semantics
11558 @item -mfp-mode=@var{mode}
11560 Set the prevailing mode of the floating-point unit.
11561 This determines the floating-point mode that is provided and expected
11562 at function call and return time. Making this mode match the mode you
11563 predominantly need at function start can make your programs smaller and
11564 faster by avoiding unnecessary mode switches.
11566 @var{mode} can be set to one the following values:
11570 Any mode at function entry is valid, and retained or restored when
11571 the function returns, and when it calls other functions.
11572 This mode is useful for compiling libraries or other compilation units
11573 you might want to incorporate into different programs with different
11574 prevailing FPU modes, and the convenience of being able to use a single
11575 object file outweighs the size and speed overhead for any extra
11576 mode switching that might be needed, compared with what would be needed
11577 with a more specific choice of prevailing FPU mode.
11580 This is the mode used for floating-point calculations with
11581 truncating (i.e.@: round towards zero) rounding mode. That includes
11582 conversion from floating point to integer.
11584 @item round-nearest
11585 This is the mode used for floating-point calculations with
11586 round-to-nearest-or-even rounding mode.
11589 This is the mode used to perform integer calculations in the FPU, e.g.@:
11590 integer multiply, or integer multiply-and-accumulate.
11593 The default is @option{-mfp-mode=caller}
11595 @item -mnosplit-lohi
11596 @itemx -mno-postinc
11597 @itemx -mno-postmodify
11598 @opindex mnosplit-lohi
11599 @opindex mno-postinc
11600 @opindex mno-postmodify
11601 Code generation tweaks that disable, respectively, splitting of 32-bit
11602 loads, generation of post-increment addresses, and generation of
11603 post-modify addresses. The defaults are @option{msplit-lohi},
11604 @option{-mpost-inc}, and @option{-mpost-modify}.
11606 @item -mnovect-double
11607 @opindex mno-vect-double
11608 Change the preferred SIMD mode to SImode. The default is
11609 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
11611 @item -max-vect-align=@var{num}
11612 @opindex max-vect-align
11613 The maximum alignment for SIMD vector mode types.
11614 @var{num} may be 4 or 8. The default is 8.
11615 Note that this is an ABI change, even though many library function
11616 interfaces are unaffected if they don't use SIMD vector modes
11617 in places that affect size and/or alignment of relevant types.
11619 @item -msplit-vecmove-early
11620 @opindex msplit-vecmove-early
11621 Split vector moves into single word moves before reload. In theory this
11622 can give better register allocation, but so far the reverse seems to be
11623 generally the case.
11625 @item -m1reg-@var{reg}
11627 Specify a register to hold the constant @minus{}1, which makes loading small negative
11628 constants and certain bitmasks faster.
11629 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
11630 which specify use of that register as a fixed register,
11631 and @samp{none}, which means that no register is used for this
11632 purpose. The default is @option{-m1reg-none}.
11637 @subsection ARC Options
11638 @cindex ARC options
11640 The following options control the architecture variant for which code
11643 @c architecture variants
11646 @item -mbarrel-shifter
11647 @opindex mbarrel-shifter
11648 Generate instructions supported by barrel shifter. This is the default
11649 unless @samp{-mcpu=ARC601} is in effect.
11651 @item -mcpu=@var{cpu}
11653 Set architecture type, register usage, and instruction scheduling
11654 parameters for @var{cpu}. There are also shortcut alias options
11655 available for backward compatibility and convenience. Supported
11656 values for @var{cpu} are
11662 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
11666 Compile for ARC601. Alias: @option{-mARC601}.
11671 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
11672 This is the default when configured with @samp{--with-cpu=arc700}@.
11677 @itemx -mdpfp-compact
11678 @opindex mdpfp-compact
11679 FPX: Generate Double Precision FPX instructions, tuned for the compact
11683 @opindex mdpfp-fast
11684 FPX: Generate Double Precision FPX instructions, tuned for the fast
11687 @item -mno-dpfp-lrsr
11688 @opindex mno-dpfp-lrsr
11689 Disable LR and SR instructions from using FPX extension aux registers.
11693 Generate Extended arithmetic instructions. Currently only
11694 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
11695 supported. This is always enabled for @samp{-mcpu=ARC700}.
11699 Do not generate mpy instructions for ARC700.
11703 Generate 32x16 bit multiply and mac instructions.
11707 Generate mul64 and mulu64 instructions. Only valid for @samp{-mcpu=ARC600}.
11711 Generate norm instruction. This is the default if @samp{-mcpu=ARC700}
11716 @itemx -mspfp-compact
11717 @opindex mspfp-compact
11718 FPX: Generate Single Precision FPX instructions, tuned for the compact
11722 @opindex mspfp-fast
11723 FPX: Generate Single Precision FPX instructions, tuned for the fast
11728 Enable generation of ARC SIMD instructions via target-specific
11729 builtins. Only valid for @samp{-mcpu=ARC700}.
11732 @opindex msoft-float
11733 This option ignored; it is provided for compatibility purposes only.
11734 Software floating point code is emitted by default, and this default
11735 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
11736 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
11737 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
11741 Generate swap instructions.
11745 The following options are passed through to the assembler, and also
11746 define preprocessor macro symbols.
11748 @c Flags used by the assembler, but for which we define preprocessor
11749 @c macro symbols as well.
11752 @opindex mdsp-packa
11753 Passed down to the assembler to enable the DSP Pack A extensions.
11754 Also sets the preprocessor symbol @code{__Xdsp_packa}.
11758 Passed down to the assembler to enable the dual viterbi butterfly
11759 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
11761 @c ARC700 4.10 extension instruction
11764 Passed down to the assembler to enable the Locked Load/Store
11765 Conditional extension. Also sets the preprocessor symbol
11770 Passed down to the assembler. Also sets the preprocessor symbol
11771 @code{__Xxmac_d16}.
11775 Passed down to the assembler. Also sets the preprocessor symbol
11778 @c ARC700 4.10 extension instruction
11781 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
11782 extension instruction. Also sets the preprocessor symbol
11785 @c ARC700 4.10 extension instruction
11788 Passed down to the assembler to enable the swap byte ordering
11789 extension instruction. Also sets the preprocessor symbol
11793 @opindex mtelephony
11794 Passed down to the assembler to enable dual and single operand
11795 instructions for telephony. Also sets the preprocessor symbol
11796 @code{__Xtelephony}.
11800 Passed down to the assembler to enable the XY Memory extension. Also
11801 sets the preprocessor symbol @code{__Xxy}.
11805 The following options control how the assembly code is annotated:
11807 @c Assembly annotation options
11811 Annotate assembler instructions with estimated addresses.
11813 @item -mannotate-align
11814 @opindex mannotate-align
11815 Explain what alignment considerations lead to the decision to make an
11816 instruction short or long.
11820 The following options are passed through to the linker:
11822 @c options passed through to the linker
11826 Passed through to the linker, to specify use of the @code{arclinux} emulation.
11827 This option is enabled by default in tool chains built for
11828 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
11829 when profiling is not requested.
11831 @item -marclinux_prof
11832 @opindex marclinux_prof
11833 Passed through to the linker, to specify use of the
11834 @code{arclinux_prof} emulation. This option is enabled by default in
11835 tool chains built for @w{@code{arc-linux-uclibc}} and
11836 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
11840 The following options control the semantics of generated code:
11842 @c semantically relevant code generation options
11844 @item -mepilogue-cfi
11845 @opindex mepilogue-cfi
11846 Enable generation of call frame information for epilogues.
11848 @item -mno-epilogue-cfi
11849 @opindex mno-epilogue-cfi
11850 Disable generation of call frame information for epilogues.
11853 @opindex mlong-calls
11854 Generate call insns as register indirect calls, thus providing access
11855 to the full 32-bit address range.
11857 @item -mmedium-calls
11858 @opindex mmedium-calls
11859 Don't use less than 25 bit addressing range for calls, which is the
11860 offset available for an unconditional branch-and-link
11861 instruction. Conditional execution of function calls is suppressed, to
11862 allow use of the 25-bit range, rather than the 21-bit range with
11863 conditional branch-and-link. This is the default for tool chains built
11864 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
11868 Do not generate sdata references. This is the default for tool chains
11869 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
11873 @opindex mucb-mcount
11874 Instrument with mcount calls as used in UCB code. I.e. do the
11875 counting in the callee, not the caller. By default ARC instrumentation
11876 counts in the caller.
11878 @item -mvolatile-cache
11879 @opindex mvolatile-cache
11880 Use ordinarily cached memory accesses for volatile references. This is the
11883 @item -mno-volatile-cache
11884 @opindex mno-volatile-cache
11885 Enable cache bypass for volatile references.
11889 The following options fine tune code generation:
11890 @c code generation tuning options
11893 @opindex malign-call
11894 Do alignment optimizations for call instructions.
11896 @item -mauto-modify-reg
11897 @opindex mauto-modify-reg
11898 Enable the use of pre/post modify with register displacement.
11900 @item -mbbit-peephole
11901 @opindex mbbit-peephole
11902 Enable bbit peephole2.
11906 This option disables a target-specific pass in @file{arc_reorg} to
11907 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
11908 generation driven by the combiner pass.
11910 @item -mcase-vector-pcrel
11911 @opindex mcase-vector-pcrel
11912 Use pc-relative switch case tables - this enables case table shortening.
11913 This is the default for @option{-Os}.
11915 @item -mcompact-casesi
11916 @opindex mcompact-casesi
11917 Enable compact casesi pattern.
11918 This is the default for @option{-Os}.
11920 @item -mno-cond-exec
11921 @opindex mno-cond-exec
11922 Disable ARCompact specific pass to generate conditional execution instructions.
11923 Due to delay slot scheduling and interactions between operand numbers,
11924 literal sizes, instruction lengths, and the support for conditional execution,
11925 the target-independent pass to generate conditional execution is often lacking,
11926 so the ARC port has kept a special pass around that tries to find more
11927 conditional execution generating opportunities after register allocation,
11928 branch shortening, and delay slot scheduling have been done. This pass
11929 generally, but not always, improves performance and code size, at the cost of
11930 extra compilation time, which is why there is an option to switch it off.
11931 If you have a problem with call instructions exceeding their allowable
11932 offset range because they are conditionalized, you should consider using
11933 @option{-mmedium-calls} instead.
11935 @item -mearly-cbranchsi
11936 @opindex mearly-cbranchsi
11937 Enable pre-reload use of the cbranchsi pattern.
11939 @item -mexpand-adddi
11940 @opindex mexpand-adddi
11941 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
11942 @code{add.f}, @code{adc} etc.
11944 @item -mindexed-loads
11945 @opindex mindexed-loads
11946 Enable the use of indexed loads. This can be problematic because some
11947 optimizers will then assume the that indexed stores exist, which is not
11952 Enable Local Register Allocation. This is still experimental for ARC,
11953 so by default the compiler uses standard reload
11954 (i.e. @samp{-mno-lra}).
11956 @item -mlra-priority-none
11957 @opindex mlra-priority-none
11958 Don't indicate any priority for target registers.
11960 @item -mlra-priority-compact
11961 @opindex mlra-priority-compact
11962 Indicate target register priority for r0..r3 / r12..r15.
11964 @item -mlra-priority-noncompact
11965 @opindex mlra-priority-noncompact
11966 Reduce target regsiter priority for r0..r3 / r12..r15.
11968 @item -mno-millicode
11969 @opindex mno-millicode
11970 When optimizing for size (using @option{-Os}), prologues and epilogues
11971 that have to save or restore a large number of registers are often
11972 shortened by using call to a special function in libgcc; this is
11973 referred to as a @emph{millicode} call. As these calls can pose
11974 performance issues, and/or cause linking issues when linking in a
11975 nonstandard way, this option is provided to turn off millicode call
11979 @opindex mmixed-code
11980 Tweak register allocation to help 16-bit instruction generation.
11981 This generally has the effect of decreasing the average instruction size
11982 while increasing the instruction count.
11986 Enable 'q' instruction alternatives.
11987 This is the default for @option{-Os}.
11991 Enable Rcq constraint handling - most short code generation depends on this.
11992 This is the default.
11996 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
11997 This is the default.
11999 @item -msize-level=@var{level}
12000 @opindex msize-level
12001 Fine-tune size optimization with regards to instruction lengths and alignment.
12002 The recognized values for @var{level} are:
12005 No size optimization. This level is deprecated and treated like @samp{1}.
12008 Short instructions are used opportunistically.
12011 In addition, alignment of loops and of code after barriers are dropped.
12014 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12018 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12019 the behavior when this is not set is equivalent to level @samp{1}.
12021 @item -mtune=@var{cpu}
12023 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12024 by @option{-mcpu=}.
12026 Supported values for @var{cpu} are
12030 Tune for ARC600 cpu.
12033 Tune for ARC601 cpu.
12036 Tune for ARC700 cpu with standard multiplier block.
12039 Tune for ARC700 cpu with XMAC block.
12042 Tune for ARC725D cpu.
12045 Tune for ARC750D cpu.
12049 @item -mmultcost=@var{num}
12051 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12052 normal instruction.
12054 @item -munalign-prob-threshold=@var{probability}
12055 @opindex munalign-prob-threshold
12056 Set probability threshold for unaligning branches.
12057 When tuning for @samp{ARC700} and optimizing for speed, branches without
12058 filled delay slot are preferably emitted unaligned and long, unless
12059 profiling indicates that the probability for the branch to be taken
12060 is below @var{probability}. @xref{Cross-profiling}.
12061 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12065 The following options are maintained for backward compatibility, but
12066 are now deprecated and will be removed in a future release:
12068 @c Deprecated options
12076 @opindex mbig-endian
12079 Compile code for big endian targets. Use of these options is now
12080 deprecated. Users wanting big-endian code, should use the
12081 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12082 building the tool chain, for which big-endian is the default.
12084 @item -mlittle-endian
12085 @opindex mlittle-endian
12088 Compile code for little endian targets. Use of these options is now
12089 deprecated. Users wanting little-endian code should use the
12090 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12091 building the tool chain, for which little-endian is the default.
12093 @item -mbarrel_shifter
12094 @opindex mbarrel_shifter
12095 Replaced by @samp{-mbarrel-shifter}
12097 @item -mdpfp_compact
12098 @opindex mdpfp_compact
12099 Replaced by @samp{-mdpfp-compact}
12102 @opindex mdpfp_fast
12103 Replaced by @samp{-mdpfp-fast}
12106 @opindex mdsp_packa
12107 Replaced by @samp{-mdsp-packa}
12111 Replaced by @samp{-mea}
12115 Replaced by @samp{-mmac-24}
12119 Replaced by @samp{-mmac-d16}
12121 @item -mspfp_compact
12122 @opindex mspfp_compact
12123 Replaced by @samp{-mspfp-compact}
12126 @opindex mspfp_fast
12127 Replaced by @samp{-mspfp-fast}
12129 @item -mtune=@var{cpu}
12131 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
12132 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
12133 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
12135 @item -multcost=@var{num}
12137 Replaced by @samp{-mmultcost}.
12142 @subsection ARM Options
12143 @cindex ARM options
12145 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
12149 @item -mabi=@var{name}
12151 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
12152 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
12155 @opindex mapcs-frame
12156 Generate a stack frame that is compliant with the ARM Procedure Call
12157 Standard for all functions, even if this is not strictly necessary for
12158 correct execution of the code. Specifying @option{-fomit-frame-pointer}
12159 with this option causes the stack frames not to be generated for
12160 leaf functions. The default is @option{-mno-apcs-frame}.
12164 This is a synonym for @option{-mapcs-frame}.
12167 @c not currently implemented
12168 @item -mapcs-stack-check
12169 @opindex mapcs-stack-check
12170 Generate code to check the amount of stack space available upon entry to
12171 every function (that actually uses some stack space). If there is
12172 insufficient space available then either the function
12173 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} is
12174 called, depending upon the amount of stack space required. The runtime
12175 system is required to provide these functions. The default is
12176 @option{-mno-apcs-stack-check}, since this produces smaller code.
12178 @c not currently implemented
12180 @opindex mapcs-float
12181 Pass floating-point arguments using the floating-point registers. This is
12182 one of the variants of the APCS@. This option is recommended if the
12183 target hardware has a floating-point unit or if a lot of floating-point
12184 arithmetic is going to be performed by the code. The default is
12185 @option{-mno-apcs-float}, since the size of integer-only code is
12186 slightly increased if @option{-mapcs-float} is used.
12188 @c not currently implemented
12189 @item -mapcs-reentrant
12190 @opindex mapcs-reentrant
12191 Generate reentrant, position-independent code. The default is
12192 @option{-mno-apcs-reentrant}.
12195 @item -mthumb-interwork
12196 @opindex mthumb-interwork
12197 Generate code that supports calling between the ARM and Thumb
12198 instruction sets. Without this option, on pre-v5 architectures, the
12199 two instruction sets cannot be reliably used inside one program. The
12200 default is @option{-mno-thumb-interwork}, since slightly larger code
12201 is generated when @option{-mthumb-interwork} is specified. In AAPCS
12202 configurations this option is meaningless.
12204 @item -mno-sched-prolog
12205 @opindex mno-sched-prolog
12206 Prevent the reordering of instructions in the function prologue, or the
12207 merging of those instruction with the instructions in the function's
12208 body. This means that all functions start with a recognizable set
12209 of instructions (or in fact one of a choice from a small set of
12210 different function prologues), and this information can be used to
12211 locate the start of functions inside an executable piece of code. The
12212 default is @option{-msched-prolog}.
12214 @item -mfloat-abi=@var{name}
12215 @opindex mfloat-abi
12216 Specifies which floating-point ABI to use. Permissible values
12217 are: @samp{soft}, @samp{softfp} and @samp{hard}.
12219 Specifying @samp{soft} causes GCC to generate output containing
12220 library calls for floating-point operations.
12221 @samp{softfp} allows the generation of code using hardware floating-point
12222 instructions, but still uses the soft-float calling conventions.
12223 @samp{hard} allows generation of floating-point instructions
12224 and uses FPU-specific calling conventions.
12226 The default depends on the specific target configuration. Note that
12227 the hard-float and soft-float ABIs are not link-compatible; you must
12228 compile your entire program with the same ABI, and link with a
12229 compatible set of libraries.
12231 @item -mlittle-endian
12232 @opindex mlittle-endian
12233 Generate code for a processor running in little-endian mode. This is
12234 the default for all standard configurations.
12237 @opindex mbig-endian
12238 Generate code for a processor running in big-endian mode; the default is
12239 to compile code for a little-endian processor.
12241 @item -mwords-little-endian
12242 @opindex mwords-little-endian
12243 This option only applies when generating code for big-endian processors.
12244 Generate code for a little-endian word order but a big-endian byte
12245 order. That is, a byte order of the form @samp{32107654}. Note: this
12246 option should only be used if you require compatibility with code for
12247 big-endian ARM processors generated by versions of the compiler prior to
12248 2.8. This option is now deprecated.
12250 @item -march=@var{name}
12252 This specifies the name of the target ARM architecture. GCC uses this
12253 name to determine what kind of instructions it can emit when generating
12254 assembly code. This option can be used in conjunction with or instead
12255 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
12256 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
12257 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
12258 @samp{armv6}, @samp{armv6j},
12259 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
12260 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
12261 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
12262 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
12264 @option{-march=armv7ve} is the armv7-a architecture with virtualization
12267 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
12268 architecture together with the optional CRC32 extensions.
12270 @option{-march=native} causes the compiler to auto-detect the architecture
12271 of the build computer. At present, this feature is only supported on
12272 Linux, and not all architectures are recognized. If the auto-detect is
12273 unsuccessful the option has no effect.
12275 @item -mtune=@var{name}
12277 This option specifies the name of the target ARM processor for
12278 which GCC should tune the performance of the code.
12279 For some ARM implementations better performance can be obtained by using
12281 Permissible names are: @samp{arm2}, @samp{arm250},
12282 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
12283 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
12284 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
12285 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
12287 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
12288 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
12289 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
12290 @samp{strongarm1110},
12291 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
12292 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
12293 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
12294 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
12295 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
12296 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
12297 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
12298 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
12299 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53}, @samp{cortex-a57},
12301 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m4},
12305 @samp{cortex-m0plus},
12306 @samp{marvell-pj4},
12307 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
12308 @samp{fa526}, @samp{fa626},
12309 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
12311 Additionally, this option can specify that GCC should tune the performance
12312 of the code for a big.LITTLE system. Permissible names are:
12313 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53}.
12315 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
12316 performance for a blend of processors within architecture @var{arch}.
12317 The aim is to generate code that run well on the current most popular
12318 processors, balancing between optimizations that benefit some CPUs in the
12319 range, and avoiding performance pitfalls of other CPUs. The effects of
12320 this option may change in future GCC versions as CPU models come and go.
12322 @option{-mtune=native} causes the compiler to auto-detect the CPU
12323 of the build computer. At present, this feature is only supported on
12324 Linux, and not all architectures are recognized. If the auto-detect is
12325 unsuccessful the option has no effect.
12327 @item -mcpu=@var{name}
12329 This specifies the name of the target ARM processor. GCC uses this name
12330 to derive the name of the target ARM architecture (as if specified
12331 by @option{-march}) and the ARM processor type for which to tune for
12332 performance (as if specified by @option{-mtune}). Where this option
12333 is used in conjunction with @option{-march} or @option{-mtune},
12334 those options take precedence over the appropriate part of this option.
12336 Permissible names for this option are the same as those for
12339 @option{-mcpu=generic-@var{arch}} is also permissible, and is
12340 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
12341 See @option{-mtune} for more information.
12343 @option{-mcpu=native} causes the compiler to auto-detect the CPU
12344 of the build computer. At present, this feature is only supported on
12345 Linux, and not all architectures are recognized. If the auto-detect is
12346 unsuccessful the option has no effect.
12348 @item -mfpu=@var{name}
12350 This specifies what floating-point hardware (or hardware emulation) is
12351 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
12352 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
12353 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
12354 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
12355 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
12357 If @option{-msoft-float} is specified this specifies the format of
12358 floating-point values.
12360 If the selected floating-point hardware includes the NEON extension
12361 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
12362 operations are not generated by GCC's auto-vectorization pass unless
12363 @option{-funsafe-math-optimizations} is also specified. This is
12364 because NEON hardware does not fully implement the IEEE 754 standard for
12365 floating-point arithmetic (in particular denormal values are treated as
12366 zero), so the use of NEON instructions may lead to a loss of precision.
12368 @item -mfp16-format=@var{name}
12369 @opindex mfp16-format
12370 Specify the format of the @code{__fp16} half-precision floating-point type.
12371 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
12372 the default is @samp{none}, in which case the @code{__fp16} type is not
12373 defined. @xref{Half-Precision}, for more information.
12375 @item -mstructure-size-boundary=@var{n}
12376 @opindex mstructure-size-boundary
12377 The sizes of all structures and unions are rounded up to a multiple
12378 of the number of bits set by this option. Permissible values are 8, 32
12379 and 64. The default value varies for different toolchains. For the COFF
12380 targeted toolchain the default value is 8. A value of 64 is only allowed
12381 if the underlying ABI supports it.
12383 Specifying a larger number can produce faster, more efficient code, but
12384 can also increase the size of the program. Different values are potentially
12385 incompatible. Code compiled with one value cannot necessarily expect to
12386 work with code or libraries compiled with another value, if they exchange
12387 information using structures or unions.
12389 @item -mabort-on-noreturn
12390 @opindex mabort-on-noreturn
12391 Generate a call to the function @code{abort} at the end of a
12392 @code{noreturn} function. It is executed if the function tries to
12396 @itemx -mno-long-calls
12397 @opindex mlong-calls
12398 @opindex mno-long-calls
12399 Tells the compiler to perform function calls by first loading the
12400 address of the function into a register and then performing a subroutine
12401 call on this register. This switch is needed if the target function
12402 lies outside of the 64-megabyte addressing range of the offset-based
12403 version of subroutine call instruction.
12405 Even if this switch is enabled, not all function calls are turned
12406 into long calls. The heuristic is that static functions, functions
12407 that have the @samp{short-call} attribute, functions that are inside
12408 the scope of a @samp{#pragma no_long_calls} directive, and functions whose
12409 definitions have already been compiled within the current compilation
12410 unit are not turned into long calls. The exceptions to this rule are
12411 that weak function definitions, functions with the @samp{long-call}
12412 attribute or the @samp{section} attribute, and functions that are within
12413 the scope of a @samp{#pragma long_calls} directive are always
12414 turned into long calls.
12416 This feature is not enabled by default. Specifying
12417 @option{-mno-long-calls} restores the default behavior, as does
12418 placing the function calls within the scope of a @samp{#pragma
12419 long_calls_off} directive. Note these switches have no effect on how
12420 the compiler generates code to handle function calls via function
12423 @item -msingle-pic-base
12424 @opindex msingle-pic-base
12425 Treat the register used for PIC addressing as read-only, rather than
12426 loading it in the prologue for each function. The runtime system is
12427 responsible for initializing this register with an appropriate value
12428 before execution begins.
12430 @item -mpic-register=@var{reg}
12431 @opindex mpic-register
12432 Specify the register to be used for PIC addressing.
12433 For standard PIC base case, the default will be any suitable register
12434 determined by compiler. For single PIC base case, the default is
12435 @samp{R9} if target is EABI based or stack-checking is enabled,
12436 otherwise the default is @samp{R10}.
12438 @item -mpic-data-is-text-relative
12439 @opindex mpic-data-is-text-relative
12440 Assume that each data segments are relative to text segment at load time.
12441 Therefore, it permits addressing data using PC-relative operations.
12442 This option is on by default for targets other than VxWorks RTP.
12444 @item -mpoke-function-name
12445 @opindex mpoke-function-name
12446 Write the name of each function into the text section, directly
12447 preceding the function prologue. The generated code is similar to this:
12451 .ascii "arm_poke_function_name", 0
12454 .word 0xff000000 + (t1 - t0)
12455 arm_poke_function_name
12457 stmfd sp!, @{fp, ip, lr, pc@}
12461 When performing a stack backtrace, code can inspect the value of
12462 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
12463 location @code{pc - 12} and the top 8 bits are set, then we know that
12464 there is a function name embedded immediately preceding this location
12465 and has length @code{((pc[-3]) & 0xff000000)}.
12472 Select between generating code that executes in ARM and Thumb
12473 states. The default for most configurations is to generate code
12474 that executes in ARM state, but the default can be changed by
12475 configuring GCC with the @option{--with-mode=}@var{state}
12479 @opindex mtpcs-frame
12480 Generate a stack frame that is compliant with the Thumb Procedure Call
12481 Standard for all non-leaf functions. (A leaf function is one that does
12482 not call any other functions.) The default is @option{-mno-tpcs-frame}.
12484 @item -mtpcs-leaf-frame
12485 @opindex mtpcs-leaf-frame
12486 Generate a stack frame that is compliant with the Thumb Procedure Call
12487 Standard for all leaf functions. (A leaf function is one that does
12488 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
12490 @item -mcallee-super-interworking
12491 @opindex mcallee-super-interworking
12492 Gives all externally visible functions in the file being compiled an ARM
12493 instruction set header which switches to Thumb mode before executing the
12494 rest of the function. This allows these functions to be called from
12495 non-interworking code. This option is not valid in AAPCS configurations
12496 because interworking is enabled by default.
12498 @item -mcaller-super-interworking
12499 @opindex mcaller-super-interworking
12500 Allows calls via function pointers (including virtual functions) to
12501 execute correctly regardless of whether the target code has been
12502 compiled for interworking or not. There is a small overhead in the cost
12503 of executing a function pointer if this option is enabled. This option
12504 is not valid in AAPCS configurations because interworking is enabled
12507 @item -mtp=@var{name}
12509 Specify the access model for the thread local storage pointer. The valid
12510 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
12511 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
12512 (supported in the arm6k architecture), and @option{auto}, which uses the
12513 best available method for the selected processor. The default setting is
12516 @item -mtls-dialect=@var{dialect}
12517 @opindex mtls-dialect
12518 Specify the dialect to use for accessing thread local storage. Two
12519 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
12520 @samp{gnu} dialect selects the original GNU scheme for supporting
12521 local and global dynamic TLS models. The @samp{gnu2} dialect
12522 selects the GNU descriptor scheme, which provides better performance
12523 for shared libraries. The GNU descriptor scheme is compatible with
12524 the original scheme, but does require new assembler, linker and
12525 library support. Initial and local exec TLS models are unaffected by
12526 this option and always use the original scheme.
12528 @item -mword-relocations
12529 @opindex mword-relocations
12530 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
12531 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
12532 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
12535 @item -mfix-cortex-m3-ldrd
12536 @opindex mfix-cortex-m3-ldrd
12537 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
12538 with overlapping destination and base registers are used. This option avoids
12539 generating these instructions. This option is enabled by default when
12540 @option{-mcpu=cortex-m3} is specified.
12542 @item -munaligned-access
12543 @itemx -mno-unaligned-access
12544 @opindex munaligned-access
12545 @opindex mno-unaligned-access
12546 Enables (or disables) reading and writing of 16- and 32- bit values
12547 from addresses that are not 16- or 32- bit aligned. By default
12548 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
12549 architectures, and enabled for all other architectures. If unaligned
12550 access is not enabled then words in packed data structures will be
12551 accessed a byte at a time.
12553 The ARM attribute @code{Tag_CPU_unaligned_access} will be set in the
12554 generated object file to either true or false, depending upon the
12555 setting of this option. If unaligned access is enabled then the
12556 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} will also be
12559 @item -mneon-for-64bits
12560 @opindex mneon-for-64bits
12561 Enables using Neon to handle scalar 64-bits operations. This is
12562 disabled by default since the cost of moving data from core registers
12565 @item -mslow-flash-data
12566 @opindex mslow-flash-data
12567 Assume loading data from flash is slower than fetching instruction.
12568 Therefore literal load is minimized for better performance.
12569 This option is only supported when compiling for ARMv7 M-profile and
12572 @item -mrestrict-it
12573 @opindex mrestrict-it
12574 Restricts generation of IT blocks to conform to the rules of ARMv8.
12575 IT blocks can only contain a single 16-bit instruction from a select
12576 set of instructions. This option is on by default for ARMv8 Thumb mode.
12580 @subsection AVR Options
12581 @cindex AVR Options
12583 These options are defined for AVR implementations:
12586 @item -mmcu=@var{mcu}
12588 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
12590 The default for this option is@tie{}@code{avr2}.
12592 GCC supports the following AVR devices and ISAs:
12594 @include avr-mmcu.texi
12596 @item -maccumulate-args
12597 @opindex maccumulate-args
12598 Accumulate outgoing function arguments and acquire/release the needed
12599 stack space for outgoing function arguments once in function
12600 prologue/epilogue. Without this option, outgoing arguments are pushed
12601 before calling a function and popped afterwards.
12603 Popping the arguments after the function call can be expensive on
12604 AVR so that accumulating the stack space might lead to smaller
12605 executables because arguments need not to be removed from the
12606 stack after such a function call.
12608 This option can lead to reduced code size for functions that perform
12609 several calls to functions that get their arguments on the stack like
12610 calls to printf-like functions.
12612 @item -mbranch-cost=@var{cost}
12613 @opindex mbranch-cost
12614 Set the branch costs for conditional branch instructions to
12615 @var{cost}. Reasonable values for @var{cost} are small, non-negative
12616 integers. The default branch cost is 0.
12618 @item -mcall-prologues
12619 @opindex mcall-prologues
12620 Functions prologues/epilogues are expanded as calls to appropriate
12621 subroutines. Code size is smaller.
12625 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
12626 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
12627 and @code{long long} is 4 bytes. Please note that this option does not
12628 conform to the C standards, but it results in smaller code
12631 @item -mno-interrupts
12632 @opindex mno-interrupts
12633 Generated code is not compatible with hardware interrupts.
12634 Code size is smaller.
12638 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
12639 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
12640 Setting @code{-mrelax} just adds the @code{--relax} option to the
12641 linker command line when the linker is called.
12643 Jump relaxing is performed by the linker because jump offsets are not
12644 known before code is located. Therefore, the assembler code generated by the
12645 compiler is the same, but the instructions in the executable may
12646 differ from instructions in the assembler code.
12648 Relaxing must be turned on if linker stubs are needed, see the
12649 section on @code{EIND} and linker stubs below.
12653 Treat the stack pointer register as an 8-bit register,
12654 i.e.@: assume the high byte of the stack pointer is zero.
12655 In general, you don't need to set this option by hand.
12657 This option is used internally by the compiler to select and
12658 build multilibs for architectures @code{avr2} and @code{avr25}.
12659 These architectures mix devices with and without @code{SPH}.
12660 For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25}
12661 the compiler driver will add or remove this option from the compiler
12662 proper's command line, because the compiler then knows if the device
12663 or architecture has an 8-bit stack pointer and thus no @code{SPH}
12668 Use address register @code{X} in a way proposed by the hardware. This means
12669 that @code{X} is only used in indirect, post-increment or
12670 pre-decrement addressing.
12672 Without this option, the @code{X} register may be used in the same way
12673 as @code{Y} or @code{Z} which then is emulated by additional
12675 For example, loading a value with @code{X+const} addressing with a
12676 small non-negative @code{const < 64} to a register @var{Rn} is
12680 adiw r26, const ; X += const
12681 ld @var{Rn}, X ; @var{Rn} = *X
12682 sbiw r26, const ; X -= const
12686 @opindex mtiny-stack
12687 Only change the lower 8@tie{}bits of the stack pointer.
12689 @item -Waddr-space-convert
12690 @opindex Waddr-space-convert
12691 Warn about conversions between address spaces in the case where the
12692 resulting address space is not contained in the incoming address space.
12695 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
12696 @cindex @code{EIND}
12697 Pointers in the implementation are 16@tie{}bits wide.
12698 The address of a function or label is represented as word address so
12699 that indirect jumps and calls can target any code address in the
12700 range of 64@tie{}Ki words.
12702 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
12703 bytes of program memory space, there is a special function register called
12704 @code{EIND} that serves as most significant part of the target address
12705 when @code{EICALL} or @code{EIJMP} instructions are used.
12707 Indirect jumps and calls on these devices are handled as follows by
12708 the compiler and are subject to some limitations:
12713 The compiler never sets @code{EIND}.
12716 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
12717 instructions or might read @code{EIND} directly in order to emulate an
12718 indirect call/jump by means of a @code{RET} instruction.
12721 The compiler assumes that @code{EIND} never changes during the startup
12722 code or during the application. In particular, @code{EIND} is not
12723 saved/restored in function or interrupt service routine
12727 For indirect calls to functions and computed goto, the linker
12728 generates @emph{stubs}. Stubs are jump pads sometimes also called
12729 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
12730 The stub contains a direct jump to the desired address.
12733 Linker relaxation must be turned on so that the linker will generate
12734 the stubs correctly an all situaltion. See the compiler option
12735 @code{-mrelax} and the linler option @code{--relax}.
12736 There are corner cases where the linker is supposed to generate stubs
12737 but aborts without relaxation and without a helpful error message.
12740 The default linker script is arranged for code with @code{EIND = 0}.
12741 If code is supposed to work for a setup with @code{EIND != 0}, a custom
12742 linker script has to be used in order to place the sections whose
12743 name start with @code{.trampolines} into the segment where @code{EIND}
12747 The startup code from libgcc never sets @code{EIND}.
12748 Notice that startup code is a blend of code from libgcc and AVR-LibC.
12749 For the impact of AVR-LibC on @code{EIND}, see the
12750 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
12753 It is legitimate for user-specific startup code to set up @code{EIND}
12754 early, for example by means of initialization code located in
12755 section @code{.init3}. Such code runs prior to general startup code
12756 that initializes RAM and calls constructors, but after the bit
12757 of startup code from AVR-LibC that sets @code{EIND} to the segment
12758 where the vector table is located.
12760 #include <avr/io.h>
12763 __attribute__((section(".init3"),naked,used,no_instrument_function))
12764 init3_set_eind (void)
12766 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
12767 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
12772 The @code{__trampolines_start} symbol is defined in the linker script.
12775 Stubs are generated automatically by the linker if
12776 the following two conditions are met:
12779 @item The address of a label is taken by means of the @code{gs} modifier
12780 (short for @emph{generate stubs}) like so:
12782 LDI r24, lo8(gs(@var{func}))
12783 LDI r25, hi8(gs(@var{func}))
12785 @item The final location of that label is in a code segment
12786 @emph{outside} the segment where the stubs are located.
12790 The compiler emits such @code{gs} modifiers for code labels in the
12791 following situations:
12793 @item Taking address of a function or code label.
12794 @item Computed goto.
12795 @item If prologue-save function is used, see @option{-mcall-prologues}
12796 command-line option.
12797 @item Switch/case dispatch tables. If you do not want such dispatch
12798 tables you can specify the @option{-fno-jump-tables} command-line option.
12799 @item C and C++ constructors/destructors called during startup/shutdown.
12800 @item If the tools hit a @code{gs()} modifier explained above.
12804 Jumping to non-symbolic addresses like so is @emph{not} supported:
12809 /* Call function at word address 0x2 */
12810 return ((int(*)(void)) 0x2)();
12814 Instead, a stub has to be set up, i.e.@: the function has to be called
12815 through a symbol (@code{func_4} in the example):
12820 extern int func_4 (void);
12822 /* Call function at byte address 0x4 */
12827 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
12828 Alternatively, @code{func_4} can be defined in the linker script.
12831 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
12832 @cindex @code{RAMPD}
12833 @cindex @code{RAMPX}
12834 @cindex @code{RAMPY}
12835 @cindex @code{RAMPZ}
12836 Some AVR devices support memories larger than the 64@tie{}KiB range
12837 that can be accessed with 16-bit pointers. To access memory locations
12838 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
12839 register is used as high part of the address:
12840 The @code{X}, @code{Y}, @code{Z} address register is concatenated
12841 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
12842 register, respectively, to get a wide address. Similarly,
12843 @code{RAMPD} is used together with direct addressing.
12847 The startup code initializes the @code{RAMP} special function
12848 registers with zero.
12851 If a @ref{AVR Named Address Spaces,named address space} other than
12852 generic or @code{__flash} is used, then @code{RAMPZ} is set
12853 as needed before the operation.
12856 If the device supports RAM larger than 64@tie{}KiB and the compiler
12857 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
12858 is reset to zero after the operation.
12861 If the device comes with a specific @code{RAMP} register, the ISR
12862 prologue/epilogue saves/restores that SFR and initializes it with
12863 zero in case the ISR code might (implicitly) use it.
12866 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
12867 If you use inline assembler to read from locations outside the
12868 16-bit address range and change one of the @code{RAMP} registers,
12869 you must reset it to zero after the access.
12873 @subsubsection AVR Built-in Macros
12875 GCC defines several built-in macros so that the user code can test
12876 for the presence or absence of features. Almost any of the following
12877 built-in macros are deduced from device capabilities and thus
12878 triggered by the @code{-mmcu=} command-line option.
12880 For even more AVR-specific built-in macros see
12881 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
12886 Build-in macro that resolves to a decimal number that identifies the
12887 architecture and depends on the @code{-mmcu=@var{mcu}} option.
12888 Possible values are:
12890 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
12891 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
12892 @code{105}, @code{106}, @code{107}
12894 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
12895 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
12896 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
12897 @code{avrxmega6}, @code{avrxmega7}, respectively.
12898 If @var{mcu} specifies a device, this built-in macro is set
12899 accordingly. For example, with @code{-mmcu=atmega8} the macro will be
12900 defined to @code{4}.
12902 @item __AVR_@var{Device}__
12903 Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects
12904 the device's name. For example, @code{-mmcu=atmega8} defines the
12905 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
12906 @code{__AVR_ATtiny261A__}, etc.
12908 The built-in macros' names follow
12909 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
12910 the device name as from the AVR user manual. The difference between
12911 @var{Device} in the built-in macro and @var{device} in
12912 @code{-mmcu=@var{device}} is that the latter is always lowercase.
12914 If @var{device} is not a device but only a core architecture like
12915 @code{avr51}, this macro will not be defined.
12917 @item __AVR_XMEGA__
12918 The device / architecture belongs to the XMEGA family of devices.
12920 @item __AVR_HAVE_ELPM__
12921 The device has the the @code{ELPM} instruction.
12923 @item __AVR_HAVE_ELPMX__
12924 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
12925 R@var{n},Z+} instructions.
12927 @item __AVR_HAVE_MOVW__
12928 The device has the @code{MOVW} instruction to perform 16-bit
12929 register-register moves.
12931 @item __AVR_HAVE_LPMX__
12932 The device has the @code{LPM R@var{n},Z} and
12933 @code{LPM R@var{n},Z+} instructions.
12935 @item __AVR_HAVE_MUL__
12936 The device has a hardware multiplier.
12938 @item __AVR_HAVE_JMP_CALL__
12939 The device has the @code{JMP} and @code{CALL} instructions.
12940 This is the case for devices with at least 16@tie{}KiB of program
12943 @item __AVR_HAVE_EIJMP_EICALL__
12944 @itemx __AVR_3_BYTE_PC__
12945 The device has the @code{EIJMP} and @code{EICALL} instructions.
12946 This is the case for devices with more than 128@tie{}KiB of program memory.
12947 This also means that the program counter
12948 (PC) is 3@tie{}bytes wide.
12950 @item __AVR_2_BYTE_PC__
12951 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
12952 with up to 128@tie{}KiB of program memory.
12954 @item __AVR_HAVE_8BIT_SP__
12955 @itemx __AVR_HAVE_16BIT_SP__
12956 The stack pointer (SP) register is treated as 8-bit respectively
12957 16-bit register by the compiler.
12958 The definition of these macros is affected by @code{-mtiny-stack}.
12960 @item __AVR_HAVE_SPH__
12962 The device has the SPH (high part of stack pointer) special function
12963 register or has an 8-bit stack pointer, respectively.
12964 The definition of these macros is affected by @code{-mmcu=} and
12965 in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also
12968 @item __AVR_HAVE_RAMPD__
12969 @itemx __AVR_HAVE_RAMPX__
12970 @itemx __AVR_HAVE_RAMPY__
12971 @itemx __AVR_HAVE_RAMPZ__
12972 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
12973 @code{RAMPZ} special function register, respectively.
12975 @item __NO_INTERRUPTS__
12976 This macro reflects the @code{-mno-interrupts} command line option.
12978 @item __AVR_ERRATA_SKIP__
12979 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
12980 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
12981 instructions because of a hardware erratum. Skip instructions are
12982 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
12983 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
12986 @item __AVR_SFR_OFFSET__=@var{offset}
12987 Instructions that can address I/O special function registers directly
12988 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
12989 address as if addressed by an instruction to access RAM like @code{LD}
12990 or @code{STS}. This offset depends on the device architecture and has
12991 to be subtracted from the RAM address in order to get the
12992 respective I/O@tie{}address.
12994 @item __WITH_AVRLIBC__
12995 The compiler is configured to be used together with AVR-Libc.
12996 See the @code{--with-avrlibc} configure option.
13000 @node Blackfin Options
13001 @subsection Blackfin Options
13002 @cindex Blackfin Options
13005 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13007 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13008 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13009 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13010 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13011 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13012 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13013 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13014 @samp{bf561}, @samp{bf592}.
13016 The optional @var{sirevision} specifies the silicon revision of the target
13017 Blackfin processor. Any workarounds available for the targeted silicon revision
13018 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13019 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13020 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13021 hexadecimal digits representing the major and minor numbers in the silicon
13022 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13023 is not defined. If @var{sirevision} is @samp{any}, the
13024 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13025 If this optional @var{sirevision} is not used, GCC assumes the latest known
13026 silicon revision of the targeted Blackfin processor.
13028 GCC defines a preprocessor macro for the specified @var{cpu}.
13029 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13030 provided by libgloss to be linked in if @option{-msim} is not given.
13032 Without this option, @samp{bf532} is used as the processor by default.
13034 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13035 only the preprocessor macro is defined.
13039 Specifies that the program will be run on the simulator. This causes
13040 the simulator BSP provided by libgloss to be linked in. This option
13041 has effect only for @samp{bfin-elf} toolchain.
13042 Certain other options, such as @option{-mid-shared-library} and
13043 @option{-mfdpic}, imply @option{-msim}.
13045 @item -momit-leaf-frame-pointer
13046 @opindex momit-leaf-frame-pointer
13047 Don't keep the frame pointer in a register for leaf functions. This
13048 avoids the instructions to save, set up and restore frame pointers and
13049 makes an extra register available in leaf functions. The option
13050 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13051 which might make debugging harder.
13053 @item -mspecld-anomaly
13054 @opindex mspecld-anomaly
13055 When enabled, the compiler ensures that the generated code does not
13056 contain speculative loads after jump instructions. If this option is used,
13057 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13059 @item -mno-specld-anomaly
13060 @opindex mno-specld-anomaly
13061 Don't generate extra code to prevent speculative loads from occurring.
13063 @item -mcsync-anomaly
13064 @opindex mcsync-anomaly
13065 When enabled, the compiler ensures that the generated code does not
13066 contain CSYNC or SSYNC instructions too soon after conditional branches.
13067 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13069 @item -mno-csync-anomaly
13070 @opindex mno-csync-anomaly
13071 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13072 occurring too soon after a conditional branch.
13076 When enabled, the compiler is free to take advantage of the knowledge that
13077 the entire program fits into the low 64k of memory.
13080 @opindex mno-low-64k
13081 Assume that the program is arbitrarily large. This is the default.
13083 @item -mstack-check-l1
13084 @opindex mstack-check-l1
13085 Do stack checking using information placed into L1 scratchpad memory by the
13088 @item -mid-shared-library
13089 @opindex mid-shared-library
13090 Generate code that supports shared libraries via the library ID method.
13091 This allows for execute in place and shared libraries in an environment
13092 without virtual memory management. This option implies @option{-fPIC}.
13093 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13095 @item -mno-id-shared-library
13096 @opindex mno-id-shared-library
13097 Generate code that doesn't assume ID-based shared libraries are being used.
13098 This is the default.
13100 @item -mleaf-id-shared-library
13101 @opindex mleaf-id-shared-library
13102 Generate code that supports shared libraries via the library ID method,
13103 but assumes that this library or executable won't link against any other
13104 ID shared libraries. That allows the compiler to use faster code for jumps
13107 @item -mno-leaf-id-shared-library
13108 @opindex mno-leaf-id-shared-library
13109 Do not assume that the code being compiled won't link against any ID shared
13110 libraries. Slower code is generated for jump and call insns.
13112 @item -mshared-library-id=n
13113 @opindex mshared-library-id
13114 Specifies the identification number of the ID-based shared library being
13115 compiled. Specifying a value of 0 generates more compact code; specifying
13116 other values forces the allocation of that number to the current
13117 library but is no more space- or time-efficient than omitting this option.
13121 Generate code that allows the data segment to be located in a different
13122 area of memory from the text segment. This allows for execute in place in
13123 an environment without virtual memory management by eliminating relocations
13124 against the text section.
13126 @item -mno-sep-data
13127 @opindex mno-sep-data
13128 Generate code that assumes that the data segment follows the text segment.
13129 This is the default.
13132 @itemx -mno-long-calls
13133 @opindex mlong-calls
13134 @opindex mno-long-calls
13135 Tells the compiler to perform function calls by first loading the
13136 address of the function into a register and then performing a subroutine
13137 call on this register. This switch is needed if the target function
13138 lies outside of the 24-bit addressing range of the offset-based
13139 version of subroutine call instruction.
13141 This feature is not enabled by default. Specifying
13142 @option{-mno-long-calls} restores the default behavior. Note these
13143 switches have no effect on how the compiler generates code to handle
13144 function calls via function pointers.
13148 Link with the fast floating-point library. This library relaxes some of
13149 the IEEE floating-point standard's rules for checking inputs against
13150 Not-a-Number (NAN), in the interest of performance.
13153 @opindex minline-plt
13154 Enable inlining of PLT entries in function calls to functions that are
13155 not known to bind locally. It has no effect without @option{-mfdpic}.
13158 @opindex mmulticore
13159 Build a standalone application for multicore Blackfin processors.
13160 This option causes proper start files and link scripts supporting
13161 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
13162 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
13164 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
13165 selects the one-application-per-core programming model. Without
13166 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
13167 programming model is used. In this model, the main function of Core B
13168 should be named as @code{coreb_main}.
13170 If this option is not used, the single-core application programming
13175 Build a standalone application for Core A of BF561 when using
13176 the one-application-per-core programming model. Proper start files
13177 and link scripts are used to support Core A, and the macro
13178 @code{__BFIN_COREA} is defined.
13179 This option can only be used in conjunction with @option{-mmulticore}.
13183 Build a standalone application for Core B of BF561 when using
13184 the one-application-per-core programming model. Proper start files
13185 and link scripts are used to support Core B, and the macro
13186 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
13187 should be used instead of @code{main}.
13188 This option can only be used in conjunction with @option{-mmulticore}.
13192 Build a standalone application for SDRAM. Proper start files and
13193 link scripts are used to put the application into SDRAM, and the macro
13194 @code{__BFIN_SDRAM} is defined.
13195 The loader should initialize SDRAM before loading the application.
13199 Assume that ICPLBs are enabled at run time. This has an effect on certain
13200 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
13201 are enabled; for standalone applications the default is off.
13205 @subsection C6X Options
13206 @cindex C6X Options
13209 @item -march=@var{name}
13211 This specifies the name of the target architecture. GCC uses this
13212 name to determine what kind of instructions it can emit when generating
13213 assembly code. Permissible names are: @samp{c62x},
13214 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
13217 @opindex mbig-endian
13218 Generate code for a big-endian target.
13220 @item -mlittle-endian
13221 @opindex mlittle-endian
13222 Generate code for a little-endian target. This is the default.
13226 Choose startup files and linker script suitable for the simulator.
13228 @item -msdata=default
13229 @opindex msdata=default
13230 Put small global and static data in the @samp{.neardata} section,
13231 which is pointed to by register @code{B14}. Put small uninitialized
13232 global and static data in the @samp{.bss} section, which is adjacent
13233 to the @samp{.neardata} section. Put small read-only data into the
13234 @samp{.rodata} section. The corresponding sections used for large
13235 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
13238 @opindex msdata=all
13239 Put all data, not just small objects, into the sections reserved for
13240 small data, and use addressing relative to the @code{B14} register to
13244 @opindex msdata=none
13245 Make no use of the sections reserved for small data, and use absolute
13246 addresses to access all data. Put all initialized global and static
13247 data in the @samp{.fardata} section, and all uninitialized data in the
13248 @samp{.far} section. Put all constant data into the @samp{.const}
13253 @subsection CRIS Options
13254 @cindex CRIS Options
13256 These options are defined specifically for the CRIS ports.
13259 @item -march=@var{architecture-type}
13260 @itemx -mcpu=@var{architecture-type}
13263 Generate code for the specified architecture. The choices for
13264 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
13265 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
13266 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
13269 @item -mtune=@var{architecture-type}
13271 Tune to @var{architecture-type} everything applicable about the generated
13272 code, except for the ABI and the set of available instructions. The
13273 choices for @var{architecture-type} are the same as for
13274 @option{-march=@var{architecture-type}}.
13276 @item -mmax-stack-frame=@var{n}
13277 @opindex mmax-stack-frame
13278 Warn when the stack frame of a function exceeds @var{n} bytes.
13284 The options @option{-metrax4} and @option{-metrax100} are synonyms for
13285 @option{-march=v3} and @option{-march=v8} respectively.
13287 @item -mmul-bug-workaround
13288 @itemx -mno-mul-bug-workaround
13289 @opindex mmul-bug-workaround
13290 @opindex mno-mul-bug-workaround
13291 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
13292 models where it applies. This option is active by default.
13296 Enable CRIS-specific verbose debug-related information in the assembly
13297 code. This option also has the effect of turning off the @samp{#NO_APP}
13298 formatted-code indicator to the assembler at the beginning of the
13303 Do not use condition-code results from previous instruction; always emit
13304 compare and test instructions before use of condition codes.
13306 @item -mno-side-effects
13307 @opindex mno-side-effects
13308 Do not emit instructions with side effects in addressing modes other than
13311 @item -mstack-align
13312 @itemx -mno-stack-align
13313 @itemx -mdata-align
13314 @itemx -mno-data-align
13315 @itemx -mconst-align
13316 @itemx -mno-const-align
13317 @opindex mstack-align
13318 @opindex mno-stack-align
13319 @opindex mdata-align
13320 @opindex mno-data-align
13321 @opindex mconst-align
13322 @opindex mno-const-align
13323 These options (@samp{no-} options) arrange (eliminate arrangements) for the
13324 stack frame, individual data and constants to be aligned for the maximum
13325 single data access size for the chosen CPU model. The default is to
13326 arrange for 32-bit alignment. ABI details such as structure layout are
13327 not affected by these options.
13335 Similar to the stack- data- and const-align options above, these options
13336 arrange for stack frame, writable data and constants to all be 32-bit,
13337 16-bit or 8-bit aligned. The default is 32-bit alignment.
13339 @item -mno-prologue-epilogue
13340 @itemx -mprologue-epilogue
13341 @opindex mno-prologue-epilogue
13342 @opindex mprologue-epilogue
13343 With @option{-mno-prologue-epilogue}, the normal function prologue and
13344 epilogue which set up the stack frame are omitted and no return
13345 instructions or return sequences are generated in the code. Use this
13346 option only together with visual inspection of the compiled code: no
13347 warnings or errors are generated when call-saved registers must be saved,
13348 or storage for local variables needs to be allocated.
13352 @opindex mno-gotplt
13354 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
13355 instruction sequences that load addresses for functions from the PLT part
13356 of the GOT rather than (traditional on other architectures) calls to the
13357 PLT@. The default is @option{-mgotplt}.
13361 Legacy no-op option only recognized with the cris-axis-elf and
13362 cris-axis-linux-gnu targets.
13366 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
13370 This option, recognized for the cris-axis-elf, arranges
13371 to link with input-output functions from a simulator library. Code,
13372 initialized data and zero-initialized data are allocated consecutively.
13376 Like @option{-sim}, but pass linker options to locate initialized data at
13377 0x40000000 and zero-initialized data at 0x80000000.
13381 @subsection CR16 Options
13382 @cindex CR16 Options
13384 These options are defined specifically for the CR16 ports.
13390 Enable the use of multiply-accumulate instructions. Disabled by default.
13394 @opindex mcr16cplus
13396 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
13401 Links the library libsim.a which is in compatible with simulator. Applicable
13402 to ELF compiler only.
13406 Choose integer type as 32-bit wide.
13410 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
13412 @item -mdata-model=@var{model}
13413 @opindex mdata-model
13414 Choose a data model. The choices for @var{model} are @samp{near},
13415 @samp{far} or @samp{medium}. @samp{medium} is default.
13416 However, @samp{far} is not valid with @option{-mcr16c}, as the
13417 CR16C architecture does not support the far data model.
13420 @node Darwin Options
13421 @subsection Darwin Options
13422 @cindex Darwin options
13424 These options are defined for all architectures running the Darwin operating
13427 FSF GCC on Darwin does not create ``fat'' object files; it creates
13428 an object file for the single architecture that GCC was built to
13429 target. Apple's GCC on Darwin does create ``fat'' files if multiple
13430 @option{-arch} options are used; it does so by running the compiler or
13431 linker multiple times and joining the results together with
13434 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
13435 @samp{i686}) is determined by the flags that specify the ISA
13436 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
13437 @option{-force_cpusubtype_ALL} option can be used to override this.
13439 The Darwin tools vary in their behavior when presented with an ISA
13440 mismatch. The assembler, @file{as}, only permits instructions to
13441 be used that are valid for the subtype of the file it is generating,
13442 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
13443 The linker for shared libraries, @file{/usr/bin/libtool}, fails
13444 and prints an error if asked to create a shared library with a less
13445 restrictive subtype than its input files (for instance, trying to put
13446 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
13447 for executables, @command{ld}, quietly gives the executable the most
13448 restrictive subtype of any of its input files.
13453 Add the framework directory @var{dir} to the head of the list of
13454 directories to be searched for header files. These directories are
13455 interleaved with those specified by @option{-I} options and are
13456 scanned in a left-to-right order.
13458 A framework directory is a directory with frameworks in it. A
13459 framework is a directory with a @file{Headers} and/or
13460 @file{PrivateHeaders} directory contained directly in it that ends
13461 in @file{.framework}. The name of a framework is the name of this
13462 directory excluding the @file{.framework}. Headers associated with
13463 the framework are found in one of those two directories, with
13464 @file{Headers} being searched first. A subframework is a framework
13465 directory that is in a framework's @file{Frameworks} directory.
13466 Includes of subframework headers can only appear in a header of a
13467 framework that contains the subframework, or in a sibling subframework
13468 header. Two subframeworks are siblings if they occur in the same
13469 framework. A subframework should not have the same name as a
13470 framework; a warning is issued if this is violated. Currently a
13471 subframework cannot have subframeworks; in the future, the mechanism
13472 may be extended to support this. The standard frameworks can be found
13473 in @file{/System/Library/Frameworks} and
13474 @file{/Library/Frameworks}. An example include looks like
13475 @code{#include <Framework/header.h>}, where @file{Framework} denotes
13476 the name of the framework and @file{header.h} is found in the
13477 @file{PrivateHeaders} or @file{Headers} directory.
13479 @item -iframework@var{dir}
13480 @opindex iframework
13481 Like @option{-F} except the directory is a treated as a system
13482 directory. The main difference between this @option{-iframework} and
13483 @option{-F} is that with @option{-iframework} the compiler does not
13484 warn about constructs contained within header files found via
13485 @var{dir}. This option is valid only for the C family of languages.
13489 Emit debugging information for symbols that are used. For stabs
13490 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
13491 This is by default ON@.
13495 Emit debugging information for all symbols and types.
13497 @item -mmacosx-version-min=@var{version}
13498 The earliest version of MacOS X that this executable will run on
13499 is @var{version}. Typical values of @var{version} include @code{10.1},
13500 @code{10.2}, and @code{10.3.9}.
13502 If the compiler was built to use the system's headers by default,
13503 then the default for this option is the system version on which the
13504 compiler is running, otherwise the default is to make choices that
13505 are compatible with as many systems and code bases as possible.
13509 Enable kernel development mode. The @option{-mkernel} option sets
13510 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
13511 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
13512 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
13513 applicable. This mode also sets @option{-mno-altivec},
13514 @option{-msoft-float}, @option{-fno-builtin} and
13515 @option{-mlong-branch} for PowerPC targets.
13517 @item -mone-byte-bool
13518 @opindex mone-byte-bool
13519 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
13520 By default @samp{sizeof(bool)} is @samp{4} when compiling for
13521 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
13522 option has no effect on x86.
13524 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
13525 to generate code that is not binary compatible with code generated
13526 without that switch. Using this switch may require recompiling all
13527 other modules in a program, including system libraries. Use this
13528 switch to conform to a non-default data model.
13530 @item -mfix-and-continue
13531 @itemx -ffix-and-continue
13532 @itemx -findirect-data
13533 @opindex mfix-and-continue
13534 @opindex ffix-and-continue
13535 @opindex findirect-data
13536 Generate code suitable for fast turnaround development, such as to
13537 allow GDB to dynamically load @code{.o} files into already-running
13538 programs. @option{-findirect-data} and @option{-ffix-and-continue}
13539 are provided for backwards compatibility.
13543 Loads all members of static archive libraries.
13544 See man ld(1) for more information.
13546 @item -arch_errors_fatal
13547 @opindex arch_errors_fatal
13548 Cause the errors having to do with files that have the wrong architecture
13551 @item -bind_at_load
13552 @opindex bind_at_load
13553 Causes the output file to be marked such that the dynamic linker will
13554 bind all undefined references when the file is loaded or launched.
13558 Produce a Mach-o bundle format file.
13559 See man ld(1) for more information.
13561 @item -bundle_loader @var{executable}
13562 @opindex bundle_loader
13563 This option specifies the @var{executable} that will load the build
13564 output file being linked. See man ld(1) for more information.
13567 @opindex dynamiclib
13568 When passed this option, GCC produces a dynamic library instead of
13569 an executable when linking, using the Darwin @file{libtool} command.
13571 @item -force_cpusubtype_ALL
13572 @opindex force_cpusubtype_ALL
13573 This causes GCC's output file to have the @var{ALL} subtype, instead of
13574 one controlled by the @option{-mcpu} or @option{-march} option.
13576 @item -allowable_client @var{client_name}
13577 @itemx -client_name
13578 @itemx -compatibility_version
13579 @itemx -current_version
13581 @itemx -dependency-file
13583 @itemx -dylinker_install_name
13585 @itemx -exported_symbols_list
13588 @itemx -flat_namespace
13589 @itemx -force_flat_namespace
13590 @itemx -headerpad_max_install_names
13593 @itemx -install_name
13594 @itemx -keep_private_externs
13595 @itemx -multi_module
13596 @itemx -multiply_defined
13597 @itemx -multiply_defined_unused
13600 @itemx -no_dead_strip_inits_and_terms
13601 @itemx -nofixprebinding
13602 @itemx -nomultidefs
13604 @itemx -noseglinkedit
13605 @itemx -pagezero_size
13607 @itemx -prebind_all_twolevel_modules
13608 @itemx -private_bundle
13610 @itemx -read_only_relocs
13612 @itemx -sectobjectsymbols
13616 @itemx -sectobjectsymbols
13619 @itemx -segs_read_only_addr
13621 @itemx -segs_read_write_addr
13622 @itemx -seg_addr_table
13623 @itemx -seg_addr_table_filename
13624 @itemx -seglinkedit
13626 @itemx -segs_read_only_addr
13627 @itemx -segs_read_write_addr
13628 @itemx -single_module
13630 @itemx -sub_library
13632 @itemx -sub_umbrella
13633 @itemx -twolevel_namespace
13636 @itemx -unexported_symbols_list
13637 @itemx -weak_reference_mismatches
13638 @itemx -whatsloaded
13639 @opindex allowable_client
13640 @opindex client_name
13641 @opindex compatibility_version
13642 @opindex current_version
13643 @opindex dead_strip
13644 @opindex dependency-file
13645 @opindex dylib_file
13646 @opindex dylinker_install_name
13648 @opindex exported_symbols_list
13650 @opindex flat_namespace
13651 @opindex force_flat_namespace
13652 @opindex headerpad_max_install_names
13653 @opindex image_base
13655 @opindex install_name
13656 @opindex keep_private_externs
13657 @opindex multi_module
13658 @opindex multiply_defined
13659 @opindex multiply_defined_unused
13660 @opindex noall_load
13661 @opindex no_dead_strip_inits_and_terms
13662 @opindex nofixprebinding
13663 @opindex nomultidefs
13665 @opindex noseglinkedit
13666 @opindex pagezero_size
13668 @opindex prebind_all_twolevel_modules
13669 @opindex private_bundle
13670 @opindex read_only_relocs
13672 @opindex sectobjectsymbols
13675 @opindex sectcreate
13676 @opindex sectobjectsymbols
13679 @opindex segs_read_only_addr
13680 @opindex segs_read_write_addr
13681 @opindex seg_addr_table
13682 @opindex seg_addr_table_filename
13683 @opindex seglinkedit
13685 @opindex segs_read_only_addr
13686 @opindex segs_read_write_addr
13687 @opindex single_module
13689 @opindex sub_library
13690 @opindex sub_umbrella
13691 @opindex twolevel_namespace
13694 @opindex unexported_symbols_list
13695 @opindex weak_reference_mismatches
13696 @opindex whatsloaded
13697 These options are passed to the Darwin linker. The Darwin linker man page
13698 describes them in detail.
13701 @node DEC Alpha Options
13702 @subsection DEC Alpha Options
13704 These @samp{-m} options are defined for the DEC Alpha implementations:
13707 @item -mno-soft-float
13708 @itemx -msoft-float
13709 @opindex mno-soft-float
13710 @opindex msoft-float
13711 Use (do not use) the hardware floating-point instructions for
13712 floating-point operations. When @option{-msoft-float} is specified,
13713 functions in @file{libgcc.a} are used to perform floating-point
13714 operations. Unless they are replaced by routines that emulate the
13715 floating-point operations, or compiled in such a way as to call such
13716 emulations routines, these routines issue floating-point
13717 operations. If you are compiling for an Alpha without floating-point
13718 operations, you must ensure that the library is built so as not to call
13721 Note that Alpha implementations without floating-point operations are
13722 required to have floating-point registers.
13725 @itemx -mno-fp-regs
13727 @opindex mno-fp-regs
13728 Generate code that uses (does not use) the floating-point register set.
13729 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
13730 register set is not used, floating-point operands are passed in integer
13731 registers as if they were integers and floating-point results are passed
13732 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
13733 so any function with a floating-point argument or return value called by code
13734 compiled with @option{-mno-fp-regs} must also be compiled with that
13737 A typical use of this option is building a kernel that does not use,
13738 and hence need not save and restore, any floating-point registers.
13742 The Alpha architecture implements floating-point hardware optimized for
13743 maximum performance. It is mostly compliant with the IEEE floating-point
13744 standard. However, for full compliance, software assistance is
13745 required. This option generates code fully IEEE-compliant code
13746 @emph{except} that the @var{inexact-flag} is not maintained (see below).
13747 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
13748 defined during compilation. The resulting code is less efficient but is
13749 able to correctly support denormalized numbers and exceptional IEEE
13750 values such as not-a-number and plus/minus infinity. Other Alpha
13751 compilers call this option @option{-ieee_with_no_inexact}.
13753 @item -mieee-with-inexact
13754 @opindex mieee-with-inexact
13755 This is like @option{-mieee} except the generated code also maintains
13756 the IEEE @var{inexact-flag}. Turning on this option causes the
13757 generated code to implement fully-compliant IEEE math. In addition to
13758 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
13759 macro. On some Alpha implementations the resulting code may execute
13760 significantly slower than the code generated by default. Since there is
13761 very little code that depends on the @var{inexact-flag}, you should
13762 normally not specify this option. Other Alpha compilers call this
13763 option @option{-ieee_with_inexact}.
13765 @item -mfp-trap-mode=@var{trap-mode}
13766 @opindex mfp-trap-mode
13767 This option controls what floating-point related traps are enabled.
13768 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
13769 The trap mode can be set to one of four values:
13773 This is the default (normal) setting. The only traps that are enabled
13774 are the ones that cannot be disabled in software (e.g., division by zero
13778 In addition to the traps enabled by @samp{n}, underflow traps are enabled
13782 Like @samp{u}, but the instructions are marked to be safe for software
13783 completion (see Alpha architecture manual for details).
13786 Like @samp{su}, but inexact traps are enabled as well.
13789 @item -mfp-rounding-mode=@var{rounding-mode}
13790 @opindex mfp-rounding-mode
13791 Selects the IEEE rounding mode. Other Alpha compilers call this option
13792 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
13797 Normal IEEE rounding mode. Floating-point numbers are rounded towards
13798 the nearest machine number or towards the even machine number in case
13802 Round towards minus infinity.
13805 Chopped rounding mode. Floating-point numbers are rounded towards zero.
13808 Dynamic rounding mode. A field in the floating-point control register
13809 (@var{fpcr}, see Alpha architecture reference manual) controls the
13810 rounding mode in effect. The C library initializes this register for
13811 rounding towards plus infinity. Thus, unless your program modifies the
13812 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
13815 @item -mtrap-precision=@var{trap-precision}
13816 @opindex mtrap-precision
13817 In the Alpha architecture, floating-point traps are imprecise. This
13818 means without software assistance it is impossible to recover from a
13819 floating trap and program execution normally needs to be terminated.
13820 GCC can generate code that can assist operating system trap handlers
13821 in determining the exact location that caused a floating-point trap.
13822 Depending on the requirements of an application, different levels of
13823 precisions can be selected:
13827 Program precision. This option is the default and means a trap handler
13828 can only identify which program caused a floating-point exception.
13831 Function precision. The trap handler can determine the function that
13832 caused a floating-point exception.
13835 Instruction precision. The trap handler can determine the exact
13836 instruction that caused a floating-point exception.
13839 Other Alpha compilers provide the equivalent options called
13840 @option{-scope_safe} and @option{-resumption_safe}.
13842 @item -mieee-conformant
13843 @opindex mieee-conformant
13844 This option marks the generated code as IEEE conformant. You must not
13845 use this option unless you also specify @option{-mtrap-precision=i} and either
13846 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
13847 is to emit the line @samp{.eflag 48} in the function prologue of the
13848 generated assembly file.
13850 @item -mbuild-constants
13851 @opindex mbuild-constants
13852 Normally GCC examines a 32- or 64-bit integer constant to
13853 see if it can construct it from smaller constants in two or three
13854 instructions. If it cannot, it outputs the constant as a literal and
13855 generates code to load it from the data segment at run time.
13857 Use this option to require GCC to construct @emph{all} integer constants
13858 using code, even if it takes more instructions (the maximum is six).
13860 You typically use this option to build a shared library dynamic
13861 loader. Itself a shared library, it must relocate itself in memory
13862 before it can find the variables and constants in its own data segment.
13880 Indicate whether GCC should generate code to use the optional BWX,
13881 CIX, FIX and MAX instruction sets. The default is to use the instruction
13882 sets supported by the CPU type specified via @option{-mcpu=} option or that
13883 of the CPU on which GCC was built if none is specified.
13886 @itemx -mfloat-ieee
13887 @opindex mfloat-vax
13888 @opindex mfloat-ieee
13889 Generate code that uses (does not use) VAX F and G floating-point
13890 arithmetic instead of IEEE single and double precision.
13892 @item -mexplicit-relocs
13893 @itemx -mno-explicit-relocs
13894 @opindex mexplicit-relocs
13895 @opindex mno-explicit-relocs
13896 Older Alpha assemblers provided no way to generate symbol relocations
13897 except via assembler macros. Use of these macros does not allow
13898 optimal instruction scheduling. GNU binutils as of version 2.12
13899 supports a new syntax that allows the compiler to explicitly mark
13900 which relocations should apply to which instructions. This option
13901 is mostly useful for debugging, as GCC detects the capabilities of
13902 the assembler when it is built and sets the default accordingly.
13905 @itemx -mlarge-data
13906 @opindex msmall-data
13907 @opindex mlarge-data
13908 When @option{-mexplicit-relocs} is in effect, static data is
13909 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
13910 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
13911 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
13912 16-bit relocations off of the @code{$gp} register. This limits the
13913 size of the small data area to 64KB, but allows the variables to be
13914 directly accessed via a single instruction.
13916 The default is @option{-mlarge-data}. With this option the data area
13917 is limited to just below 2GB@. Programs that require more than 2GB of
13918 data must use @code{malloc} or @code{mmap} to allocate the data in the
13919 heap instead of in the program's data segment.
13921 When generating code for shared libraries, @option{-fpic} implies
13922 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
13925 @itemx -mlarge-text
13926 @opindex msmall-text
13927 @opindex mlarge-text
13928 When @option{-msmall-text} is used, the compiler assumes that the
13929 code of the entire program (or shared library) fits in 4MB, and is
13930 thus reachable with a branch instruction. When @option{-msmall-data}
13931 is used, the compiler can assume that all local symbols share the
13932 same @code{$gp} value, and thus reduce the number of instructions
13933 required for a function call from 4 to 1.
13935 The default is @option{-mlarge-text}.
13937 @item -mcpu=@var{cpu_type}
13939 Set the instruction set and instruction scheduling parameters for
13940 machine type @var{cpu_type}. You can specify either the @samp{EV}
13941 style name or the corresponding chip number. GCC supports scheduling
13942 parameters for the EV4, EV5 and EV6 family of processors and
13943 chooses the default values for the instruction set from the processor
13944 you specify. If you do not specify a processor type, GCC defaults
13945 to the processor on which the compiler was built.
13947 Supported values for @var{cpu_type} are
13953 Schedules as an EV4 and has no instruction set extensions.
13957 Schedules as an EV5 and has no instruction set extensions.
13961 Schedules as an EV5 and supports the BWX extension.
13966 Schedules as an EV5 and supports the BWX and MAX extensions.
13970 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
13974 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
13977 Native toolchains also support the value @samp{native},
13978 which selects the best architecture option for the host processor.
13979 @option{-mcpu=native} has no effect if GCC does not recognize
13982 @item -mtune=@var{cpu_type}
13984 Set only the instruction scheduling parameters for machine type
13985 @var{cpu_type}. The instruction set is not changed.
13987 Native toolchains also support the value @samp{native},
13988 which selects the best architecture option for the host processor.
13989 @option{-mtune=native} has no effect if GCC does not recognize
13992 @item -mmemory-latency=@var{time}
13993 @opindex mmemory-latency
13994 Sets the latency the scheduler should assume for typical memory
13995 references as seen by the application. This number is highly
13996 dependent on the memory access patterns used by the application
13997 and the size of the external cache on the machine.
13999 Valid options for @var{time} are
14003 A decimal number representing clock cycles.
14009 The compiler contains estimates of the number of clock cycles for
14010 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14011 (also called Dcache, Scache, and Bcache), as well as to main memory.
14012 Note that L3 is only valid for EV5.
14018 @subsection FR30 Options
14019 @cindex FR30 Options
14021 These options are defined specifically for the FR30 port.
14025 @item -msmall-model
14026 @opindex msmall-model
14027 Use the small address space model. This can produce smaller code, but
14028 it does assume that all symbolic values and addresses fit into a
14033 Assume that runtime support has been provided and so there is no need
14034 to include the simulator library (@file{libsim.a}) on the linker
14040 @subsection FRV Options
14041 @cindex FRV Options
14047 Only use the first 32 general-purpose registers.
14052 Use all 64 general-purpose registers.
14057 Use only the first 32 floating-point registers.
14062 Use all 64 floating-point registers.
14065 @opindex mhard-float
14067 Use hardware instructions for floating-point operations.
14070 @opindex msoft-float
14072 Use library routines for floating-point operations.
14077 Dynamically allocate condition code registers.
14082 Do not try to dynamically allocate condition code registers, only
14083 use @code{icc0} and @code{fcc0}.
14088 Change ABI to use double word insns.
14093 Do not use double word instructions.
14098 Use floating-point double instructions.
14101 @opindex mno-double
14103 Do not use floating-point double instructions.
14108 Use media instructions.
14113 Do not use media instructions.
14118 Use multiply and add/subtract instructions.
14121 @opindex mno-muladd
14123 Do not use multiply and add/subtract instructions.
14128 Select the FDPIC ABI, which uses function descriptors to represent
14129 pointers to functions. Without any PIC/PIE-related options, it
14130 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
14131 assumes GOT entries and small data are within a 12-bit range from the
14132 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
14133 are computed with 32 bits.
14134 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14137 @opindex minline-plt
14139 Enable inlining of PLT entries in function calls to functions that are
14140 not known to bind locally. It has no effect without @option{-mfdpic}.
14141 It's enabled by default if optimizing for speed and compiling for
14142 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
14143 optimization option such as @option{-O3} or above is present in the
14149 Assume a large TLS segment when generating thread-local code.
14154 Do not assume a large TLS segment when generating thread-local code.
14159 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
14160 that is known to be in read-only sections. It's enabled by default,
14161 except for @option{-fpic} or @option{-fpie}: even though it may help
14162 make the global offset table smaller, it trades 1 instruction for 4.
14163 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
14164 one of which may be shared by multiple symbols, and it avoids the need
14165 for a GOT entry for the referenced symbol, so it's more likely to be a
14166 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
14168 @item -multilib-library-pic
14169 @opindex multilib-library-pic
14171 Link with the (library, not FD) pic libraries. It's implied by
14172 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
14173 @option{-fpic} without @option{-mfdpic}. You should never have to use
14177 @opindex mlinked-fp
14179 Follow the EABI requirement of always creating a frame pointer whenever
14180 a stack frame is allocated. This option is enabled by default and can
14181 be disabled with @option{-mno-linked-fp}.
14184 @opindex mlong-calls
14186 Use indirect addressing to call functions outside the current
14187 compilation unit. This allows the functions to be placed anywhere
14188 within the 32-bit address space.
14190 @item -malign-labels
14191 @opindex malign-labels
14193 Try to align labels to an 8-byte boundary by inserting NOPs into the
14194 previous packet. This option only has an effect when VLIW packing
14195 is enabled. It doesn't create new packets; it merely adds NOPs to
14198 @item -mlibrary-pic
14199 @opindex mlibrary-pic
14201 Generate position-independent EABI code.
14206 Use only the first four media accumulator registers.
14211 Use all eight media accumulator registers.
14216 Pack VLIW instructions.
14221 Do not pack VLIW instructions.
14224 @opindex mno-eflags
14226 Do not mark ABI switches in e_flags.
14229 @opindex mcond-move
14231 Enable the use of conditional-move instructions (default).
14233 This switch is mainly for debugging the compiler and will likely be removed
14234 in a future version.
14236 @item -mno-cond-move
14237 @opindex mno-cond-move
14239 Disable the use of conditional-move instructions.
14241 This switch is mainly for debugging the compiler and will likely be removed
14242 in a future version.
14247 Enable the use of conditional set instructions (default).
14249 This switch is mainly for debugging the compiler and will likely be removed
14250 in a future version.
14255 Disable the use of conditional set instructions.
14257 This switch is mainly for debugging the compiler and will likely be removed
14258 in a future version.
14261 @opindex mcond-exec
14263 Enable the use of conditional execution (default).
14265 This switch is mainly for debugging the compiler and will likely be removed
14266 in a future version.
14268 @item -mno-cond-exec
14269 @opindex mno-cond-exec
14271 Disable the use of conditional execution.
14273 This switch is mainly for debugging the compiler and will likely be removed
14274 in a future version.
14276 @item -mvliw-branch
14277 @opindex mvliw-branch
14279 Run a pass to pack branches into VLIW instructions (default).
14281 This switch is mainly for debugging the compiler and will likely be removed
14282 in a future version.
14284 @item -mno-vliw-branch
14285 @opindex mno-vliw-branch
14287 Do not run a pass to pack branches into VLIW instructions.
14289 This switch is mainly for debugging the compiler and will likely be removed
14290 in a future version.
14292 @item -mmulti-cond-exec
14293 @opindex mmulti-cond-exec
14295 Enable optimization of @code{&&} and @code{||} in conditional execution
14298 This switch is mainly for debugging the compiler and will likely be removed
14299 in a future version.
14301 @item -mno-multi-cond-exec
14302 @opindex mno-multi-cond-exec
14304 Disable optimization of @code{&&} and @code{||} in conditional execution.
14306 This switch is mainly for debugging the compiler and will likely be removed
14307 in a future version.
14309 @item -mnested-cond-exec
14310 @opindex mnested-cond-exec
14312 Enable nested conditional execution optimizations (default).
14314 This switch is mainly for debugging the compiler and will likely be removed
14315 in a future version.
14317 @item -mno-nested-cond-exec
14318 @opindex mno-nested-cond-exec
14320 Disable nested conditional execution optimizations.
14322 This switch is mainly for debugging the compiler and will likely be removed
14323 in a future version.
14325 @item -moptimize-membar
14326 @opindex moptimize-membar
14328 This switch removes redundant @code{membar} instructions from the
14329 compiler-generated code. It is enabled by default.
14331 @item -mno-optimize-membar
14332 @opindex mno-optimize-membar
14334 This switch disables the automatic removal of redundant @code{membar}
14335 instructions from the generated code.
14337 @item -mtomcat-stats
14338 @opindex mtomcat-stats
14340 Cause gas to print out tomcat statistics.
14342 @item -mcpu=@var{cpu}
14345 Select the processor type for which to generate code. Possible values are
14346 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
14347 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
14351 @node GNU/Linux Options
14352 @subsection GNU/Linux Options
14354 These @samp{-m} options are defined for GNU/Linux targets:
14359 Use the GNU C library. This is the default except
14360 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
14364 Use uClibc C library. This is the default on
14365 @samp{*-*-linux-*uclibc*} targets.
14369 Use Bionic C library. This is the default on
14370 @samp{*-*-linux-*android*} targets.
14374 Compile code compatible with Android platform. This is the default on
14375 @samp{*-*-linux-*android*} targets.
14377 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
14378 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
14379 this option makes the GCC driver pass Android-specific options to the linker.
14380 Finally, this option causes the preprocessor macro @code{__ANDROID__}
14383 @item -tno-android-cc
14384 @opindex tno-android-cc
14385 Disable compilation effects of @option{-mandroid}, i.e., do not enable
14386 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
14387 @option{-fno-rtti} by default.
14389 @item -tno-android-ld
14390 @opindex tno-android-ld
14391 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
14392 linking options to the linker.
14396 @node H8/300 Options
14397 @subsection H8/300 Options
14399 These @samp{-m} options are defined for the H8/300 implementations:
14404 Shorten some address references at link time, when possible; uses the
14405 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
14406 ld, Using ld}, for a fuller description.
14410 Generate code for the H8/300H@.
14414 Generate code for the H8S@.
14418 Generate code for the H8S and H8/300H in the normal mode. This switch
14419 must be used either with @option{-mh} or @option{-ms}.
14423 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
14427 Extended registers are stored on stack before execution of function
14428 with monitor attribute. Default option is @option{-mexr}.
14429 This option is valid only for H8S targets.
14433 Extended registers are not stored on stack before execution of function
14434 with monitor attribute. Default option is @option{-mno-exr}.
14435 This option is valid only for H8S targets.
14439 Make @code{int} data 32 bits by default.
14442 @opindex malign-300
14443 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
14444 The default for the H8/300H and H8S is to align longs and floats on
14446 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
14447 This option has no effect on the H8/300.
14451 @subsection HPPA Options
14452 @cindex HPPA Options
14454 These @samp{-m} options are defined for the HPPA family of computers:
14457 @item -march=@var{architecture-type}
14459 Generate code for the specified architecture. The choices for
14460 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
14461 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
14462 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
14463 architecture option for your machine. Code compiled for lower numbered
14464 architectures runs on higher numbered architectures, but not the
14467 @item -mpa-risc-1-0
14468 @itemx -mpa-risc-1-1
14469 @itemx -mpa-risc-2-0
14470 @opindex mpa-risc-1-0
14471 @opindex mpa-risc-1-1
14472 @opindex mpa-risc-2-0
14473 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
14475 @item -mjump-in-delay
14476 @opindex mjump-in-delay
14477 Fill delay slots of function calls with unconditional jump instructions
14478 by modifying the return pointer for the function call to be the target
14479 of the conditional jump.
14481 @item -mdisable-fpregs
14482 @opindex mdisable-fpregs
14483 Prevent floating-point registers from being used in any manner. This is
14484 necessary for compiling kernels that perform lazy context switching of
14485 floating-point registers. If you use this option and attempt to perform
14486 floating-point operations, the compiler aborts.
14488 @item -mdisable-indexing
14489 @opindex mdisable-indexing
14490 Prevent the compiler from using indexing address modes. This avoids some
14491 rather obscure problems when compiling MIG generated code under MACH@.
14493 @item -mno-space-regs
14494 @opindex mno-space-regs
14495 Generate code that assumes the target has no space registers. This allows
14496 GCC to generate faster indirect calls and use unscaled index address modes.
14498 Such code is suitable for level 0 PA systems and kernels.
14500 @item -mfast-indirect-calls
14501 @opindex mfast-indirect-calls
14502 Generate code that assumes calls never cross space boundaries. This
14503 allows GCC to emit code that performs faster indirect calls.
14505 This option does not work in the presence of shared libraries or nested
14508 @item -mfixed-range=@var{register-range}
14509 @opindex mfixed-range
14510 Generate code treating the given register range as fixed registers.
14511 A fixed register is one that the register allocator cannot use. This is
14512 useful when compiling kernel code. A register range is specified as
14513 two registers separated by a dash. Multiple register ranges can be
14514 specified separated by a comma.
14516 @item -mlong-load-store
14517 @opindex mlong-load-store
14518 Generate 3-instruction load and store sequences as sometimes required by
14519 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
14522 @item -mportable-runtime
14523 @opindex mportable-runtime
14524 Use the portable calling conventions proposed by HP for ELF systems.
14528 Enable the use of assembler directives only GAS understands.
14530 @item -mschedule=@var{cpu-type}
14532 Schedule code according to the constraints for the machine type
14533 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
14534 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
14535 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
14536 proper scheduling option for your machine. The default scheduling is
14540 @opindex mlinker-opt
14541 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
14542 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
14543 linkers in which they give bogus error messages when linking some programs.
14546 @opindex msoft-float
14547 Generate output containing library calls for floating point.
14548 @strong{Warning:} the requisite libraries are not available for all HPPA
14549 targets. Normally the facilities of the machine's usual C compiler are
14550 used, but this cannot be done directly in cross-compilation. You must make
14551 your own arrangements to provide suitable library functions for
14554 @option{-msoft-float} changes the calling convention in the output file;
14555 therefore, it is only useful if you compile @emph{all} of a program with
14556 this option. In particular, you need to compile @file{libgcc.a}, the
14557 library that comes with GCC, with @option{-msoft-float} in order for
14562 Generate the predefine, @code{_SIO}, for server IO@. The default is
14563 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
14564 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
14565 options are available under HP-UX and HI-UX@.
14569 Use options specific to GNU @command{ld}.
14570 This passes @option{-shared} to @command{ld} when
14571 building a shared library. It is the default when GCC is configured,
14572 explicitly or implicitly, with the GNU linker. This option does not
14573 affect which @command{ld} is called; it only changes what parameters
14574 are passed to that @command{ld}.
14575 The @command{ld} that is called is determined by the
14576 @option{--with-ld} configure option, GCC's program search path, and
14577 finally by the user's @env{PATH}. The linker used by GCC can be printed
14578 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
14579 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
14583 Use options specific to HP @command{ld}.
14584 This passes @option{-b} to @command{ld} when building
14585 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
14586 links. It is the default when GCC is configured, explicitly or
14587 implicitly, with the HP linker. This option does not affect
14588 which @command{ld} is called; it only changes what parameters are passed to that
14590 The @command{ld} that is called is determined by the @option{--with-ld}
14591 configure option, GCC's program search path, and finally by the user's
14592 @env{PATH}. The linker used by GCC can be printed using @samp{which
14593 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
14594 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
14597 @opindex mno-long-calls
14598 Generate code that uses long call sequences. This ensures that a call
14599 is always able to reach linker generated stubs. The default is to generate
14600 long calls only when the distance from the call site to the beginning
14601 of the function or translation unit, as the case may be, exceeds a
14602 predefined limit set by the branch type being used. The limits for
14603 normal calls are 7,600,000 and 240,000 bytes, respectively for the
14604 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
14607 Distances are measured from the beginning of functions when using the
14608 @option{-ffunction-sections} option, or when using the @option{-mgas}
14609 and @option{-mno-portable-runtime} options together under HP-UX with
14612 It is normally not desirable to use this option as it degrades
14613 performance. However, it may be useful in large applications,
14614 particularly when partial linking is used to build the application.
14616 The types of long calls used depends on the capabilities of the
14617 assembler and linker, and the type of code being generated. The
14618 impact on systems that support long absolute calls, and long pic
14619 symbol-difference or pc-relative calls should be relatively small.
14620 However, an indirect call is used on 32-bit ELF systems in pic code
14621 and it is quite long.
14623 @item -munix=@var{unix-std}
14625 Generate compiler predefines and select a startfile for the specified
14626 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
14627 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
14628 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
14629 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
14630 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
14633 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
14634 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
14635 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
14636 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
14637 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
14638 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
14640 It is @emph{important} to note that this option changes the interfaces
14641 for various library routines. It also affects the operational behavior
14642 of the C library. Thus, @emph{extreme} care is needed in using this
14645 Library code that is intended to operate with more than one UNIX
14646 standard must test, set and restore the variable @var{__xpg4_extended_mask}
14647 as appropriate. Most GNU software doesn't provide this capability.
14651 Suppress the generation of link options to search libdld.sl when the
14652 @option{-static} option is specified on HP-UX 10 and later.
14656 The HP-UX implementation of setlocale in libc has a dependency on
14657 libdld.sl. There isn't an archive version of libdld.sl. Thus,
14658 when the @option{-static} option is specified, special link options
14659 are needed to resolve this dependency.
14661 On HP-UX 10 and later, the GCC driver adds the necessary options to
14662 link with libdld.sl when the @option{-static} option is specified.
14663 This causes the resulting binary to be dynamic. On the 64-bit port,
14664 the linkers generate dynamic binaries by default in any case. The
14665 @option{-nolibdld} option can be used to prevent the GCC driver from
14666 adding these link options.
14670 Add support for multithreading with the @dfn{dce thread} library
14671 under HP-UX@. This option sets flags for both the preprocessor and
14675 @node i386 and x86-64 Options
14676 @subsection Intel 386 and AMD x86-64 Options
14677 @cindex i386 Options
14678 @cindex x86-64 Options
14679 @cindex Intel 386 Options
14680 @cindex AMD x86-64 Options
14682 These @samp{-m} options are defined for the i386 and x86-64 family of
14687 @item -march=@var{cpu-type}
14689 Generate instructions for the machine type @var{cpu-type}. In contrast to
14690 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
14691 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
14692 to generate code that may not run at all on processors other than the one
14693 indicated. Specifying @option{-march=@var{cpu-type}} implies
14694 @option{-mtune=@var{cpu-type}}.
14696 The choices for @var{cpu-type} are:
14700 This selects the CPU to generate code for at compilation time by determining
14701 the processor type of the compiling machine. Using @option{-march=native}
14702 enables all instruction subsets supported by the local machine (hence
14703 the result might not run on different machines). Using @option{-mtune=native}
14704 produces code optimized for the local machine under the constraints
14705 of the selected instruction set.
14708 Original Intel i386 CPU@.
14711 Intel i486 CPU@. (No scheduling is implemented for this chip.)
14715 Intel Pentium CPU with no MMX support.
14718 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
14721 Intel Pentium Pro CPU@.
14724 When used with @option{-march}, the Pentium Pro
14725 instruction set is used, so the code runs on all i686 family chips.
14726 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
14729 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
14734 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
14738 Intel Pentium M; low-power version of Intel Pentium III CPU
14739 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
14743 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
14746 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
14750 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
14751 SSE2 and SSE3 instruction set support.
14754 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
14755 instruction set support.
14758 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14759 SSE4.1, SSE4.2 and POPCNT instruction set support.
14762 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14763 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
14766 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14767 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
14770 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14771 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
14772 instruction set support.
14775 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
14776 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
14777 BMI, BMI2 and F16C instruction set support.
14780 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
14781 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
14782 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
14785 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
14786 instruction set support.
14789 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
14790 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
14793 AMD K6 CPU with MMX instruction set support.
14797 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
14800 @itemx athlon-tbird
14801 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
14807 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
14808 instruction set support.
14814 Processors based on the AMD K8 core with x86-64 instruction set support,
14815 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
14816 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
14817 instruction set extensions.)
14820 @itemx opteron-sse3
14821 @itemx athlon64-sse3
14822 Improved versions of AMD K8 cores with SSE3 instruction set support.
14826 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
14827 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
14828 instruction set extensions.)
14831 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
14832 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
14833 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
14835 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
14836 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
14837 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
14840 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
14841 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
14842 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
14843 64-bit instruction set extensions.
14845 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
14846 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
14847 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
14848 SSE4.2, ABM and 64-bit instruction set extensions.
14851 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
14852 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
14853 instruction set extensions.)
14856 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
14857 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
14858 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
14861 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
14865 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
14866 instruction set support.
14869 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
14870 implemented for this chip.)
14873 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
14875 implemented for this chip.)
14878 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
14881 @item -mtune=@var{cpu-type}
14883 Tune to @var{cpu-type} everything applicable about the generated code, except
14884 for the ABI and the set of available instructions.
14885 While picking a specific @var{cpu-type} schedules things appropriately
14886 for that particular chip, the compiler does not generate any code that
14887 cannot run on the default machine type unless you use a
14888 @option{-march=@var{cpu-type}} option.
14889 For example, if GCC is configured for i686-pc-linux-gnu
14890 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
14891 but still runs on i686 machines.
14893 The choices for @var{cpu-type} are the same as for @option{-march}.
14894 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
14898 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
14899 If you know the CPU on which your code will run, then you should use
14900 the corresponding @option{-mtune} or @option{-march} option instead of
14901 @option{-mtune=generic}. But, if you do not know exactly what CPU users
14902 of your application will have, then you should use this option.
14904 As new processors are deployed in the marketplace, the behavior of this
14905 option will change. Therefore, if you upgrade to a newer version of
14906 GCC, code generation controlled by this option will change to reflect
14908 that are most common at the time that version of GCC is released.
14910 There is no @option{-march=generic} option because @option{-march}
14911 indicates the instruction set the compiler can use, and there is no
14912 generic instruction set applicable to all processors. In contrast,
14913 @option{-mtune} indicates the processor (or, in this case, collection of
14914 processors) for which the code is optimized.
14917 Produce code optimized for the most current Intel processors, which are
14918 Haswell and Silvermont for this version of GCC. If you know the CPU
14919 on which your code will run, then you should use the corresponding
14920 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
14921 But, if you want your application performs better on both Haswell and
14922 Silvermont, then you should use this option.
14924 As new Intel processors are deployed in the marketplace, the behavior of
14925 this option will change. Therefore, if you upgrade to a newer version of
14926 GCC, code generation controlled by this option will change to reflect
14927 the most current Intel processors at the time that version of GCC is
14930 There is no @option{-march=intel} option because @option{-march} indicates
14931 the instruction set the compiler can use, and there is no common
14932 instruction set applicable to all processors. In contrast,
14933 @option{-mtune} indicates the processor (or, in this case, collection of
14934 processors) for which the code is optimized.
14937 @item -mcpu=@var{cpu-type}
14939 A deprecated synonym for @option{-mtune}.
14941 @item -mfpmath=@var{unit}
14943 Generate floating-point arithmetic for selected unit @var{unit}. The choices
14944 for @var{unit} are:
14948 Use the standard 387 floating-point coprocessor present on the majority of chips and
14949 emulated otherwise. Code compiled with this option runs almost everywhere.
14950 The temporary results are computed in 80-bit precision instead of the precision
14951 specified by the type, resulting in slightly different results compared to most
14952 of other chips. See @option{-ffloat-store} for more detailed description.
14954 This is the default choice for i386 compiler.
14957 Use scalar floating-point instructions present in the SSE instruction set.
14958 This instruction set is supported by Pentium III and newer chips,
14959 and in the AMD line
14960 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
14961 instruction set supports only single-precision arithmetic, thus the double and
14962 extended-precision arithmetic are still done using 387. A later version, present
14963 only in Pentium 4 and AMD x86-64 chips, supports double-precision
14966 For the i386 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
14967 or @option{-msse2} switches to enable SSE extensions and make this option
14968 effective. For the x86-64 compiler, these extensions are enabled by default.
14970 The resulting code should be considerably faster in the majority of cases and avoid
14971 the numerical instability problems of 387 code, but may break some existing
14972 code that expects temporaries to be 80 bits.
14974 This is the default choice for the x86-64 compiler.
14979 Attempt to utilize both instruction sets at once. This effectively doubles the
14980 amount of available registers, and on chips with separate execution units for
14981 387 and SSE the execution resources too. Use this option with care, as it is
14982 still experimental, because the GCC register allocator does not model separate
14983 functional units well, resulting in unstable performance.
14986 @item -masm=@var{dialect}
14987 @opindex masm=@var{dialect}
14988 Output assembly instructions using selected @var{dialect}. Supported
14989 choices are @samp{intel} or @samp{att} (the default). Darwin does
14990 not support @samp{intel}.
14993 @itemx -mno-ieee-fp
14995 @opindex mno-ieee-fp
14996 Control whether or not the compiler uses IEEE floating-point
14997 comparisons. These correctly handle the case where the result of a
14998 comparison is unordered.
15001 @opindex msoft-float
15002 Generate output containing library calls for floating point.
15004 @strong{Warning:} the requisite libraries are not part of GCC@.
15005 Normally the facilities of the machine's usual C compiler are used, but
15006 this can't be done directly in cross-compilation. You must make your
15007 own arrangements to provide suitable library functions for
15010 On machines where a function returns floating-point results in the 80387
15011 register stack, some floating-point opcodes may be emitted even if
15012 @option{-msoft-float} is used.
15014 @item -mno-fp-ret-in-387
15015 @opindex mno-fp-ret-in-387
15016 Do not use the FPU registers for return values of functions.
15018 The usual calling convention has functions return values of types
15019 @code{float} and @code{double} in an FPU register, even if there
15020 is no FPU@. The idea is that the operating system should emulate
15023 The option @option{-mno-fp-ret-in-387} causes such values to be returned
15024 in ordinary CPU registers instead.
15026 @item -mno-fancy-math-387
15027 @opindex mno-fancy-math-387
15028 Some 387 emulators do not support the @code{sin}, @code{cos} and
15029 @code{sqrt} instructions for the 387. Specify this option to avoid
15030 generating those instructions. This option is the default on FreeBSD,
15031 OpenBSD and NetBSD@. This option is overridden when @option{-march}
15032 indicates that the target CPU always has an FPU and so the
15033 instruction does not need emulation. These
15034 instructions are not generated unless you also use the
15035 @option{-funsafe-math-optimizations} switch.
15037 @item -malign-double
15038 @itemx -mno-align-double
15039 @opindex malign-double
15040 @opindex mno-align-double
15041 Control whether GCC aligns @code{double}, @code{long double}, and
15042 @code{long long} variables on a two-word boundary or a one-word
15043 boundary. Aligning @code{double} variables on a two-word boundary
15044 produces code that runs somewhat faster on a Pentium at the
15045 expense of more memory.
15047 On x86-64, @option{-malign-double} is enabled by default.
15049 @strong{Warning:} if you use the @option{-malign-double} switch,
15050 structures containing the above types are aligned differently than
15051 the published application binary interface specifications for the 386
15052 and are not binary compatible with structures in code compiled
15053 without that switch.
15055 @item -m96bit-long-double
15056 @itemx -m128bit-long-double
15057 @opindex m96bit-long-double
15058 @opindex m128bit-long-double
15059 These switches control the size of @code{long double} type. The i386
15060 application binary interface specifies the size to be 96 bits,
15061 so @option{-m96bit-long-double} is the default in 32-bit mode.
15063 Modern architectures (Pentium and newer) prefer @code{long double}
15064 to be aligned to an 8- or 16-byte boundary. In arrays or structures
15065 conforming to the ABI, this is not possible. So specifying
15066 @option{-m128bit-long-double} aligns @code{long double}
15067 to a 16-byte boundary by padding the @code{long double} with an additional
15070 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
15071 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
15073 Notice that neither of these options enable any extra precision over the x87
15074 standard of 80 bits for a @code{long double}.
15076 @strong{Warning:} if you override the default value for your target ABI, this
15077 changes the size of
15078 structures and arrays containing @code{long double} variables,
15079 as well as modifying the function calling convention for functions taking
15080 @code{long double}. Hence they are not binary-compatible
15081 with code compiled without that switch.
15083 @item -mlong-double-64
15084 @itemx -mlong-double-80
15085 @itemx -mlong-double-128
15086 @opindex mlong-double-64
15087 @opindex mlong-double-80
15088 @opindex mlong-double-128
15089 These switches control the size of @code{long double} type. A size
15090 of 64 bits makes the @code{long double} type equivalent to the @code{double}
15091 type. This is the default for 32-bit Bionic C library. A size
15092 of 128 bits makes the @code{long double} type equivalent to the
15093 @code{__float128} type. This is the default for 64-bit Bionic C library.
15095 @strong{Warning:} if you override the default value for your target ABI, this
15096 changes the size of
15097 structures and arrays containing @code{long double} variables,
15098 as well as modifying the function calling convention for functions taking
15099 @code{long double}. Hence they are not binary-compatible
15100 with code compiled without that switch.
15102 @item -mlarge-data-threshold=@var{threshold}
15103 @opindex mlarge-data-threshold
15104 When @option{-mcmodel=medium} is specified, data objects larger than
15105 @var{threshold} are placed in the large data section. This value must be the
15106 same across all objects linked into the binary, and defaults to 65535.
15110 Use a different function-calling convention, in which functions that
15111 take a fixed number of arguments return with the @code{ret @var{num}}
15112 instruction, which pops their arguments while returning. This saves one
15113 instruction in the caller since there is no need to pop the arguments
15116 You can specify that an individual function is called with this calling
15117 sequence with the function attribute @samp{stdcall}. You can also
15118 override the @option{-mrtd} option by using the function attribute
15119 @samp{cdecl}. @xref{Function Attributes}.
15121 @strong{Warning:} this calling convention is incompatible with the one
15122 normally used on Unix, so you cannot use it if you need to call
15123 libraries compiled with the Unix compiler.
15125 Also, you must provide function prototypes for all functions that
15126 take variable numbers of arguments (including @code{printf});
15127 otherwise incorrect code is generated for calls to those
15130 In addition, seriously incorrect code results if you call a
15131 function with too many arguments. (Normally, extra arguments are
15132 harmlessly ignored.)
15134 @item -mregparm=@var{num}
15136 Control how many registers are used to pass integer arguments. By
15137 default, no registers are used to pass arguments, and at most 3
15138 registers can be used. You can control this behavior for a specific
15139 function by using the function attribute @samp{regparm}.
15140 @xref{Function Attributes}.
15142 @strong{Warning:} if you use this switch, and
15143 @var{num} is nonzero, then you must build all modules with the same
15144 value, including any libraries. This includes the system libraries and
15148 @opindex msseregparm
15149 Use SSE register passing conventions for float and double arguments
15150 and return values. You can control this behavior for a specific
15151 function by using the function attribute @samp{sseregparm}.
15152 @xref{Function Attributes}.
15154 @strong{Warning:} if you use this switch then you must build all
15155 modules with the same value, including any libraries. This includes
15156 the system libraries and startup modules.
15158 @item -mvect8-ret-in-mem
15159 @opindex mvect8-ret-in-mem
15160 Return 8-byte vectors in memory instead of MMX registers. This is the
15161 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
15162 Studio compilers until version 12. Later compiler versions (starting
15163 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
15164 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
15165 you need to remain compatible with existing code produced by those
15166 previous compiler versions or older versions of GCC@.
15175 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
15176 is specified, the significands of results of floating-point operations are
15177 rounded to 24 bits (single precision); @option{-mpc64} rounds the
15178 significands of results of floating-point operations to 53 bits (double
15179 precision) and @option{-mpc80} rounds the significands of results of
15180 floating-point operations to 64 bits (extended double precision), which is
15181 the default. When this option is used, floating-point operations in higher
15182 precisions are not available to the programmer without setting the FPU
15183 control word explicitly.
15185 Setting the rounding of floating-point operations to less than the default
15186 80 bits can speed some programs by 2% or more. Note that some mathematical
15187 libraries assume that extended-precision (80-bit) floating-point operations
15188 are enabled by default; routines in such libraries could suffer significant
15189 loss of accuracy, typically through so-called ``catastrophic cancellation'',
15190 when this option is used to set the precision to less than extended precision.
15192 @item -mstackrealign
15193 @opindex mstackrealign
15194 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
15195 option generates an alternate prologue and epilogue that realigns the
15196 run-time stack if necessary. This supports mixing legacy codes that keep
15197 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
15198 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
15199 applicable to individual functions.
15201 @item -mpreferred-stack-boundary=@var{num}
15202 @opindex mpreferred-stack-boundary
15203 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
15204 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
15205 the default is 4 (16 bytes or 128 bits).
15207 @strong{Warning:} When generating code for the x86-64 architecture with
15208 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
15209 used to keep the stack boundary aligned to 8 byte boundary. Since
15210 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
15211 intended to be used in controlled environment where stack space is
15212 important limitation. This option will lead to wrong code when functions
15213 compiled with 16 byte stack alignment (such as functions from a standard
15214 library) are called with misaligned stack. In this case, SSE
15215 instructions may lead to misaligned memory access traps. In addition,
15216 variable arguments will be handled incorrectly for 16 byte aligned
15217 objects (including x87 long double and __int128), leading to wrong
15218 results. You must build all modules with
15219 @option{-mpreferred-stack-boundary=3}, including any libraries. This
15220 includes the system libraries and startup modules.
15222 @item -mincoming-stack-boundary=@var{num}
15223 @opindex mincoming-stack-boundary
15224 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
15225 boundary. If @option{-mincoming-stack-boundary} is not specified,
15226 the one specified by @option{-mpreferred-stack-boundary} is used.
15228 On Pentium and Pentium Pro, @code{double} and @code{long double} values
15229 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
15230 suffer significant run time performance penalties. On Pentium III, the
15231 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
15232 properly if it is not 16-byte aligned.
15234 To ensure proper alignment of this values on the stack, the stack boundary
15235 must be as aligned as that required by any value stored on the stack.
15236 Further, every function must be generated such that it keeps the stack
15237 aligned. Thus calling a function compiled with a higher preferred
15238 stack boundary from a function compiled with a lower preferred stack
15239 boundary most likely misaligns the stack. It is recommended that
15240 libraries that use callbacks always use the default setting.
15242 This extra alignment does consume extra stack space, and generally
15243 increases code size. Code that is sensitive to stack space usage, such
15244 as embedded systems and operating system kernels, may want to reduce the
15245 preferred alignment to @option{-mpreferred-stack-boundary=2}.
15269 @itemx -mno-avx512f
15272 @itemx -mno-avx512pf
15274 @itemx -mno-avx512er
15276 @itemx -mno-avx512cd
15285 @itemx -mno-fsgsbase
15292 @itemx -mprefetchwt1
15293 @itemx -mno-prefetchwt1
15327 These switches enable or disable the use of instructions in the MMX, SSE,
15328 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
15329 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
15330 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, or 3DNow!@:
15331 extended instruction sets.
15332 These extensions are also available as built-in functions: see
15333 @ref{X86 Built-in Functions}, for details of the functions enabled and
15334 disabled by these switches.
15336 To generate SSE/SSE2 instructions automatically from floating-point
15337 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
15339 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
15340 generates new AVX instructions or AVX equivalence for all SSEx instructions
15343 These options enable GCC to use these extended instructions in
15344 generated code, even without @option{-mfpmath=sse}. Applications that
15345 perform run-time CPU detection must compile separate files for each
15346 supported architecture, using the appropriate flags. In particular,
15347 the file containing the CPU detection code should be compiled without
15350 @item -mdump-tune-features
15351 @opindex mdump-tune-features
15352 This option instructs GCC to dump the names of the x86 performance
15353 tuning features and default settings. The names can be used in
15354 @option{-mtune-ctrl=@var{feature-list}}.
15356 @item -mtune-ctrl=@var{feature-list}
15357 @opindex mtune-ctrl=@var{feature-list}
15358 This option is used to do fine grain control of x86 code generation features.
15359 @var{feature-list} is a comma separated list of @var{feature} names. See also
15360 @option{-mdump-tune-features}. When specified, the @var{feature} will be turned
15361 on if it is not preceded with @code{^}, otherwise, it will be turned off.
15362 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
15363 developers. Using it may lead to code paths not covered by testing and can
15364 potentially result in compiler ICEs or runtime errors.
15367 @opindex mno-default
15368 This option instructs GCC to turn off all tunable features. See also
15369 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
15373 This option instructs GCC to emit a @code{cld} instruction in the prologue
15374 of functions that use string instructions. String instructions depend on
15375 the DF flag to select between autoincrement or autodecrement mode. While the
15376 ABI specifies the DF flag to be cleared on function entry, some operating
15377 systems violate this specification by not clearing the DF flag in their
15378 exception dispatchers. The exception handler can be invoked with the DF flag
15379 set, which leads to wrong direction mode when string instructions are used.
15380 This option can be enabled by default on 32-bit x86 targets by configuring
15381 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
15382 instructions can be suppressed with the @option{-mno-cld} compiler option
15386 @opindex mvzeroupper
15387 This option instructs GCC to emit a @code{vzeroupper} instruction
15388 before a transfer of control flow out of the function to minimize
15389 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
15392 @item -mprefer-avx128
15393 @opindex mprefer-avx128
15394 This option instructs GCC to use 128-bit AVX instructions instead of
15395 256-bit AVX instructions in the auto-vectorizer.
15399 This option enables GCC to generate @code{CMPXCHG16B} instructions.
15400 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
15401 (or oword) data types.
15402 This is useful for high-resolution counters that can be updated
15403 by multiple processors (or cores). This instruction is generated as part of
15404 atomic built-in functions: see @ref{__sync Builtins} or
15405 @ref{__atomic Builtins} for details.
15409 This option enables generation of @code{SAHF} instructions in 64-bit code.
15410 Early Intel Pentium 4 CPUs with Intel 64 support,
15411 prior to the introduction of Pentium 4 G1 step in December 2005,
15412 lacked the @code{LAHF} and @code{SAHF} instructions
15413 which were supported by AMD64.
15414 These are load and store instructions, respectively, for certain status flags.
15415 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
15416 @code{drem}, and @code{remainder} built-in functions;
15417 see @ref{Other Builtins} for details.
15421 This option enables use of the @code{movbe} instruction to implement
15422 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
15426 This option enables built-in functions @code{__builtin_ia32_crc32qi},
15427 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
15428 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
15432 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
15433 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
15434 with an additional Newton-Raphson step
15435 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
15436 (and their vectorized
15437 variants) for single-precision floating-point arguments. These instructions
15438 are generated only when @option{-funsafe-math-optimizations} is enabled
15439 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
15440 Note that while the throughput of the sequence is higher than the throughput
15441 of the non-reciprocal instruction, the precision of the sequence can be
15442 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
15444 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
15445 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
15446 combination), and doesn't need @option{-mrecip}.
15448 Also note that GCC emits the above sequence with additional Newton-Raphson step
15449 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
15450 already with @option{-ffast-math} (or the above option combination), and
15451 doesn't need @option{-mrecip}.
15453 @item -mrecip=@var{opt}
15454 @opindex mrecip=opt
15455 This option controls which reciprocal estimate instructions
15456 may be used. @var{opt} is a comma-separated list of options, which may
15457 be preceded by a @samp{!} to invert the option:
15461 Enable all estimate instructions.
15464 Enable the default instructions, equivalent to @option{-mrecip}.
15467 Disable all estimate instructions, equivalent to @option{-mno-recip}.
15470 Enable the approximation for scalar division.
15473 Enable the approximation for vectorized division.
15476 Enable the approximation for scalar square root.
15479 Enable the approximation for vectorized square root.
15482 So, for example, @option{-mrecip=all,!sqrt} enables
15483 all of the reciprocal approximations, except for square root.
15485 @item -mveclibabi=@var{type}
15486 @opindex mveclibabi
15487 Specifies the ABI type to use for vectorizing intrinsics using an
15488 external library. Supported values for @var{type} are @samp{svml}
15489 for the Intel short
15490 vector math library and @samp{acml} for the AMD math core library.
15491 To use this option, both @option{-ftree-vectorize} and
15492 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
15493 ABI-compatible library must be specified at link time.
15495 GCC currently emits calls to @code{vmldExp2},
15496 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
15497 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
15498 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
15499 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
15500 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
15501 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
15502 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
15503 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
15504 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
15505 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
15506 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
15507 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
15508 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
15509 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
15510 when @option{-mveclibabi=acml} is used.
15512 @item -mabi=@var{name}
15514 Generate code for the specified calling convention. Permissible values
15515 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
15516 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
15517 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
15518 You can control this behavior for a specific function by
15519 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
15520 @xref{Function Attributes}.
15522 @item -mtls-dialect=@var{type}
15523 @opindex mtls-dialect
15524 Generate code to access thread-local storage using the @samp{gnu} or
15525 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
15526 @samp{gnu2} is more efficient, but it may add compile- and run-time
15527 requirements that cannot be satisfied on all systems.
15530 @itemx -mno-push-args
15531 @opindex mpush-args
15532 @opindex mno-push-args
15533 Use PUSH operations to store outgoing parameters. This method is shorter
15534 and usually equally fast as method using SUB/MOV operations and is enabled
15535 by default. In some cases disabling it may improve performance because of
15536 improved scheduling and reduced dependencies.
15538 @item -maccumulate-outgoing-args
15539 @opindex maccumulate-outgoing-args
15540 If enabled, the maximum amount of space required for outgoing arguments is
15541 computed in the function prologue. This is faster on most modern CPUs
15542 because of reduced dependencies, improved scheduling and reduced stack usage
15543 when the preferred stack boundary is not equal to 2. The drawback is a notable
15544 increase in code size. This switch implies @option{-mno-push-args}.
15548 Support thread-safe exception handling on MinGW. Programs that rely
15549 on thread-safe exception handling must compile and link all code with the
15550 @option{-mthreads} option. When compiling, @option{-mthreads} defines
15551 @code{-D_MT}; when linking, it links in a special thread helper library
15552 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
15554 @item -mno-align-stringops
15555 @opindex mno-align-stringops
15556 Do not align the destination of inlined string operations. This switch reduces
15557 code size and improves performance in case the destination is already aligned,
15558 but GCC doesn't know about it.
15560 @item -minline-all-stringops
15561 @opindex minline-all-stringops
15562 By default GCC inlines string operations only when the destination is
15563 known to be aligned to least a 4-byte boundary.
15564 This enables more inlining and increases code
15565 size, but may improve performance of code that depends on fast
15566 @code{memcpy}, @code{strlen},
15567 and @code{memset} for short lengths.
15569 @item -minline-stringops-dynamically
15570 @opindex minline-stringops-dynamically
15571 For string operations of unknown size, use run-time checks with
15572 inline code for small blocks and a library call for large blocks.
15574 @item -mstringop-strategy=@var{alg}
15575 @opindex mstringop-strategy=@var{alg}
15576 Override the internal decision heuristic for the particular algorithm to use
15577 for inlining string operations. The allowed values for @var{alg} are:
15583 Expand using i386 @code{rep} prefix of the specified size.
15587 @itemx unrolled_loop
15588 Expand into an inline loop.
15591 Always use a library call.
15594 @item -mmemcpy-strategy=@var{strategy}
15595 @opindex mmemcpy-strategy=@var{strategy}
15596 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
15597 should be inlined and what inline algorithm to use when the expected size
15598 of the copy operation is known. @var{strategy}
15599 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
15600 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
15601 the max byte size with which inline algorithm @var{alg} is allowed. For the last
15602 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
15603 in the list must be specified in increasing order. The minimal byte size for
15604 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
15607 @item -mmemset-strategy=@var{strategy}
15608 @opindex mmemset-strategy=@var{strategy}
15609 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
15610 @code{__builtin_memset} expansion.
15612 @item -momit-leaf-frame-pointer
15613 @opindex momit-leaf-frame-pointer
15614 Don't keep the frame pointer in a register for leaf functions. This
15615 avoids the instructions to save, set up, and restore frame pointers and
15616 makes an extra register available in leaf functions. The option
15617 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
15618 which might make debugging harder.
15620 @item -mtls-direct-seg-refs
15621 @itemx -mno-tls-direct-seg-refs
15622 @opindex mtls-direct-seg-refs
15623 Controls whether TLS variables may be accessed with offsets from the
15624 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
15625 or whether the thread base pointer must be added. Whether or not this
15626 is valid depends on the operating system, and whether it maps the
15627 segment to cover the entire TLS area.
15629 For systems that use the GNU C Library, the default is on.
15632 @itemx -mno-sse2avx
15634 Specify that the assembler should encode SSE instructions with VEX
15635 prefix. The option @option{-mavx} turns this on by default.
15640 If profiling is active (@option{-pg}), put the profiling
15641 counter call before the prologue.
15642 Note: On x86 architectures the attribute @code{ms_hook_prologue}
15643 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
15646 @itemx -mno-8bit-idiv
15648 On some processors, like Intel Atom, 8-bit unsigned integer divide is
15649 much faster than 32-bit/64-bit integer divide. This option generates a
15650 run-time check. If both dividend and divisor are within range of 0
15651 to 255, 8-bit unsigned integer divide is used instead of
15652 32-bit/64-bit integer divide.
15654 @item -mavx256-split-unaligned-load
15655 @itemx -mavx256-split-unaligned-store
15656 @opindex avx256-split-unaligned-load
15657 @opindex avx256-split-unaligned-store
15658 Split 32-byte AVX unaligned load and store.
15660 @item -mstack-protector-guard=@var{guard}
15661 @opindex mstack-protector-guard=@var{guard}
15662 Generate stack protection code using canary at @var{guard}. Supported
15663 locations are @samp{global} for global canary or @samp{tls} for per-thread
15664 canary in the TLS block (the default). This option has effect only when
15665 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
15669 These @samp{-m} switches are supported in addition to the above
15670 on x86-64 processors in 64-bit environments.
15681 Generate code for a 16-bit, 32-bit or 64-bit environment.
15682 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
15684 generates code that runs on any i386 system.
15686 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
15687 types to 64 bits, and generates code for the x86-64 architecture.
15688 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
15689 and @option{-mdynamic-no-pic} options.
15691 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
15693 generates code for the x86-64 architecture.
15695 The @option{-m16} option is the same as @option{-m32}, except for that
15696 it outputs the @code{.code16gcc} assembly directive at the beginning of
15697 the assembly output so that the binary can run in 16-bit mode.
15699 @item -mno-red-zone
15700 @opindex mno-red-zone
15701 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
15702 by the x86-64 ABI; it is a 128-byte area beyond the location of the
15703 stack pointer that is not modified by signal or interrupt handlers
15704 and therefore can be used for temporary data without adjusting the stack
15705 pointer. The flag @option{-mno-red-zone} disables this red zone.
15707 @item -mcmodel=small
15708 @opindex mcmodel=small
15709 Generate code for the small code model: the program and its symbols must
15710 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
15711 Programs can be statically or dynamically linked. This is the default
15714 @item -mcmodel=kernel
15715 @opindex mcmodel=kernel
15716 Generate code for the kernel code model. The kernel runs in the
15717 negative 2 GB of the address space.
15718 This model has to be used for Linux kernel code.
15720 @item -mcmodel=medium
15721 @opindex mcmodel=medium
15722 Generate code for the medium model: the program is linked in the lower 2
15723 GB of the address space. Small symbols are also placed there. Symbols
15724 with sizes larger than @option{-mlarge-data-threshold} are put into
15725 large data or BSS sections and can be located above 2GB. Programs can
15726 be statically or dynamically linked.
15728 @item -mcmodel=large
15729 @opindex mcmodel=large
15730 Generate code for the large model. This model makes no assumptions
15731 about addresses and sizes of sections.
15733 @item -maddress-mode=long
15734 @opindex maddress-mode=long
15735 Generate code for long address mode. This is only supported for 64-bit
15736 and x32 environments. It is the default address mode for 64-bit
15739 @item -maddress-mode=short
15740 @opindex maddress-mode=short
15741 Generate code for short address mode. This is only supported for 32-bit
15742 and x32 environments. It is the default address mode for 32-bit and
15746 @node i386 and x86-64 Windows Options
15747 @subsection i386 and x86-64 Windows Options
15748 @cindex i386 and x86-64 Windows Options
15750 These additional options are available for Microsoft Windows targets:
15756 specifies that a console application is to be generated, by
15757 instructing the linker to set the PE header subsystem type
15758 required for console applications.
15759 This option is available for Cygwin and MinGW targets and is
15760 enabled by default on those targets.
15764 This option is available for Cygwin and MinGW targets. It
15765 specifies that a DLL---a dynamic link library---is to be
15766 generated, enabling the selection of the required runtime
15767 startup object and entry point.
15769 @item -mnop-fun-dllimport
15770 @opindex mnop-fun-dllimport
15771 This option is available for Cygwin and MinGW targets. It
15772 specifies that the @code{dllimport} attribute should be ignored.
15776 This option is available for MinGW targets. It specifies
15777 that MinGW-specific thread support is to be used.
15781 This option is available for MinGW-w64 targets. It causes
15782 the @code{UNICODE} preprocessor macro to be predefined, and
15783 chooses Unicode-capable runtime startup code.
15787 This option is available for Cygwin and MinGW targets. It
15788 specifies that the typical Microsoft Windows predefined macros are to
15789 be set in the pre-processor, but does not influence the choice
15790 of runtime library/startup code.
15794 This option is available for Cygwin and MinGW targets. It
15795 specifies that a GUI application is to be generated by
15796 instructing the linker to set the PE header subsystem type
15799 @item -fno-set-stack-executable
15800 @opindex fno-set-stack-executable
15801 This option is available for MinGW targets. It specifies that
15802 the executable flag for the stack used by nested functions isn't
15803 set. This is necessary for binaries running in kernel mode of
15804 Microsoft Windows, as there the User32 API, which is used to set executable
15805 privileges, isn't available.
15807 @item -fwritable-relocated-rdata
15808 @opindex fno-writable-relocated-rdata
15809 This option is available for MinGW and Cygwin targets. It specifies
15810 that relocated-data in read-only section is put into .data
15811 section. This is a necessary for older runtimes not supporting
15812 modification of .rdata sections for pseudo-relocation.
15814 @item -mpe-aligned-commons
15815 @opindex mpe-aligned-commons
15816 This option is available for Cygwin and MinGW targets. It
15817 specifies that the GNU extension to the PE file format that
15818 permits the correct alignment of COMMON variables should be
15819 used when generating code. It is enabled by default if
15820 GCC detects that the target assembler found during configuration
15821 supports the feature.
15824 See also under @ref{i386 and x86-64 Options} for standard options.
15826 @node IA-64 Options
15827 @subsection IA-64 Options
15828 @cindex IA-64 Options
15830 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15834 @opindex mbig-endian
15835 Generate code for a big-endian target. This is the default for HP-UX@.
15837 @item -mlittle-endian
15838 @opindex mlittle-endian
15839 Generate code for a little-endian target. This is the default for AIX5
15845 @opindex mno-gnu-as
15846 Generate (or don't) code for the GNU assembler. This is the default.
15847 @c Also, this is the default if the configure option @option{--with-gnu-as}
15853 @opindex mno-gnu-ld
15854 Generate (or don't) code for the GNU linker. This is the default.
15855 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15860 Generate code that does not use a global pointer register. The result
15861 is not position independent code, and violates the IA-64 ABI@.
15863 @item -mvolatile-asm-stop
15864 @itemx -mno-volatile-asm-stop
15865 @opindex mvolatile-asm-stop
15866 @opindex mno-volatile-asm-stop
15867 Generate (or don't) a stop bit immediately before and after volatile asm
15870 @item -mregister-names
15871 @itemx -mno-register-names
15872 @opindex mregister-names
15873 @opindex mno-register-names
15874 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15875 the stacked registers. This may make assembler output more readable.
15881 Disable (or enable) optimizations that use the small data section. This may
15882 be useful for working around optimizer bugs.
15884 @item -mconstant-gp
15885 @opindex mconstant-gp
15886 Generate code that uses a single constant global pointer value. This is
15887 useful when compiling kernel code.
15891 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15892 This is useful when compiling firmware code.
15894 @item -minline-float-divide-min-latency
15895 @opindex minline-float-divide-min-latency
15896 Generate code for inline divides of floating-point values
15897 using the minimum latency algorithm.
15899 @item -minline-float-divide-max-throughput
15900 @opindex minline-float-divide-max-throughput
15901 Generate code for inline divides of floating-point values
15902 using the maximum throughput algorithm.
15904 @item -mno-inline-float-divide
15905 @opindex mno-inline-float-divide
15906 Do not generate inline code for divides of floating-point values.
15908 @item -minline-int-divide-min-latency
15909 @opindex minline-int-divide-min-latency
15910 Generate code for inline divides of integer values
15911 using the minimum latency algorithm.
15913 @item -minline-int-divide-max-throughput
15914 @opindex minline-int-divide-max-throughput
15915 Generate code for inline divides of integer values
15916 using the maximum throughput algorithm.
15918 @item -mno-inline-int-divide
15919 @opindex mno-inline-int-divide
15920 Do not generate inline code for divides of integer values.
15922 @item -minline-sqrt-min-latency
15923 @opindex minline-sqrt-min-latency
15924 Generate code for inline square roots
15925 using the minimum latency algorithm.
15927 @item -minline-sqrt-max-throughput
15928 @opindex minline-sqrt-max-throughput
15929 Generate code for inline square roots
15930 using the maximum throughput algorithm.
15932 @item -mno-inline-sqrt
15933 @opindex mno-inline-sqrt
15934 Do not generate inline code for @code{sqrt}.
15937 @itemx -mno-fused-madd
15938 @opindex mfused-madd
15939 @opindex mno-fused-madd
15940 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
15941 instructions. The default is to use these instructions.
15943 @item -mno-dwarf2-asm
15944 @itemx -mdwarf2-asm
15945 @opindex mno-dwarf2-asm
15946 @opindex mdwarf2-asm
15947 Don't (or do) generate assembler code for the DWARF 2 line number debugging
15948 info. This may be useful when not using the GNU assembler.
15950 @item -mearly-stop-bits
15951 @itemx -mno-early-stop-bits
15952 @opindex mearly-stop-bits
15953 @opindex mno-early-stop-bits
15954 Allow stop bits to be placed earlier than immediately preceding the
15955 instruction that triggered the stop bit. This can improve instruction
15956 scheduling, but does not always do so.
15958 @item -mfixed-range=@var{register-range}
15959 @opindex mfixed-range
15960 Generate code treating the given register range as fixed registers.
15961 A fixed register is one that the register allocator cannot use. This is
15962 useful when compiling kernel code. A register range is specified as
15963 two registers separated by a dash. Multiple register ranges can be
15964 specified separated by a comma.
15966 @item -mtls-size=@var{tls-size}
15968 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
15971 @item -mtune=@var{cpu-type}
15973 Tune the instruction scheduling for a particular CPU, Valid values are
15974 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
15975 and @samp{mckinley}.
15981 Generate code for a 32-bit or 64-bit environment.
15982 The 32-bit environment sets int, long and pointer to 32 bits.
15983 The 64-bit environment sets int to 32 bits and long and pointer
15984 to 64 bits. These are HP-UX specific flags.
15986 @item -mno-sched-br-data-spec
15987 @itemx -msched-br-data-spec
15988 @opindex mno-sched-br-data-spec
15989 @opindex msched-br-data-spec
15990 (Dis/En)able data speculative scheduling before reload.
15991 This results in generation of @code{ld.a} instructions and
15992 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15993 The default is 'disable'.
15995 @item -msched-ar-data-spec
15996 @itemx -mno-sched-ar-data-spec
15997 @opindex msched-ar-data-spec
15998 @opindex mno-sched-ar-data-spec
15999 (En/Dis)able data speculative scheduling after reload.
16000 This results in generation of @code{ld.a} instructions and
16001 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16002 The default is 'enable'.
16004 @item -mno-sched-control-spec
16005 @itemx -msched-control-spec
16006 @opindex mno-sched-control-spec
16007 @opindex msched-control-spec
16008 (Dis/En)able control speculative scheduling. This feature is
16009 available only during region scheduling (i.e.@: before reload).
16010 This results in generation of the @code{ld.s} instructions and
16011 the corresponding check instructions @code{chk.s}.
16012 The default is 'disable'.
16014 @item -msched-br-in-data-spec
16015 @itemx -mno-sched-br-in-data-spec
16016 @opindex msched-br-in-data-spec
16017 @opindex mno-sched-br-in-data-spec
16018 (En/Dis)able speculative scheduling of the instructions that
16019 are dependent on the data speculative loads before reload.
16020 This is effective only with @option{-msched-br-data-spec} enabled.
16021 The default is 'enable'.
16023 @item -msched-ar-in-data-spec
16024 @itemx -mno-sched-ar-in-data-spec
16025 @opindex msched-ar-in-data-spec
16026 @opindex mno-sched-ar-in-data-spec
16027 (En/Dis)able speculative scheduling of the instructions that
16028 are dependent on the data speculative loads after reload.
16029 This is effective only with @option{-msched-ar-data-spec} enabled.
16030 The default is 'enable'.
16032 @item -msched-in-control-spec
16033 @itemx -mno-sched-in-control-spec
16034 @opindex msched-in-control-spec
16035 @opindex mno-sched-in-control-spec
16036 (En/Dis)able speculative scheduling of the instructions that
16037 are dependent on the control speculative loads.
16038 This is effective only with @option{-msched-control-spec} enabled.
16039 The default is 'enable'.
16041 @item -mno-sched-prefer-non-data-spec-insns
16042 @itemx -msched-prefer-non-data-spec-insns
16043 @opindex mno-sched-prefer-non-data-spec-insns
16044 @opindex msched-prefer-non-data-spec-insns
16045 If enabled, data-speculative instructions are chosen for schedule
16046 only if there are no other choices at the moment. This makes
16047 the use of the data speculation much more conservative.
16048 The default is 'disable'.
16050 @item -mno-sched-prefer-non-control-spec-insns
16051 @itemx -msched-prefer-non-control-spec-insns
16052 @opindex mno-sched-prefer-non-control-spec-insns
16053 @opindex msched-prefer-non-control-spec-insns
16054 If enabled, control-speculative instructions are chosen for schedule
16055 only if there are no other choices at the moment. This makes
16056 the use of the control speculation much more conservative.
16057 The default is 'disable'.
16059 @item -mno-sched-count-spec-in-critical-path
16060 @itemx -msched-count-spec-in-critical-path
16061 @opindex mno-sched-count-spec-in-critical-path
16062 @opindex msched-count-spec-in-critical-path
16063 If enabled, speculative dependencies are considered during
16064 computation of the instructions priorities. This makes the use of the
16065 speculation a bit more conservative.
16066 The default is 'disable'.
16068 @item -msched-spec-ldc
16069 @opindex msched-spec-ldc
16070 Use a simple data speculation check. This option is on by default.
16072 @item -msched-control-spec-ldc
16073 @opindex msched-spec-ldc
16074 Use a simple check for control speculation. This option is on by default.
16076 @item -msched-stop-bits-after-every-cycle
16077 @opindex msched-stop-bits-after-every-cycle
16078 Place a stop bit after every cycle when scheduling. This option is on
16081 @item -msched-fp-mem-deps-zero-cost
16082 @opindex msched-fp-mem-deps-zero-cost
16083 Assume that floating-point stores and loads are not likely to cause a conflict
16084 when placed into the same instruction group. This option is disabled by
16087 @item -msel-sched-dont-check-control-spec
16088 @opindex msel-sched-dont-check-control-spec
16089 Generate checks for control speculation in selective scheduling.
16090 This flag is disabled by default.
16092 @item -msched-max-memory-insns=@var{max-insns}
16093 @opindex msched-max-memory-insns
16094 Limit on the number of memory insns per instruction group, giving lower
16095 priority to subsequent memory insns attempting to schedule in the same
16096 instruction group. Frequently useful to prevent cache bank conflicts.
16097 The default value is 1.
16099 @item -msched-max-memory-insns-hard-limit
16100 @opindex msched-max-memory-insns-hard-limit
16101 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16102 disallowing more than that number in an instruction group.
16103 Otherwise, the limit is ``soft'', meaning that non-memory operations
16104 are preferred when the limit is reached, but memory operations may still
16110 @subsection LM32 Options
16111 @cindex LM32 options
16113 These @option{-m} options are defined for the LatticeMico32 architecture:
16116 @item -mbarrel-shift-enabled
16117 @opindex mbarrel-shift-enabled
16118 Enable barrel-shift instructions.
16120 @item -mdivide-enabled
16121 @opindex mdivide-enabled
16122 Enable divide and modulus instructions.
16124 @item -mmultiply-enabled
16125 @opindex multiply-enabled
16126 Enable multiply instructions.
16128 @item -msign-extend-enabled
16129 @opindex msign-extend-enabled
16130 Enable sign extend instructions.
16132 @item -muser-enabled
16133 @opindex muser-enabled
16134 Enable user-defined instructions.
16139 @subsection M32C Options
16140 @cindex M32C options
16143 @item -mcpu=@var{name}
16145 Select the CPU for which code is generated. @var{name} may be one of
16146 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16147 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16148 the M32C/80 series.
16152 Specifies that the program will be run on the simulator. This causes
16153 an alternate runtime library to be linked in which supports, for
16154 example, file I/O@. You must not use this option when generating
16155 programs that will run on real hardware; you must provide your own
16156 runtime library for whatever I/O functions are needed.
16158 @item -memregs=@var{number}
16160 Specifies the number of memory-based pseudo-registers GCC uses
16161 during code generation. These pseudo-registers are used like real
16162 registers, so there is a tradeoff between GCC's ability to fit the
16163 code into available registers, and the performance penalty of using
16164 memory instead of registers. Note that all modules in a program must
16165 be compiled with the same value for this option. Because of that, you
16166 must not use this option with GCC's default runtime libraries.
16170 @node M32R/D Options
16171 @subsection M32R/D Options
16172 @cindex M32R/D options
16174 These @option{-m} options are defined for Renesas M32R/D architectures:
16179 Generate code for the M32R/2@.
16183 Generate code for the M32R/X@.
16187 Generate code for the M32R@. This is the default.
16189 @item -mmodel=small
16190 @opindex mmodel=small
16191 Assume all objects live in the lower 16MB of memory (so that their addresses
16192 can be loaded with the @code{ld24} instruction), and assume all subroutines
16193 are reachable with the @code{bl} instruction.
16194 This is the default.
16196 The addressability of a particular object can be set with the
16197 @code{model} attribute.
16199 @item -mmodel=medium
16200 @opindex mmodel=medium
16201 Assume objects may be anywhere in the 32-bit address space (the compiler
16202 generates @code{seth/add3} instructions to load their addresses), and
16203 assume all subroutines are reachable with the @code{bl} instruction.
16205 @item -mmodel=large
16206 @opindex mmodel=large
16207 Assume objects may be anywhere in the 32-bit address space (the compiler
16208 generates @code{seth/add3} instructions to load their addresses), and
16209 assume subroutines may not be reachable with the @code{bl} instruction
16210 (the compiler generates the much slower @code{seth/add3/jl}
16211 instruction sequence).
16214 @opindex msdata=none
16215 Disable use of the small data area. Variables are put into
16216 one of @samp{.data}, @samp{.bss}, or @samp{.rodata} (unless the
16217 @code{section} attribute has been specified).
16218 This is the default.
16220 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
16221 Objects may be explicitly put in the small data area with the
16222 @code{section} attribute using one of these sections.
16224 @item -msdata=sdata
16225 @opindex msdata=sdata
16226 Put small global and static data in the small data area, but do not
16227 generate special code to reference them.
16230 @opindex msdata=use
16231 Put small global and static data in the small data area, and generate
16232 special instructions to reference them.
16236 @cindex smaller data references
16237 Put global and static objects less than or equal to @var{num} bytes
16238 into the small data or BSS sections instead of the normal data or BSS
16239 sections. The default value of @var{num} is 8.
16240 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16241 for this option to have any effect.
16243 All modules should be compiled with the same @option{-G @var{num}} value.
16244 Compiling with different values of @var{num} may or may not work; if it
16245 doesn't the linker gives an error message---incorrect code is not
16250 Makes the M32R-specific code in the compiler display some statistics
16251 that might help in debugging programs.
16253 @item -malign-loops
16254 @opindex malign-loops
16255 Align all loops to a 32-byte boundary.
16257 @item -mno-align-loops
16258 @opindex mno-align-loops
16259 Do not enforce a 32-byte alignment for loops. This is the default.
16261 @item -missue-rate=@var{number}
16262 @opindex missue-rate=@var{number}
16263 Issue @var{number} instructions per cycle. @var{number} can only be 1
16266 @item -mbranch-cost=@var{number}
16267 @opindex mbranch-cost=@var{number}
16268 @var{number} can only be 1 or 2. If it is 1 then branches are
16269 preferred over conditional code, if it is 2, then the opposite applies.
16271 @item -mflush-trap=@var{number}
16272 @opindex mflush-trap=@var{number}
16273 Specifies the trap number to use to flush the cache. The default is
16274 12. Valid numbers are between 0 and 15 inclusive.
16276 @item -mno-flush-trap
16277 @opindex mno-flush-trap
16278 Specifies that the cache cannot be flushed by using a trap.
16280 @item -mflush-func=@var{name}
16281 @opindex mflush-func=@var{name}
16282 Specifies the name of the operating system function to call to flush
16283 the cache. The default is @emph{_flush_cache}, but a function call
16284 is only used if a trap is not available.
16286 @item -mno-flush-func
16287 @opindex mno-flush-func
16288 Indicates that there is no OS function for flushing the cache.
16292 @node M680x0 Options
16293 @subsection M680x0 Options
16294 @cindex M680x0 options
16296 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16297 The default settings depend on which architecture was selected when
16298 the compiler was configured; the defaults for the most common choices
16302 @item -march=@var{arch}
16304 Generate code for a specific M680x0 or ColdFire instruction set
16305 architecture. Permissible values of @var{arch} for M680x0
16306 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16307 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16308 architectures are selected according to Freescale's ISA classification
16309 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16310 @samp{isab} and @samp{isac}.
16312 GCC defines a macro @samp{__mcf@var{arch}__} whenever it is generating
16313 code for a ColdFire target. The @var{arch} in this macro is one of the
16314 @option{-march} arguments given above.
16316 When used together, @option{-march} and @option{-mtune} select code
16317 that runs on a family of similar processors but that is optimized
16318 for a particular microarchitecture.
16320 @item -mcpu=@var{cpu}
16322 Generate code for a specific M680x0 or ColdFire processor.
16323 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16324 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16325 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16326 below, which also classifies the CPUs into families:
16328 @multitable @columnfractions 0.20 0.80
16329 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16330 @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}
16331 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16332 @item @samp{5206e} @tab @samp{5206e}
16333 @item @samp{5208} @tab @samp{5207} @samp{5208}
16334 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16335 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16336 @item @samp{5216} @tab @samp{5214} @samp{5216}
16337 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16338 @item @samp{5225} @tab @samp{5224} @samp{5225}
16339 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16340 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16341 @item @samp{5249} @tab @samp{5249}
16342 @item @samp{5250} @tab @samp{5250}
16343 @item @samp{5271} @tab @samp{5270} @samp{5271}
16344 @item @samp{5272} @tab @samp{5272}
16345 @item @samp{5275} @tab @samp{5274} @samp{5275}
16346 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16347 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16348 @item @samp{5307} @tab @samp{5307}
16349 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16350 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16351 @item @samp{5407} @tab @samp{5407}
16352 @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}
16355 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16356 @var{arch} is compatible with @var{cpu}. Other combinations of
16357 @option{-mcpu} and @option{-march} are rejected.
16359 GCC defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
16360 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
16361 where the value of @var{family} is given by the table above.
16363 @item -mtune=@var{tune}
16365 Tune the code for a particular microarchitecture within the
16366 constraints set by @option{-march} and @option{-mcpu}.
16367 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16368 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16369 and @samp{cpu32}. The ColdFire microarchitectures
16370 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16372 You can also use @option{-mtune=68020-40} for code that needs
16373 to run relatively well on 68020, 68030 and 68040 targets.
16374 @option{-mtune=68020-60} is similar but includes 68060 targets
16375 as well. These two options select the same tuning decisions as
16376 @option{-m68020-40} and @option{-m68020-60} respectively.
16378 GCC defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
16379 when tuning for 680x0 architecture @var{arch}. It also defines
16380 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16381 option is used. If GCC is tuning for a range of architectures,
16382 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16383 it defines the macros for every architecture in the range.
16385 GCC also defines the macro @samp{__m@var{uarch}__} when tuning for
16386 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16387 of the arguments given above.
16393 Generate output for a 68000. This is the default
16394 when the compiler is configured for 68000-based systems.
16395 It is equivalent to @option{-march=68000}.
16397 Use this option for microcontrollers with a 68000 or EC000 core,
16398 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16402 Generate output for a 68010. This is the default
16403 when the compiler is configured for 68010-based systems.
16404 It is equivalent to @option{-march=68010}.
16410 Generate output for a 68020. This is the default
16411 when the compiler is configured for 68020-based systems.
16412 It is equivalent to @option{-march=68020}.
16416 Generate output for a 68030. This is the default when the compiler is
16417 configured for 68030-based systems. It is equivalent to
16418 @option{-march=68030}.
16422 Generate output for a 68040. This is the default when the compiler is
16423 configured for 68040-based systems. It is equivalent to
16424 @option{-march=68040}.
16426 This option inhibits the use of 68881/68882 instructions that have to be
16427 emulated by software on the 68040. Use this option if your 68040 does not
16428 have code to emulate those instructions.
16432 Generate output for a 68060. This is the default when the compiler is
16433 configured for 68060-based systems. It is equivalent to
16434 @option{-march=68060}.
16436 This option inhibits the use of 68020 and 68881/68882 instructions that
16437 have to be emulated by software on the 68060. Use this option if your 68060
16438 does not have code to emulate those instructions.
16442 Generate output for a CPU32. This is the default
16443 when the compiler is configured for CPU32-based systems.
16444 It is equivalent to @option{-march=cpu32}.
16446 Use this option for microcontrollers with a
16447 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16448 68336, 68340, 68341, 68349 and 68360.
16452 Generate output for a 520X ColdFire CPU@. This is the default
16453 when the compiler is configured for 520X-based systems.
16454 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16455 in favor of that option.
16457 Use this option for microcontroller with a 5200 core, including
16458 the MCF5202, MCF5203, MCF5204 and MCF5206.
16462 Generate output for a 5206e ColdFire CPU@. The option is now
16463 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16467 Generate output for a member of the ColdFire 528X family.
16468 The option is now deprecated in favor of the equivalent
16469 @option{-mcpu=528x}.
16473 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16474 in favor of the equivalent @option{-mcpu=5307}.
16478 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16479 in favor of the equivalent @option{-mcpu=5407}.
16483 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16484 This includes use of hardware floating-point instructions.
16485 The option is equivalent to @option{-mcpu=547x}, and is now
16486 deprecated in favor of that option.
16490 Generate output for a 68040, without using any of the new instructions.
16491 This results in code that can run relatively efficiently on either a
16492 68020/68881 or a 68030 or a 68040. The generated code does use the
16493 68881 instructions that are emulated on the 68040.
16495 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16499 Generate output for a 68060, without using any of the new instructions.
16500 This results in code that can run relatively efficiently on either a
16501 68020/68881 or a 68030 or a 68040. The generated code does use the
16502 68881 instructions that are emulated on the 68060.
16504 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16508 @opindex mhard-float
16510 Generate floating-point instructions. This is the default for 68020
16511 and above, and for ColdFire devices that have an FPU@. It defines the
16512 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
16513 on ColdFire targets.
16516 @opindex msoft-float
16517 Do not generate floating-point instructions; use library calls instead.
16518 This is the default for 68000, 68010, and 68832 targets. It is also
16519 the default for ColdFire devices that have no FPU.
16525 Generate (do not generate) ColdFire hardware divide and remainder
16526 instructions. If @option{-march} is used without @option{-mcpu},
16527 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16528 architectures. Otherwise, the default is taken from the target CPU
16529 (either the default CPU, or the one specified by @option{-mcpu}). For
16530 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16531 @option{-mcpu=5206e}.
16533 GCC defines the macro @samp{__mcfhwdiv__} when this option is enabled.
16537 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16538 Additionally, parameters passed on the stack are also aligned to a
16539 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16543 Do not consider type @code{int} to be 16 bits wide. This is the default.
16546 @itemx -mno-bitfield
16547 @opindex mnobitfield
16548 @opindex mno-bitfield
16549 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16550 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16554 Do use the bit-field instructions. The @option{-m68020} option implies
16555 @option{-mbitfield}. This is the default if you use a configuration
16556 designed for a 68020.
16560 Use a different function-calling convention, in which functions
16561 that take a fixed number of arguments return with the @code{rtd}
16562 instruction, which pops their arguments while returning. This
16563 saves one instruction in the caller since there is no need to pop
16564 the arguments there.
16566 This calling convention is incompatible with the one normally
16567 used on Unix, so you cannot use it if you need to call libraries
16568 compiled with the Unix compiler.
16570 Also, you must provide function prototypes for all functions that
16571 take variable numbers of arguments (including @code{printf});
16572 otherwise incorrect code is generated for calls to those
16575 In addition, seriously incorrect code results if you call a
16576 function with too many arguments. (Normally, extra arguments are
16577 harmlessly ignored.)
16579 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16580 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16584 Do not use the calling conventions selected by @option{-mrtd}.
16585 This is the default.
16588 @itemx -mno-align-int
16589 @opindex malign-int
16590 @opindex mno-align-int
16591 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16592 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16593 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16594 Aligning variables on 32-bit boundaries produces code that runs somewhat
16595 faster on processors with 32-bit busses at the expense of more memory.
16597 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16598 aligns structures containing the above types differently than
16599 most published application binary interface specifications for the m68k.
16603 Use the pc-relative addressing mode of the 68000 directly, instead of
16604 using a global offset table. At present, this option implies @option{-fpic},
16605 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16606 not presently supported with @option{-mpcrel}, though this could be supported for
16607 68020 and higher processors.
16609 @item -mno-strict-align
16610 @itemx -mstrict-align
16611 @opindex mno-strict-align
16612 @opindex mstrict-align
16613 Do not (do) assume that unaligned memory references are handled by
16617 Generate code that allows the data segment to be located in a different
16618 area of memory from the text segment. This allows for execute-in-place in
16619 an environment without virtual memory management. This option implies
16622 @item -mno-sep-data
16623 Generate code that assumes that the data segment follows the text segment.
16624 This is the default.
16626 @item -mid-shared-library
16627 Generate code that supports shared libraries via the library ID method.
16628 This allows for execute-in-place and shared libraries in an environment
16629 without virtual memory management. This option implies @option{-fPIC}.
16631 @item -mno-id-shared-library
16632 Generate code that doesn't assume ID-based shared libraries are being used.
16633 This is the default.
16635 @item -mshared-library-id=n
16636 Specifies the identification number of the ID-based shared library being
16637 compiled. Specifying a value of 0 generates more compact code; specifying
16638 other values forces the allocation of that number to the current
16639 library, but is no more space- or time-efficient than omitting this option.
16645 When generating position-independent code for ColdFire, generate code
16646 that works if the GOT has more than 8192 entries. This code is
16647 larger and slower than code generated without this option. On M680x0
16648 processors, this option is not needed; @option{-fPIC} suffices.
16650 GCC normally uses a single instruction to load values from the GOT@.
16651 While this is relatively efficient, it only works if the GOT
16652 is smaller than about 64k. Anything larger causes the linker
16653 to report an error such as:
16655 @cindex relocation truncated to fit (ColdFire)
16657 relocation truncated to fit: R_68K_GOT16O foobar
16660 If this happens, you should recompile your code with @option{-mxgot}.
16661 It should then work with very large GOTs. However, code generated with
16662 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16663 the value of a global symbol.
16665 Note that some linkers, including newer versions of the GNU linker,
16666 can create multiple GOTs and sort GOT entries. If you have such a linker,
16667 you should only need to use @option{-mxgot} when compiling a single
16668 object file that accesses more than 8192 GOT entries. Very few do.
16670 These options have no effect unless GCC is generating
16671 position-independent code.
16675 @node MCore Options
16676 @subsection MCore Options
16677 @cindex MCore options
16679 These are the @samp{-m} options defined for the Motorola M*Core
16685 @itemx -mno-hardlit
16687 @opindex mno-hardlit
16688 Inline constants into the code stream if it can be done in two
16689 instructions or less.
16695 Use the divide instruction. (Enabled by default).
16697 @item -mrelax-immediate
16698 @itemx -mno-relax-immediate
16699 @opindex mrelax-immediate
16700 @opindex mno-relax-immediate
16701 Allow arbitrary-sized immediates in bit operations.
16703 @item -mwide-bitfields
16704 @itemx -mno-wide-bitfields
16705 @opindex mwide-bitfields
16706 @opindex mno-wide-bitfields
16707 Always treat bit-fields as @code{int}-sized.
16709 @item -m4byte-functions
16710 @itemx -mno-4byte-functions
16711 @opindex m4byte-functions
16712 @opindex mno-4byte-functions
16713 Force all functions to be aligned to a 4-byte boundary.
16715 @item -mcallgraph-data
16716 @itemx -mno-callgraph-data
16717 @opindex mcallgraph-data
16718 @opindex mno-callgraph-data
16719 Emit callgraph information.
16722 @itemx -mno-slow-bytes
16723 @opindex mslow-bytes
16724 @opindex mno-slow-bytes
16725 Prefer word access when reading byte quantities.
16727 @item -mlittle-endian
16728 @itemx -mbig-endian
16729 @opindex mlittle-endian
16730 @opindex mbig-endian
16731 Generate code for a little-endian target.
16737 Generate code for the 210 processor.
16741 Assume that runtime support has been provided and so omit the
16742 simulator library (@file{libsim.a)} from the linker command line.
16744 @item -mstack-increment=@var{size}
16745 @opindex mstack-increment
16746 Set the maximum amount for a single stack increment operation. Large
16747 values can increase the speed of programs that contain functions
16748 that need a large amount of stack space, but they can also trigger a
16749 segmentation fault if the stack is extended too much. The default
16755 @subsection MeP Options
16756 @cindex MeP options
16762 Enables the @code{abs} instruction, which is the absolute difference
16763 between two registers.
16767 Enables all the optional instructions---average, multiply, divide, bit
16768 operations, leading zero, absolute difference, min/max, clip, and
16774 Enables the @code{ave} instruction, which computes the average of two
16777 @item -mbased=@var{n}
16779 Variables of size @var{n} bytes or smaller are placed in the
16780 @code{.based} section by default. Based variables use the @code{$tp}
16781 register as a base register, and there is a 128-byte limit to the
16782 @code{.based} section.
16786 Enables the bit operation instructions---bit test (@code{btstm}), set
16787 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16788 test-and-set (@code{tas}).
16790 @item -mc=@var{name}
16792 Selects which section constant data is placed in. @var{name} may
16793 be @code{tiny}, @code{near}, or @code{far}.
16797 Enables the @code{clip} instruction. Note that @code{-mclip} is not
16798 useful unless you also provide @code{-mminmax}.
16800 @item -mconfig=@var{name}
16802 Selects one of the built-in core configurations. Each MeP chip has
16803 one or more modules in it; each module has a core CPU and a variety of
16804 coprocessors, optional instructions, and peripherals. The
16805 @code{MeP-Integrator} tool, not part of GCC, provides these
16806 configurations through this option; using this option is the same as
16807 using all the corresponding command-line options. The default
16808 configuration is @code{default}.
16812 Enables the coprocessor instructions. By default, this is a 32-bit
16813 coprocessor. Note that the coprocessor is normally enabled via the
16814 @code{-mconfig=} option.
16818 Enables the 32-bit coprocessor's instructions.
16822 Enables the 64-bit coprocessor's instructions.
16826 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16830 Causes constant variables to be placed in the @code{.near} section.
16834 Enables the @code{div} and @code{divu} instructions.
16838 Generate big-endian code.
16842 Generate little-endian code.
16844 @item -mio-volatile
16845 @opindex mio-volatile
16846 Tells the compiler that any variable marked with the @code{io}
16847 attribute is to be considered volatile.
16851 Causes variables to be assigned to the @code{.far} section by default.
16855 Enables the @code{leadz} (leading zero) instruction.
16859 Causes variables to be assigned to the @code{.near} section by default.
16863 Enables the @code{min} and @code{max} instructions.
16867 Enables the multiplication and multiply-accumulate instructions.
16871 Disables all the optional instructions enabled by @code{-mall-opts}.
16875 Enables the @code{repeat} and @code{erepeat} instructions, used for
16876 low-overhead looping.
16880 Causes all variables to default to the @code{.tiny} section. Note
16881 that there is a 65536-byte limit to this section. Accesses to these
16882 variables use the @code{%gp} base register.
16886 Enables the saturation instructions. Note that the compiler does not
16887 currently generate these itself, but this option is included for
16888 compatibility with other tools, like @code{as}.
16892 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16896 Link the simulator run-time libraries.
16900 Link the simulator runtime libraries, excluding built-in support
16901 for reset and exception vectors and tables.
16905 Causes all functions to default to the @code{.far} section. Without
16906 this option, functions default to the @code{.near} section.
16908 @item -mtiny=@var{n}
16910 Variables that are @var{n} bytes or smaller are allocated to the
16911 @code{.tiny} section. These variables use the @code{$gp} base
16912 register. The default for this option is 4, but note that there's a
16913 65536-byte limit to the @code{.tiny} section.
16917 @node MicroBlaze Options
16918 @subsection MicroBlaze Options
16919 @cindex MicroBlaze Options
16924 @opindex msoft-float
16925 Use software emulation for floating point (default).
16928 @opindex mhard-float
16929 Use hardware floating-point instructions.
16933 Do not optimize block moves, use @code{memcpy}.
16935 @item -mno-clearbss
16936 @opindex mno-clearbss
16937 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
16939 @item -mcpu=@var{cpu-type}
16941 Use features of, and schedule code for, the given CPU.
16942 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
16943 where @var{X} is a major version, @var{YY} is the minor version, and
16944 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
16945 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
16947 @item -mxl-soft-mul
16948 @opindex mxl-soft-mul
16949 Use software multiply emulation (default).
16951 @item -mxl-soft-div
16952 @opindex mxl-soft-div
16953 Use software emulation for divides (default).
16955 @item -mxl-barrel-shift
16956 @opindex mxl-barrel-shift
16957 Use the hardware barrel shifter.
16959 @item -mxl-pattern-compare
16960 @opindex mxl-pattern-compare
16961 Use pattern compare instructions.
16963 @item -msmall-divides
16964 @opindex msmall-divides
16965 Use table lookup optimization for small signed integer divisions.
16967 @item -mxl-stack-check
16968 @opindex mxl-stack-check
16969 This option is deprecated. Use @option{-fstack-check} instead.
16972 @opindex mxl-gp-opt
16973 Use GP-relative @code{.sdata}/@code{.sbss} sections.
16975 @item -mxl-multiply-high
16976 @opindex mxl-multiply-high
16977 Use multiply high instructions for high part of 32x32 multiply.
16979 @item -mxl-float-convert
16980 @opindex mxl-float-convert
16981 Use hardware floating-point conversion instructions.
16983 @item -mxl-float-sqrt
16984 @opindex mxl-float-sqrt
16985 Use hardware floating-point square root instruction.
16988 @opindex mbig-endian
16989 Generate code for a big-endian target.
16991 @item -mlittle-endian
16992 @opindex mlittle-endian
16993 Generate code for a little-endian target.
16996 @opindex mxl-reorder
16997 Use reorder instructions (swap and byte reversed load/store).
16999 @item -mxl-mode-@var{app-model}
17000 Select application model @var{app-model}. Valid models are
17003 normal executable (default), uses startup code @file{crt0.o}.
17006 for use with Xilinx Microprocessor Debugger (XMD) based
17007 software intrusive debug agent called xmdstub. This uses startup file
17008 @file{crt1.o} and sets the start address of the program to 0x800.
17011 for applications that are loaded using a bootloader.
17012 This model uses startup file @file{crt2.o} which does not contain a processor
17013 reset vector handler. This is suitable for transferring control on a
17014 processor reset to the bootloader rather than the application.
17017 for applications that do not require any of the
17018 MicroBlaze vectors. This option may be useful for applications running
17019 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17022 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17023 @option{-mxl-mode-@var{app-model}}.
17028 @subsection MIPS Options
17029 @cindex MIPS options
17035 Generate big-endian code.
17039 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17042 @item -march=@var{arch}
17044 Generate code that runs on @var{arch}, which can be the name of a
17045 generic MIPS ISA, or the name of a particular processor.
17047 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17048 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
17049 The processor names are:
17050 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17051 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17052 @samp{5kc}, @samp{5kf},
17054 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17055 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17056 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17057 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17058 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17059 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17061 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17062 @samp{octeon}, @samp{octeon+}, @samp{octeon2},
17064 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17065 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17066 @samp{rm7000}, @samp{rm9000},
17067 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17070 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17071 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17072 @samp{xlr} and @samp{xlp}.
17073 The special value @samp{from-abi} selects the
17074 most compatible architecture for the selected ABI (that is,
17075 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17077 The native Linux/GNU toolchain also supports the value @samp{native},
17078 which selects the best architecture option for the host processor.
17079 @option{-march=native} has no effect if GCC does not recognize
17082 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17083 (for example, @option{-march=r2k}). Prefixes are optional, and
17084 @samp{vr} may be written @samp{r}.
17086 Names of the form @samp{@var{n}f2_1} refer to processors with
17087 FPUs clocked at half the rate of the core, names of the form
17088 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17089 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17090 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17091 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17092 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17093 accepted as synonyms for @samp{@var{n}f1_1}.
17095 GCC defines two macros based on the value of this option. The first
17096 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
17097 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
17098 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
17099 For example, @option{-march=r2000} sets @samp{_MIPS_ARCH}
17100 to @samp{"r2000"} and defines the macro @samp{_MIPS_ARCH_R2000}.
17102 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
17103 above. In other words, it has the full prefix and does not
17104 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17105 the macro names the resolved architecture (either @samp{"mips1"} or
17106 @samp{"mips3"}). It names the default architecture when no
17107 @option{-march} option is given.
17109 @item -mtune=@var{arch}
17111 Optimize for @var{arch}. Among other things, this option controls
17112 the way instructions are scheduled, and the perceived cost of arithmetic
17113 operations. The list of @var{arch} values is the same as for
17116 When this option is not used, GCC optimizes for the processor
17117 specified by @option{-march}. By using @option{-march} and
17118 @option{-mtune} together, it is possible to generate code that
17119 runs on a family of processors, but optimize the code for one
17120 particular member of that family.
17122 @option{-mtune} defines the macros @samp{_MIPS_TUNE} and
17123 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17124 @option{-march} ones described above.
17128 Equivalent to @option{-march=mips1}.
17132 Equivalent to @option{-march=mips2}.
17136 Equivalent to @option{-march=mips3}.
17140 Equivalent to @option{-march=mips4}.
17144 Equivalent to @option{-march=mips32}.
17148 Equivalent to @option{-march=mips32r2}.
17152 Equivalent to @option{-march=mips64}.
17156 Equivalent to @option{-march=mips64r2}.
17161 @opindex mno-mips16
17162 Generate (do not generate) MIPS16 code. If GCC is targeting a
17163 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17165 MIPS16 code generation can also be controlled on a per-function basis
17166 by means of @code{mips16} and @code{nomips16} attributes.
17167 @xref{Function Attributes}, for more information.
17169 @item -mflip-mips16
17170 @opindex mflip-mips16
17171 Generate MIPS16 code on alternating functions. This option is provided
17172 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17173 not intended for ordinary use in compiling user code.
17175 @item -minterlink-compressed
17176 @item -mno-interlink-compressed
17177 @opindex minterlink-compressed
17178 @opindex mno-interlink-compressed
17179 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17180 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17182 For example, code using the standard ISA encoding cannot jump directly
17183 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17184 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17185 knows that the target of the jump is not compressed.
17187 @item -minterlink-mips16
17188 @itemx -mno-interlink-mips16
17189 @opindex minterlink-mips16
17190 @opindex mno-interlink-mips16
17191 Aliases of @option{-minterlink-compressed} and
17192 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17193 and are retained for backwards compatibility.
17205 Generate code for the given ABI@.
17207 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17208 generates 64-bit code when you select a 64-bit architecture, but you
17209 can use @option{-mgp32} to get 32-bit code instead.
17211 For information about the O64 ABI, see
17212 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17214 GCC supports a variant of the o32 ABI in which floating-point registers
17215 are 64 rather than 32 bits wide. You can select this combination with
17216 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17217 and @code{mfhc1} instructions and is therefore only supported for
17218 MIPS32R2 processors.
17220 The register assignments for arguments and return values remain the
17221 same, but each scalar value is passed in a single 64-bit register
17222 rather than a pair of 32-bit registers. For example, scalar
17223 floating-point values are returned in @samp{$f0} only, not a
17224 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17225 remains the same, but all 64 bits are saved.
17228 @itemx -mno-abicalls
17230 @opindex mno-abicalls
17231 Generate (do not generate) code that is suitable for SVR4-style
17232 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17237 Generate (do not generate) code that is fully position-independent,
17238 and that can therefore be linked into shared libraries. This option
17239 only affects @option{-mabicalls}.
17241 All @option{-mabicalls} code has traditionally been position-independent,
17242 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17243 as an extension, the GNU toolchain allows executables to use absolute
17244 accesses for locally-binding symbols. It can also use shorter GP
17245 initialization sequences and generate direct calls to locally-defined
17246 functions. This mode is selected by @option{-mno-shared}.
17248 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17249 objects that can only be linked by the GNU linker. However, the option
17250 does not affect the ABI of the final executable; it only affects the ABI
17251 of relocatable objects. Using @option{-mno-shared} generally makes
17252 executables both smaller and quicker.
17254 @option{-mshared} is the default.
17260 Assume (do not assume) that the static and dynamic linkers
17261 support PLTs and copy relocations. This option only affects
17262 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17263 has no effect without @option{-msym32}.
17265 You can make @option{-mplt} the default by configuring
17266 GCC with @option{--with-mips-plt}. The default is
17267 @option{-mno-plt} otherwise.
17273 Lift (do not lift) the usual restrictions on the size of the global
17276 GCC normally uses a single instruction to load values from the GOT@.
17277 While this is relatively efficient, it only works if the GOT
17278 is smaller than about 64k. Anything larger causes the linker
17279 to report an error such as:
17281 @cindex relocation truncated to fit (MIPS)
17283 relocation truncated to fit: R_MIPS_GOT16 foobar
17286 If this happens, you should recompile your code with @option{-mxgot}.
17287 This works with very large GOTs, although the code is also
17288 less efficient, since it takes three instructions to fetch the
17289 value of a global symbol.
17291 Note that some linkers can create multiple GOTs. If you have such a
17292 linker, you should only need to use @option{-mxgot} when a single object
17293 file accesses more than 64k's worth of GOT entries. Very few do.
17295 These options have no effect unless GCC is generating position
17300 Assume that general-purpose registers are 32 bits wide.
17304 Assume that general-purpose registers are 64 bits wide.
17308 Assume that floating-point registers are 32 bits wide.
17312 Assume that floating-point registers are 64 bits wide.
17315 @opindex mhard-float
17316 Use floating-point coprocessor instructions.
17319 @opindex msoft-float
17320 Do not use floating-point coprocessor instructions. Implement
17321 floating-point calculations using library calls instead.
17325 Equivalent to @option{-msoft-float}, but additionally asserts that the
17326 program being compiled does not perform any floating-point operations.
17327 This option is presently supported only by some bare-metal MIPS
17328 configurations, where it may select a special set of libraries
17329 that lack all floating-point support (including, for example, the
17330 floating-point @code{printf} formats).
17331 If code compiled with @code{-mno-float} accidentally contains
17332 floating-point operations, it is likely to suffer a link-time
17333 or run-time failure.
17335 @item -msingle-float
17336 @opindex msingle-float
17337 Assume that the floating-point coprocessor only supports single-precision
17340 @item -mdouble-float
17341 @opindex mdouble-float
17342 Assume that the floating-point coprocessor supports double-precision
17343 operations. This is the default.
17346 @itemx -mabs=legacy
17348 @opindex mabs=legacy
17349 These options control the treatment of the special not-a-number (NaN)
17350 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17351 @code{neg.@i{fmt}} machine instructions.
17353 By default or when the @option{-mabs=legacy} is used the legacy
17354 treatment is selected. In this case these instructions are considered
17355 arithmetic and avoided where correct operation is required and the
17356 input operand might be a NaN. A longer sequence of instructions that
17357 manipulate the sign bit of floating-point datum manually is used
17358 instead unless the @option{-ffinite-math-only} option has also been
17361 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17362 this case these instructions are considered non-arithmetic and therefore
17363 operating correctly in all cases, including in particular where the
17364 input operand is a NaN. These instructions are therefore always used
17365 for the respective operations.
17368 @itemx -mnan=legacy
17370 @opindex mnan=legacy
17371 These options control the encoding of the special not-a-number (NaN)
17372 IEEE 754 floating-point data.
17374 The @option{-mnan=legacy} option selects the legacy encoding. In this
17375 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17376 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17377 by the first bit of their trailing significand field being 1.
17379 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17380 this case qNaNs are denoted by the first bit of their trailing
17381 significand field being 1, whereas sNaNs are denoted by the first bit of
17382 their trailing significand field being 0.
17384 The default is @option{-mnan=legacy} unless GCC has been configured with
17385 @option{--with-nan=2008}.
17391 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17392 implement atomic memory built-in functions. When neither option is
17393 specified, GCC uses the instructions if the target architecture
17396 @option{-mllsc} is useful if the runtime environment can emulate the
17397 instructions and @option{-mno-llsc} can be useful when compiling for
17398 nonstandard ISAs. You can make either option the default by
17399 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17400 respectively. @option{--with-llsc} is the default for some
17401 configurations; see the installation documentation for details.
17407 Use (do not use) revision 1 of the MIPS DSP ASE@.
17408 @xref{MIPS DSP Built-in Functions}. This option defines the
17409 preprocessor macro @samp{__mips_dsp}. It also defines
17410 @samp{__mips_dsp_rev} to 1.
17416 Use (do not use) revision 2 of the MIPS DSP ASE@.
17417 @xref{MIPS DSP Built-in Functions}. This option defines the
17418 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
17419 It also defines @samp{__mips_dsp_rev} to 2.
17422 @itemx -mno-smartmips
17423 @opindex msmartmips
17424 @opindex mno-smartmips
17425 Use (do not use) the MIPS SmartMIPS ASE.
17427 @item -mpaired-single
17428 @itemx -mno-paired-single
17429 @opindex mpaired-single
17430 @opindex mno-paired-single
17431 Use (do not use) paired-single floating-point instructions.
17432 @xref{MIPS Paired-Single Support}. This option requires
17433 hardware floating-point support to be enabled.
17439 Use (do not use) MIPS Digital Media Extension instructions.
17440 This option can only be used when generating 64-bit code and requires
17441 hardware floating-point support to be enabled.
17446 @opindex mno-mips3d
17447 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17448 The option @option{-mips3d} implies @option{-mpaired-single}.
17451 @itemx -mno-micromips
17452 @opindex mmicromips
17453 @opindex mno-mmicromips
17454 Generate (do not generate) microMIPS code.
17456 MicroMIPS code generation can also be controlled on a per-function basis
17457 by means of @code{micromips} and @code{nomicromips} attributes.
17458 @xref{Function Attributes}, for more information.
17464 Use (do not use) MT Multithreading instructions.
17470 Use (do not use) the MIPS MCU ASE instructions.
17476 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17482 Use (do not use) the MIPS Virtualization Application Specific instructions.
17486 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17487 an explanation of the default and the way that the pointer size is
17492 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17494 The default size of @code{int}s, @code{long}s and pointers depends on
17495 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17496 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17497 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17498 or the same size as integer registers, whichever is smaller.
17504 Assume (do not assume) that all symbols have 32-bit values, regardless
17505 of the selected ABI@. This option is useful in combination with
17506 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17507 to generate shorter and faster references to symbolic addresses.
17511 Put definitions of externally-visible data in a small data section
17512 if that data is no bigger than @var{num} bytes. GCC can then generate
17513 more efficient accesses to the data; see @option{-mgpopt} for details.
17515 The default @option{-G} option depends on the configuration.
17517 @item -mlocal-sdata
17518 @itemx -mno-local-sdata
17519 @opindex mlocal-sdata
17520 @opindex mno-local-sdata
17521 Extend (do not extend) the @option{-G} behavior to local data too,
17522 such as to static variables in C@. @option{-mlocal-sdata} is the
17523 default for all configurations.
17525 If the linker complains that an application is using too much small data,
17526 you might want to try rebuilding the less performance-critical parts with
17527 @option{-mno-local-sdata}. You might also want to build large
17528 libraries with @option{-mno-local-sdata}, so that the libraries leave
17529 more room for the main program.
17531 @item -mextern-sdata
17532 @itemx -mno-extern-sdata
17533 @opindex mextern-sdata
17534 @opindex mno-extern-sdata
17535 Assume (do not assume) that externally-defined data is in
17536 a small data section if the size of that data is within the @option{-G} limit.
17537 @option{-mextern-sdata} is the default for all configurations.
17539 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17540 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17541 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17542 is placed in a small data section. If @var{Var} is defined by another
17543 module, you must either compile that module with a high-enough
17544 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17545 definition. If @var{Var} is common, you must link the application
17546 with a high-enough @option{-G} setting.
17548 The easiest way of satisfying these restrictions is to compile
17549 and link every module with the same @option{-G} option. However,
17550 you may wish to build a library that supports several different
17551 small data limits. You can do this by compiling the library with
17552 the highest supported @option{-G} setting and additionally using
17553 @option{-mno-extern-sdata} to stop the library from making assumptions
17554 about externally-defined data.
17560 Use (do not use) GP-relative accesses for symbols that are known to be
17561 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17562 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17565 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17566 might not hold the value of @code{_gp}. For example, if the code is
17567 part of a library that might be used in a boot monitor, programs that
17568 call boot monitor routines pass an unknown value in @code{$gp}.
17569 (In such situations, the boot monitor itself is usually compiled
17570 with @option{-G0}.)
17572 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17573 @option{-mno-extern-sdata}.
17575 @item -membedded-data
17576 @itemx -mno-embedded-data
17577 @opindex membedded-data
17578 @opindex mno-embedded-data
17579 Allocate variables to the read-only data section first if possible, then
17580 next in the small data section if possible, otherwise in data. This gives
17581 slightly slower code than the default, but reduces the amount of RAM required
17582 when executing, and thus may be preferred for some embedded systems.
17584 @item -muninit-const-in-rodata
17585 @itemx -mno-uninit-const-in-rodata
17586 @opindex muninit-const-in-rodata
17587 @opindex mno-uninit-const-in-rodata
17588 Put uninitialized @code{const} variables in the read-only data section.
17589 This option is only meaningful in conjunction with @option{-membedded-data}.
17591 @item -mcode-readable=@var{setting}
17592 @opindex mcode-readable
17593 Specify whether GCC may generate code that reads from executable sections.
17594 There are three possible settings:
17597 @item -mcode-readable=yes
17598 Instructions may freely access executable sections. This is the
17601 @item -mcode-readable=pcrel
17602 MIPS16 PC-relative load instructions can access executable sections,
17603 but other instructions must not do so. This option is useful on 4KSc
17604 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17605 It is also useful on processors that can be configured to have a dual
17606 instruction/data SRAM interface and that, like the M4K, automatically
17607 redirect PC-relative loads to the instruction RAM.
17609 @item -mcode-readable=no
17610 Instructions must not access executable sections. This option can be
17611 useful on targets that are configured to have a dual instruction/data
17612 SRAM interface but that (unlike the M4K) do not automatically redirect
17613 PC-relative loads to the instruction RAM.
17616 @item -msplit-addresses
17617 @itemx -mno-split-addresses
17618 @opindex msplit-addresses
17619 @opindex mno-split-addresses
17620 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17621 relocation operators. This option has been superseded by
17622 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17624 @item -mexplicit-relocs
17625 @itemx -mno-explicit-relocs
17626 @opindex mexplicit-relocs
17627 @opindex mno-explicit-relocs
17628 Use (do not use) assembler relocation operators when dealing with symbolic
17629 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17630 is to use assembler macros instead.
17632 @option{-mexplicit-relocs} is the default if GCC was configured
17633 to use an assembler that supports relocation operators.
17635 @item -mcheck-zero-division
17636 @itemx -mno-check-zero-division
17637 @opindex mcheck-zero-division
17638 @opindex mno-check-zero-division
17639 Trap (do not trap) on integer division by zero.
17641 The default is @option{-mcheck-zero-division}.
17643 @item -mdivide-traps
17644 @itemx -mdivide-breaks
17645 @opindex mdivide-traps
17646 @opindex mdivide-breaks
17647 MIPS systems check for division by zero by generating either a
17648 conditional trap or a break instruction. Using traps results in
17649 smaller code, but is only supported on MIPS II and later. Also, some
17650 versions of the Linux kernel have a bug that prevents trap from
17651 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17652 allow conditional traps on architectures that support them and
17653 @option{-mdivide-breaks} to force the use of breaks.
17655 The default is usually @option{-mdivide-traps}, but this can be
17656 overridden at configure time using @option{--with-divide=breaks}.
17657 Divide-by-zero checks can be completely disabled using
17658 @option{-mno-check-zero-division}.
17663 @opindex mno-memcpy
17664 Force (do not force) the use of @code{memcpy()} for non-trivial block
17665 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17666 most constant-sized copies.
17669 @itemx -mno-long-calls
17670 @opindex mlong-calls
17671 @opindex mno-long-calls
17672 Disable (do not disable) use of the @code{jal} instruction. Calling
17673 functions using @code{jal} is more efficient but requires the caller
17674 and callee to be in the same 256 megabyte segment.
17676 This option has no effect on abicalls code. The default is
17677 @option{-mno-long-calls}.
17683 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17684 instructions, as provided by the R4650 ISA@.
17690 Enable (disable) use of the @code{madd} and @code{msub} integer
17691 instructions. The default is @option{-mimadd} on architectures
17692 that support @code{madd} and @code{msub} except for the 74k
17693 architecture where it was found to generate slower code.
17696 @itemx -mno-fused-madd
17697 @opindex mfused-madd
17698 @opindex mno-fused-madd
17699 Enable (disable) use of the floating-point multiply-accumulate
17700 instructions, when they are available. The default is
17701 @option{-mfused-madd}.
17703 On the R8000 CPU when multiply-accumulate instructions are used,
17704 the intermediate product is calculated to infinite precision
17705 and is not subject to the FCSR Flush to Zero bit. This may be
17706 undesirable in some circumstances. On other processors the result
17707 is numerically identical to the equivalent computation using
17708 separate multiply, add, subtract and negate instructions.
17712 Tell the MIPS assembler to not run its preprocessor over user
17713 assembler files (with a @samp{.s} suffix) when assembling them.
17718 @opindex mno-fix-24k
17719 Work around the 24K E48 (lost data on stores during refill) errata.
17720 The workarounds are implemented by the assembler rather than by GCC@.
17723 @itemx -mno-fix-r4000
17724 @opindex mfix-r4000
17725 @opindex mno-fix-r4000
17726 Work around certain R4000 CPU errata:
17729 A double-word or a variable shift may give an incorrect result if executed
17730 immediately after starting an integer division.
17732 A double-word or a variable shift may give an incorrect result if executed
17733 while an integer multiplication is in progress.
17735 An integer division may give an incorrect result if started in a delay slot
17736 of a taken branch or a jump.
17740 @itemx -mno-fix-r4400
17741 @opindex mfix-r4400
17742 @opindex mno-fix-r4400
17743 Work around certain R4400 CPU errata:
17746 A double-word or a variable shift may give an incorrect result if executed
17747 immediately after starting an integer division.
17751 @itemx -mno-fix-r10000
17752 @opindex mfix-r10000
17753 @opindex mno-fix-r10000
17754 Work around certain R10000 errata:
17757 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17758 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17761 This option can only be used if the target architecture supports
17762 branch-likely instructions. @option{-mfix-r10000} is the default when
17763 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17767 @itemx -mno-fix-rm7000
17768 @opindex mfix-rm7000
17769 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17770 workarounds are implemented by the assembler rather than by GCC@.
17773 @itemx -mno-fix-vr4120
17774 @opindex mfix-vr4120
17775 Work around certain VR4120 errata:
17778 @code{dmultu} does not always produce the correct result.
17780 @code{div} and @code{ddiv} do not always produce the correct result if one
17781 of the operands is negative.
17783 The workarounds for the division errata rely on special functions in
17784 @file{libgcc.a}. At present, these functions are only provided by
17785 the @code{mips64vr*-elf} configurations.
17787 Other VR4120 errata require a NOP to be inserted between certain pairs of
17788 instructions. These errata are handled by the assembler, not by GCC itself.
17791 @opindex mfix-vr4130
17792 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17793 workarounds are implemented by the assembler rather than by GCC,
17794 although GCC avoids using @code{mflo} and @code{mfhi} if the
17795 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17796 instructions are available instead.
17799 @itemx -mno-fix-sb1
17801 Work around certain SB-1 CPU core errata.
17802 (This flag currently works around the SB-1 revision 2
17803 ``F1'' and ``F2'' floating-point errata.)
17805 @item -mr10k-cache-barrier=@var{setting}
17806 @opindex mr10k-cache-barrier
17807 Specify whether GCC should insert cache barriers to avoid the
17808 side-effects of speculation on R10K processors.
17810 In common with many processors, the R10K tries to predict the outcome
17811 of a conditional branch and speculatively executes instructions from
17812 the ``taken'' branch. It later aborts these instructions if the
17813 predicted outcome is wrong. However, on the R10K, even aborted
17814 instructions can have side effects.
17816 This problem only affects kernel stores and, depending on the system,
17817 kernel loads. As an example, a speculatively-executed store may load
17818 the target memory into cache and mark the cache line as dirty, even if
17819 the store itself is later aborted. If a DMA operation writes to the
17820 same area of memory before the ``dirty'' line is flushed, the cached
17821 data overwrites the DMA-ed data. See the R10K processor manual
17822 for a full description, including other potential problems.
17824 One workaround is to insert cache barrier instructions before every memory
17825 access that might be speculatively executed and that might have side
17826 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17827 controls GCC's implementation of this workaround. It assumes that
17828 aborted accesses to any byte in the following regions does not have
17833 the memory occupied by the current function's stack frame;
17836 the memory occupied by an incoming stack argument;
17839 the memory occupied by an object with a link-time-constant address.
17842 It is the kernel's responsibility to ensure that speculative
17843 accesses to these regions are indeed safe.
17845 If the input program contains a function declaration such as:
17851 then the implementation of @code{foo} must allow @code{j foo} and
17852 @code{jal foo} to be executed speculatively. GCC honors this
17853 restriction for functions it compiles itself. It expects non-GCC
17854 functions (such as hand-written assembly code) to do the same.
17856 The option has three forms:
17859 @item -mr10k-cache-barrier=load-store
17860 Insert a cache barrier before a load or store that might be
17861 speculatively executed and that might have side effects even
17864 @item -mr10k-cache-barrier=store
17865 Insert a cache barrier before a store that might be speculatively
17866 executed and that might have side effects even if aborted.
17868 @item -mr10k-cache-barrier=none
17869 Disable the insertion of cache barriers. This is the default setting.
17872 @item -mflush-func=@var{func}
17873 @itemx -mno-flush-func
17874 @opindex mflush-func
17875 Specifies the function to call to flush the I and D caches, or to not
17876 call any such function. If called, the function must take the same
17877 arguments as the common @code{_flush_func()}, that is, the address of the
17878 memory range for which the cache is being flushed, the size of the
17879 memory range, and the number 3 (to flush both caches). The default
17880 depends on the target GCC was configured for, but commonly is either
17881 @samp{_flush_func} or @samp{__cpu_flush}.
17883 @item mbranch-cost=@var{num}
17884 @opindex mbranch-cost
17885 Set the cost of branches to roughly @var{num} ``simple'' instructions.
17886 This cost is only a heuristic and is not guaranteed to produce
17887 consistent results across releases. A zero cost redundantly selects
17888 the default, which is based on the @option{-mtune} setting.
17890 @item -mbranch-likely
17891 @itemx -mno-branch-likely
17892 @opindex mbranch-likely
17893 @opindex mno-branch-likely
17894 Enable or disable use of Branch Likely instructions, regardless of the
17895 default for the selected architecture. By default, Branch Likely
17896 instructions may be generated if they are supported by the selected
17897 architecture. An exception is for the MIPS32 and MIPS64 architectures
17898 and processors that implement those architectures; for those, Branch
17899 Likely instructions are not be generated by default because the MIPS32
17900 and MIPS64 architectures specifically deprecate their use.
17902 @item -mfp-exceptions
17903 @itemx -mno-fp-exceptions
17904 @opindex mfp-exceptions
17905 Specifies whether FP exceptions are enabled. This affects how
17906 FP instructions are scheduled for some processors.
17907 The default is that FP exceptions are
17910 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
17911 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
17914 @item -mvr4130-align
17915 @itemx -mno-vr4130-align
17916 @opindex mvr4130-align
17917 The VR4130 pipeline is two-way superscalar, but can only issue two
17918 instructions together if the first one is 8-byte aligned. When this
17919 option is enabled, GCC aligns pairs of instructions that it
17920 thinks should execute in parallel.
17922 This option only has an effect when optimizing for the VR4130.
17923 It normally makes code faster, but at the expense of making it bigger.
17924 It is enabled by default at optimization level @option{-O3}.
17929 Enable (disable) generation of @code{synci} instructions on
17930 architectures that support it. The @code{synci} instructions (if
17931 enabled) are generated when @code{__builtin___clear_cache()} is
17934 This option defaults to @code{-mno-synci}, but the default can be
17935 overridden by configuring with @code{--with-synci}.
17937 When compiling code for single processor systems, it is generally safe
17938 to use @code{synci}. However, on many multi-core (SMP) systems, it
17939 does not invalidate the instruction caches on all cores and may lead
17940 to undefined behavior.
17942 @item -mrelax-pic-calls
17943 @itemx -mno-relax-pic-calls
17944 @opindex mrelax-pic-calls
17945 Try to turn PIC calls that are normally dispatched via register
17946 @code{$25} into direct calls. This is only possible if the linker can
17947 resolve the destination at link-time and if the destination is within
17948 range for a direct call.
17950 @option{-mrelax-pic-calls} is the default if GCC was configured to use
17951 an assembler and a linker that support the @code{.reloc} assembly
17952 directive and @code{-mexplicit-relocs} is in effect. With
17953 @code{-mno-explicit-relocs}, this optimization can be performed by the
17954 assembler and the linker alone without help from the compiler.
17956 @item -mmcount-ra-address
17957 @itemx -mno-mcount-ra-address
17958 @opindex mmcount-ra-address
17959 @opindex mno-mcount-ra-address
17960 Emit (do not emit) code that allows @code{_mcount} to modify the
17961 calling function's return address. When enabled, this option extends
17962 the usual @code{_mcount} interface with a new @var{ra-address}
17963 parameter, which has type @code{intptr_t *} and is passed in register
17964 @code{$12}. @code{_mcount} can then modify the return address by
17965 doing both of the following:
17968 Returning the new address in register @code{$31}.
17970 Storing the new address in @code{*@var{ra-address}},
17971 if @var{ra-address} is nonnull.
17974 The default is @option{-mno-mcount-ra-address}.
17979 @subsection MMIX Options
17980 @cindex MMIX Options
17982 These options are defined for the MMIX:
17986 @itemx -mno-libfuncs
17988 @opindex mno-libfuncs
17989 Specify that intrinsic library functions are being compiled, passing all
17990 values in registers, no matter the size.
17993 @itemx -mno-epsilon
17995 @opindex mno-epsilon
17996 Generate floating-point comparison instructions that compare with respect
17997 to the @code{rE} epsilon register.
17999 @item -mabi=mmixware
18001 @opindex mabi=mmixware
18003 Generate code that passes function parameters and return values that (in
18004 the called function) are seen as registers @code{$0} and up, as opposed to
18005 the GNU ABI which uses global registers @code{$231} and up.
18007 @item -mzero-extend
18008 @itemx -mno-zero-extend
18009 @opindex mzero-extend
18010 @opindex mno-zero-extend
18011 When reading data from memory in sizes shorter than 64 bits, use (do not
18012 use) zero-extending load instructions by default, rather than
18013 sign-extending ones.
18016 @itemx -mno-knuthdiv
18018 @opindex mno-knuthdiv
18019 Make the result of a division yielding a remainder have the same sign as
18020 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18021 remainder follows the sign of the dividend. Both methods are
18022 arithmetically valid, the latter being almost exclusively used.
18024 @item -mtoplevel-symbols
18025 @itemx -mno-toplevel-symbols
18026 @opindex mtoplevel-symbols
18027 @opindex mno-toplevel-symbols
18028 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18029 code can be used with the @code{PREFIX} assembly directive.
18033 Generate an executable in the ELF format, rather than the default
18034 @samp{mmo} format used by the @command{mmix} simulator.
18036 @item -mbranch-predict
18037 @itemx -mno-branch-predict
18038 @opindex mbranch-predict
18039 @opindex mno-branch-predict
18040 Use (do not use) the probable-branch instructions, when static branch
18041 prediction indicates a probable branch.
18043 @item -mbase-addresses
18044 @itemx -mno-base-addresses
18045 @opindex mbase-addresses
18046 @opindex mno-base-addresses
18047 Generate (do not generate) code that uses @emph{base addresses}. Using a
18048 base address automatically generates a request (handled by the assembler
18049 and the linker) for a constant to be set up in a global register. The
18050 register is used for one or more base address requests within the range 0
18051 to 255 from the value held in the register. The generally leads to short
18052 and fast code, but the number of different data items that can be
18053 addressed is limited. This means that a program that uses lots of static
18054 data may require @option{-mno-base-addresses}.
18056 @item -msingle-exit
18057 @itemx -mno-single-exit
18058 @opindex msingle-exit
18059 @opindex mno-single-exit
18060 Force (do not force) generated code to have a single exit point in each
18064 @node MN10300 Options
18065 @subsection MN10300 Options
18066 @cindex MN10300 options
18068 These @option{-m} options are defined for Matsushita MN10300 architectures:
18073 Generate code to avoid bugs in the multiply instructions for the MN10300
18074 processors. This is the default.
18076 @item -mno-mult-bug
18077 @opindex mno-mult-bug
18078 Do not generate code to avoid bugs in the multiply instructions for the
18079 MN10300 processors.
18083 Generate code using features specific to the AM33 processor.
18087 Do not generate code using features specific to the AM33 processor. This
18092 Generate code using features specific to the AM33/2.0 processor.
18096 Generate code using features specific to the AM34 processor.
18098 @item -mtune=@var{cpu-type}
18100 Use the timing characteristics of the indicated CPU type when
18101 scheduling instructions. This does not change the targeted processor
18102 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18103 @samp{am33-2} or @samp{am34}.
18105 @item -mreturn-pointer-on-d0
18106 @opindex mreturn-pointer-on-d0
18107 When generating a function that returns a pointer, return the pointer
18108 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18109 only in @code{a0}, and attempts to call such functions without a prototype
18110 result in errors. Note that this option is on by default; use
18111 @option{-mno-return-pointer-on-d0} to disable it.
18115 Do not link in the C run-time initialization object file.
18119 Indicate to the linker that it should perform a relaxation optimization pass
18120 to shorten branches, calls and absolute memory addresses. This option only
18121 has an effect when used on the command line for the final link step.
18123 This option makes symbolic debugging impossible.
18127 Allow the compiler to generate @emph{Long Instruction Word}
18128 instructions if the target is the @samp{AM33} or later. This is the
18129 default. This option defines the preprocessor macro @samp{__LIW__}.
18133 Do not allow the compiler to generate @emph{Long Instruction Word}
18134 instructions. This option defines the preprocessor macro
18139 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18140 instructions if the target is the @samp{AM33} or later. This is the
18141 default. This option defines the preprocessor macro @samp{__SETLB__}.
18145 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18146 instructions. This option defines the preprocessor macro
18147 @samp{__NO_SETLB__}.
18151 @node Moxie Options
18152 @subsection Moxie Options
18153 @cindex Moxie Options
18159 Generate big-endian code. This is the default for @samp{moxie-*-*}
18164 Generate little-endian code.
18168 Do not link in the C run-time initialization object file.
18172 @node MSP430 Options
18173 @subsection MSP430 Options
18174 @cindex MSP430 Options
18176 These options are defined for the MSP430:
18182 Force assembly output to always use hex constants. Normally such
18183 constants are signed decimals, but this option is available for
18184 testsuite and/or aesthetic purposes.
18188 Select the MCU to target. This is used to create a C preprocessor
18189 symbol based upon the MCU name, converted to upper case and pre- and
18190 post- fixed with @code{__}. This in turn will be used by the
18191 @code{msp430.h} header file to select an MCU specific supplimentary
18194 The option also sets the ISA to use. If the MCU name is one that is
18195 known to only support the 430 ISA then that is selected, otherwise the
18196 430X ISA is selected. A generic MCU name of @code{msp430} can also be
18197 used to select the 430 ISA. Similarly the generic @code{msp430x} MCU
18198 name will select the 430X ISA.
18200 In addition an MCU specific linker script will be added to the linker
18201 command line. The script's name is the name of the MCU with
18202 @code{.ld} appended. Thus specifying @option{-mmcu=xxx} on the gcc
18203 command line will define the C preprocessor symbol @code{__XXX__} and
18204 cause the linker to search for a script called @file{xxx.ld}.
18206 This option is also passed on to the assembler.
18210 Specifies the ISA to use. Accepted values are @code{msp430},
18211 @code{msp430x} and @code{msp430xv2}. This option is deprecated. The
18212 @option{-mmcu=} option should be used to select the ISA.
18216 Link to the simulator runtime libraries and linker script. Overrides
18217 any scripts that would be selected by the @option{-mmcu=} option.
18221 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18225 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18229 This option is passed to the assembler and linker, and allows the
18230 linker to perform certain optimizations that cannot be done until
18235 @node NDS32 Options
18236 @subsection NDS32 Options
18237 @cindex NDS32 Options
18239 These options are defined for NDS32 implementations:
18244 @opindex mbig-endian
18245 Generate code in big-endian mode.
18247 @item -mlittle-endian
18248 @opindex mlittle-endian
18249 Generate code in little-endian mode.
18251 @item -mreduced-regs
18252 @opindex mreduced-regs
18253 Use reduced-set registers for register allocation.
18256 @opindex mfull-regs
18257 Use full-set registers for register allocation.
18261 Generate conditional move instructions.
18265 Do not generate conditional move instructions.
18269 Generate performance extension instructions.
18271 @item -mno-perf-ext
18272 @opindex mno-perf-ext
18273 Do not generate performance extension instructions.
18277 Generate v3 push25/pop25 instructions.
18280 @opindex mno-v3push
18281 Do not generate v3 push25/pop25 instructions.
18285 Generate 16-bit instructions.
18288 @opindex mno-16-bit
18289 Do not generate 16-bit instructions.
18292 @opindex mgp-direct
18293 Generate GP base instructions directly.
18295 @item -mno-gp-direct
18296 @opindex mno-gp-direct
18297 Do no generate GP base instructions directly.
18299 @item -misr-vector-size=@var{num}
18300 @opindex misr-vector-size
18301 Specify the size of each interrupt vector, which must be 4 or 16.
18303 @item -mcache-block-size=@var{num}
18304 @opindex mcache-block-size
18305 Specify the size of each cache block,
18306 which must be a power of 2 between 4 and 512.
18308 @item -march=@var{arch}
18310 Specify the name of the target architecture.
18312 @item -mforce-fp-as-gp
18313 @opindex mforce-fp-as-gp
18314 Prevent $fp being allocated during register allocation so that compiler
18315 is able to force performing fp-as-gp optimization.
18317 @item -mforbid-fp-as-gp
18318 @opindex mforbid-fp-as-gp
18319 Forbid using $fp to access static and global variables.
18320 This option strictly forbids fp-as-gp optimization
18321 regardless of @option{-mforce-fp-as-gp}.
18325 Use special directives to guide linker doing ex9 optimization.
18328 @opindex mctor-dtor
18329 Enable constructor/destructor feature.
18333 Guide linker to relax instructions.
18337 @node Nios II Options
18338 @subsection Nios II Options
18339 @cindex Nios II options
18340 @cindex Altera Nios II options
18342 These are the options defined for the Altera Nios II processor.
18348 @cindex smaller data references
18349 Put global and static objects less than or equal to @var{num} bytes
18350 into the small data or BSS sections instead of the normal data or BSS
18351 sections. The default value of @var{num} is 8.
18357 Generate (do not generate) GP-relative accesses for objects in the
18358 small data or BSS sections. The default is @option{-mgpopt} except
18359 when @option{-fpic} or @option{-fPIC} is specified to generate
18360 position-independent code. Note that the Nios II ABI does not permit
18361 GP-relative accesses from shared libraries.
18363 You may need to specify @option{-mno-gpopt} explicitly when building
18364 programs that include large amounts of small data, including large
18365 GOT data sections. In this case, the 16-bit offset for GP-relative
18366 addressing may not be large enough to allow access to the entire
18367 small data section.
18373 Generate little-endian (default) or big-endian (experimental) code,
18376 @item -mbypass-cache
18377 @itemx -mno-bypass-cache
18378 @opindex mno-bypass-cache
18379 @opindex mbypass-cache
18380 Force all load and store instructions to always bypass cache by
18381 using I/O variants of the instructions. The default is not to
18384 @item -mno-cache-volatile
18385 @itemx -mcache-volatile
18386 @opindex mcache-volatile
18387 @opindex mno-cache-volatile
18388 Volatile memory access bypass the cache using the I/O variants of
18389 the load and store instructions. The default is not to bypass the cache.
18391 @item -mno-fast-sw-div
18392 @itemx -mfast-sw-div
18393 @opindex mno-fast-sw-div
18394 @opindex mfast-sw-div
18395 Do not use table-based fast divide for small numbers. The default
18396 is to use the fast divide at @option{-O3} and above.
18400 @itemx -mno-hw-mulx
18404 @opindex mno-hw-mul
18406 @opindex mno-hw-mulx
18408 @opindex mno-hw-div
18410 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18411 instructions by the compiler. The default is to emit @code{mul}
18412 and not emit @code{div} and @code{mulx}.
18414 @item -mcustom-@var{insn}=@var{N}
18415 @itemx -mno-custom-@var{insn}
18416 @opindex mcustom-@var{insn}
18417 @opindex mno-custom-@var{insn}
18418 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18419 custom instruction with encoding @var{N} when generating code that uses
18420 @var{insn}. For example, @code{-mcustom-fadds=253} generates custom
18421 instruction 253 for single-precision floating-point add operations instead
18422 of the default behavior of using a library call.
18424 The following values of @var{insn} are supported. Except as otherwise
18425 noted, floating-point operations are expected to be implemented with
18426 normal IEEE 754 semantics and correspond directly to the C operators or the
18427 equivalent GCC built-in functions (@pxref{Other Builtins}).
18429 Single-precision floating point:
18432 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18433 Binary arithmetic operations.
18439 Unary absolute value.
18441 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18442 Comparison operations.
18444 @item @samp{fmins}, @samp{fmaxs}
18445 Floating-point minimum and maximum. These instructions are only
18446 generated if @option{-ffinite-math-only} is specified.
18448 @item @samp{fsqrts}
18449 Unary square root operation.
18451 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18452 Floating-point trigonometric and exponential functions. These instructions
18453 are only generated if @option{-funsafe-math-optimizations} is also specified.
18457 Double-precision floating point:
18460 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18461 Binary arithmetic operations.
18467 Unary absolute value.
18469 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18470 Comparison operations.
18472 @item @samp{fmind}, @samp{fmaxd}
18473 Double-precision minimum and maximum. These instructions are only
18474 generated if @option{-ffinite-math-only} is specified.
18476 @item @samp{fsqrtd}
18477 Unary square root operation.
18479 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18480 Double-precision trigonometric and exponential functions. These instructions
18481 are only generated if @option{-funsafe-math-optimizations} is also specified.
18487 @item @samp{fextsd}
18488 Conversion from single precision to double precision.
18490 @item @samp{ftruncds}
18491 Conversion from double precision to single precision.
18493 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18494 Conversion from floating point to signed or unsigned integer types, with
18495 truncation towards zero.
18497 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18498 Conversion from signed or unsigned integer types to floating-point types.
18502 In addition, all of the following transfer instructions for internal
18503 registers X and Y must be provided to use any of the double-precision
18504 floating-point instructions. Custom instructions taking two
18505 double-precision source operands expect the first operand in the
18506 64-bit register X. The other operand (or only operand of a unary
18507 operation) is given to the custom arithmetic instruction with the
18508 least significant half in source register @var{src1} and the most
18509 significant half in @var{src2}. A custom instruction that returns a
18510 double-precision result returns the most significant 32 bits in the
18511 destination register and the other half in 32-bit register Y.
18512 GCC automatically generates the necessary code sequences to write
18513 register X and/or read register Y when double-precision floating-point
18514 instructions are used.
18519 Write @var{src1} into the least significant half of X and @var{src2} into
18520 the most significant half of X.
18523 Write @var{src1} into Y.
18525 @item @samp{frdxhi}, @samp{frdxlo}
18526 Read the most or least (respectively) significant half of X and store it in
18530 Read the value of Y and store it into @var{dest}.
18533 Note that you can gain more local control over generation of Nios II custom
18534 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18535 and @code{target("no-custom-@var{insn}")} function attributes
18536 (@pxref{Function Attributes})
18537 or pragmas (@pxref{Function Specific Option Pragmas}).
18539 @item -mcustom-fpu-cfg=@var{name}
18540 @opindex mcustom-fpu-cfg
18542 This option enables a predefined, named set of custom instruction encodings
18543 (see @option{-mcustom-@var{insn}} above).
18544 Currently, the following sets are defined:
18546 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18547 @gccoptlist{-mcustom-fmuls=252 @gol
18548 -mcustom-fadds=253 @gol
18549 -mcustom-fsubs=254 @gol
18550 -fsingle-precision-constant}
18552 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18553 @gccoptlist{-mcustom-fmuls=252 @gol
18554 -mcustom-fadds=253 @gol
18555 -mcustom-fsubs=254 @gol
18556 -mcustom-fdivs=255 @gol
18557 -fsingle-precision-constant}
18559 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18560 @gccoptlist{-mcustom-floatus=243 @gol
18561 -mcustom-fixsi=244 @gol
18562 -mcustom-floatis=245 @gol
18563 -mcustom-fcmpgts=246 @gol
18564 -mcustom-fcmples=249 @gol
18565 -mcustom-fcmpeqs=250 @gol
18566 -mcustom-fcmpnes=251 @gol
18567 -mcustom-fmuls=252 @gol
18568 -mcustom-fadds=253 @gol
18569 -mcustom-fsubs=254 @gol
18570 -mcustom-fdivs=255 @gol
18571 -fsingle-precision-constant}
18573 Custom instruction assignments given by individual
18574 @option{-mcustom-@var{insn}=} options override those given by
18575 @option{-mcustom-fpu-cfg=}, regardless of the
18576 order of the options on the command line.
18578 Note that you can gain more local control over selection of a FPU
18579 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18580 function attribute (@pxref{Function Attributes})
18581 or pragma (@pxref{Function Specific Option Pragmas}).
18585 These additional @samp{-m} options are available for the Altera Nios II
18586 ELF (bare-metal) target:
18592 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18593 startup and termination code, and is typically used in conjunction with
18594 @option{-msys-crt0=} to specify the location of the alternate startup code
18595 provided by the HAL BSP.
18599 Link with a limited version of the C library, @option{-lsmallc}, rather than
18602 @item -msys-crt0=@var{startfile}
18604 @var{startfile} is the file name of the startfile (crt0) to use
18605 when linking. This option is only useful in conjunction with @option{-mhal}.
18607 @item -msys-lib=@var{systemlib}
18609 @var{systemlib} is the library name of the library that provides
18610 low-level system calls required by the C library,
18611 e.g. @code{read} and @code{write}.
18612 This option is typically used to link with a library provided by a HAL BSP.
18616 @node PDP-11 Options
18617 @subsection PDP-11 Options
18618 @cindex PDP-11 Options
18620 These options are defined for the PDP-11:
18625 Use hardware FPP floating point. This is the default. (FIS floating
18626 point on the PDP-11/40 is not supported.)
18629 @opindex msoft-float
18630 Do not use hardware floating point.
18634 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18638 Return floating-point results in memory. This is the default.
18642 Generate code for a PDP-11/40.
18646 Generate code for a PDP-11/45. This is the default.
18650 Generate code for a PDP-11/10.
18652 @item -mbcopy-builtin
18653 @opindex mbcopy-builtin
18654 Use inline @code{movmemhi} patterns for copying memory. This is the
18659 Do not use inline @code{movmemhi} patterns for copying memory.
18665 Use 16-bit @code{int}. This is the default.
18671 Use 32-bit @code{int}.
18674 @itemx -mno-float32
18676 @opindex mno-float32
18677 Use 64-bit @code{float}. This is the default.
18680 @itemx -mno-float64
18682 @opindex mno-float64
18683 Use 32-bit @code{float}.
18687 Use @code{abshi2} pattern. This is the default.
18691 Do not use @code{abshi2} pattern.
18693 @item -mbranch-expensive
18694 @opindex mbranch-expensive
18695 Pretend that branches are expensive. This is for experimenting with
18696 code generation only.
18698 @item -mbranch-cheap
18699 @opindex mbranch-cheap
18700 Do not pretend that branches are expensive. This is the default.
18704 Use Unix assembler syntax. This is the default when configured for
18705 @samp{pdp11-*-bsd}.
18709 Use DEC assembler syntax. This is the default when configured for any
18710 PDP-11 target other than @samp{pdp11-*-bsd}.
18713 @node picoChip Options
18714 @subsection picoChip Options
18715 @cindex picoChip options
18717 These @samp{-m} options are defined for picoChip implementations:
18721 @item -mae=@var{ae_type}
18723 Set the instruction set, register set, and instruction scheduling
18724 parameters for array element type @var{ae_type}. Supported values
18725 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18727 @option{-mae=ANY} selects a completely generic AE type. Code
18728 generated with this option runs on any of the other AE types. The
18729 code is not as efficient as it would be if compiled for a specific
18730 AE type, and some types of operation (e.g., multiplication) do not
18731 work properly on all types of AE.
18733 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
18734 for compiled code, and is the default.
18736 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
18737 option may suffer from poor performance of byte (char) manipulation,
18738 since the DSP AE does not provide hardware support for byte load/stores.
18740 @item -msymbol-as-address
18741 Enable the compiler to directly use a symbol name as an address in a
18742 load/store instruction, without first loading it into a
18743 register. Typically, the use of this option generates larger
18744 programs, which run faster than when the option isn't used. However, the
18745 results vary from program to program, so it is left as a user option,
18746 rather than being permanently enabled.
18748 @item -mno-inefficient-warnings
18749 Disables warnings about the generation of inefficient code. These
18750 warnings can be generated, for example, when compiling code that
18751 performs byte-level memory operations on the MAC AE type. The MAC AE has
18752 no hardware support for byte-level memory operations, so all byte
18753 load/stores must be synthesized from word load/store operations. This is
18754 inefficient and a warning is generated to indicate
18755 that you should rewrite the code to avoid byte operations, or to target
18756 an AE type that has the necessary hardware support. This option disables
18761 @node PowerPC Options
18762 @subsection PowerPC Options
18763 @cindex PowerPC options
18765 These are listed under @xref{RS/6000 and PowerPC Options}.
18768 @subsection RL78 Options
18769 @cindex RL78 Options
18775 Links in additional target libraries to support operation within a
18782 Specifies the type of hardware multiplication support to be used. The
18783 default is @code{none}, which uses software multiplication functions.
18784 The @code{g13} option is for the hardware multiply/divide peripheral
18785 only on the RL78/G13 targets. The @code{rl78} option is for the
18786 standard hardware multiplication defined in the RL78 software manual.
18790 @node RS/6000 and PowerPC Options
18791 @subsection IBM RS/6000 and PowerPC Options
18792 @cindex RS/6000 and PowerPC Options
18793 @cindex IBM RS/6000 and PowerPC Options
18795 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
18797 @item -mpowerpc-gpopt
18798 @itemx -mno-powerpc-gpopt
18799 @itemx -mpowerpc-gfxopt
18800 @itemx -mno-powerpc-gfxopt
18803 @itemx -mno-powerpc64
18807 @itemx -mno-popcntb
18809 @itemx -mno-popcntd
18818 @itemx -mno-hard-dfp
18819 @opindex mpowerpc-gpopt
18820 @opindex mno-powerpc-gpopt
18821 @opindex mpowerpc-gfxopt
18822 @opindex mno-powerpc-gfxopt
18823 @opindex mpowerpc64
18824 @opindex mno-powerpc64
18828 @opindex mno-popcntb
18830 @opindex mno-popcntd
18836 @opindex mno-mfpgpr
18838 @opindex mno-hard-dfp
18839 You use these options to specify which instructions are available on the
18840 processor you are using. The default value of these options is
18841 determined when configuring GCC@. Specifying the
18842 @option{-mcpu=@var{cpu_type}} overrides the specification of these
18843 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
18844 rather than the options listed above.
18846 Specifying @option{-mpowerpc-gpopt} allows
18847 GCC to use the optional PowerPC architecture instructions in the
18848 General Purpose group, including floating-point square root. Specifying
18849 @option{-mpowerpc-gfxopt} allows GCC to
18850 use the optional PowerPC architecture instructions in the Graphics
18851 group, including floating-point select.
18853 The @option{-mmfcrf} option allows GCC to generate the move from
18854 condition register field instruction implemented on the POWER4
18855 processor and other processors that support the PowerPC V2.01
18857 The @option{-mpopcntb} option allows GCC to generate the popcount and
18858 double-precision FP reciprocal estimate instruction implemented on the
18859 POWER5 processor and other processors that support the PowerPC V2.02
18861 The @option{-mpopcntd} option allows GCC to generate the popcount
18862 instruction implemented on the POWER7 processor and other processors
18863 that support the PowerPC V2.06 architecture.
18864 The @option{-mfprnd} option allows GCC to generate the FP round to
18865 integer instructions implemented on the POWER5+ processor and other
18866 processors that support the PowerPC V2.03 architecture.
18867 The @option{-mcmpb} option allows GCC to generate the compare bytes
18868 instruction implemented on the POWER6 processor and other processors
18869 that support the PowerPC V2.05 architecture.
18870 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
18871 general-purpose register instructions implemented on the POWER6X
18872 processor and other processors that support the extended PowerPC V2.05
18874 The @option{-mhard-dfp} option allows GCC to generate the decimal
18875 floating-point instructions implemented on some POWER processors.
18877 The @option{-mpowerpc64} option allows GCC to generate the additional
18878 64-bit instructions that are found in the full PowerPC64 architecture
18879 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
18880 @option{-mno-powerpc64}.
18882 @item -mcpu=@var{cpu_type}
18884 Set architecture type, register usage, and
18885 instruction scheduling parameters for machine type @var{cpu_type}.
18886 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
18887 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
18888 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
18889 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
18890 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
18891 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
18892 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
18893 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
18894 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
18895 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
18896 @samp{powerpc64}, and @samp{rs64}.
18898 @option{-mcpu=powerpc}, and @option{-mcpu=powerpc64} specify pure 32-bit
18899 PowerPC and 64-bit PowerPC architecture machine
18900 types, with an appropriate, generic processor model assumed for
18901 scheduling purposes.
18903 The other options specify a specific processor. Code generated under
18904 those options runs best on that processor, and may not run at all on
18907 The @option{-mcpu} options automatically enable or disable the
18910 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
18911 -mpopcntb -mpopcntd -mpowerpc64 @gol
18912 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
18913 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
18914 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
18915 -mquad-memory -mquad-memory-atomic}
18917 The particular options set for any particular CPU varies between
18918 compiler versions, depending on what setting seems to produce optimal
18919 code for that CPU; it doesn't necessarily reflect the actual hardware's
18920 capabilities. If you wish to set an individual option to a particular
18921 value, you may specify it after the @option{-mcpu} option, like
18922 @option{-mcpu=970 -mno-altivec}.
18924 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
18925 not enabled or disabled by the @option{-mcpu} option at present because
18926 AIX does not have full support for these options. You may still
18927 enable or disable them individually if you're sure it'll work in your
18930 @item -mtune=@var{cpu_type}
18932 Set the instruction scheduling parameters for machine type
18933 @var{cpu_type}, but do not set the architecture type or register usage,
18934 as @option{-mcpu=@var{cpu_type}} does. The same
18935 values for @var{cpu_type} are used for @option{-mtune} as for
18936 @option{-mcpu}. If both are specified, the code generated uses the
18937 architecture and registers set by @option{-mcpu}, but the
18938 scheduling parameters set by @option{-mtune}.
18940 @item -mcmodel=small
18941 @opindex mcmodel=small
18942 Generate PowerPC64 code for the small model: The TOC is limited to
18945 @item -mcmodel=medium
18946 @opindex mcmodel=medium
18947 Generate PowerPC64 code for the medium model: The TOC and other static
18948 data may be up to a total of 4G in size.
18950 @item -mcmodel=large
18951 @opindex mcmodel=large
18952 Generate PowerPC64 code for the large model: The TOC may be up to 4G
18953 in size. Other data and code is only limited by the 64-bit address
18957 @itemx -mno-altivec
18959 @opindex mno-altivec
18960 Generate code that uses (does not use) AltiVec instructions, and also
18961 enable the use of built-in functions that allow more direct access to
18962 the AltiVec instruction set. You may also need to set
18963 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
18966 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
18967 @option{-maltivec=be}, the element order for Altivec intrinsics such
18968 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert} will
18969 match array element order corresponding to the endianness of the
18970 target. That is, element zero identifies the leftmost element in a
18971 vector register when targeting a big-endian platform, and identifies
18972 the rightmost element in a vector register when targeting a
18973 little-endian platform.
18976 @opindex maltivec=be
18977 Generate Altivec instructions using big-endian element order,
18978 regardless of whether the target is big- or little-endian. This is
18979 the default when targeting a big-endian platform.
18981 The element order is used to interpret element numbers in Altivec
18982 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18983 @code{vec_insert}. By default, these will match array element order
18984 corresponding to the endianness for the target.
18987 @opindex maltivec=le
18988 Generate Altivec instructions using little-endian element order,
18989 regardless of whether the target is big- or little-endian. This is
18990 the default when targeting a little-endian platform. This option is
18991 currently ignored when targeting a big-endian platform.
18993 The element order is used to interpret element numbers in Altivec
18994 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18995 @code{vec_insert}. By default, these will match array element order
18996 corresponding to the endianness for the target.
19001 @opindex mno-vrsave
19002 Generate VRSAVE instructions when generating AltiVec code.
19004 @item -mgen-cell-microcode
19005 @opindex mgen-cell-microcode
19006 Generate Cell microcode instructions.
19008 @item -mwarn-cell-microcode
19009 @opindex mwarn-cell-microcode
19010 Warn when a Cell microcode instruction is emitted. An example
19011 of a Cell microcode instruction is a variable shift.
19014 @opindex msecure-plt
19015 Generate code that allows @command{ld} and @command{ld.so}
19016 to build executables and shared
19017 libraries with non-executable @code{.plt} and @code{.got} sections.
19019 32-bit SYSV ABI option.
19023 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19025 requires @code{.plt} and @code{.got}
19026 sections that are both writable and executable.
19027 This is a PowerPC 32-bit SYSV ABI option.
19033 This switch enables or disables the generation of ISEL instructions.
19035 @item -misel=@var{yes/no}
19036 This switch has been deprecated. Use @option{-misel} and
19037 @option{-mno-isel} instead.
19043 This switch enables or disables the generation of SPE simd
19049 @opindex mno-paired
19050 This switch enables or disables the generation of PAIRED simd
19053 @item -mspe=@var{yes/no}
19054 This option has been deprecated. Use @option{-mspe} and
19055 @option{-mno-spe} instead.
19061 Generate code that uses (does not use) vector/scalar (VSX)
19062 instructions, and also enable the use of built-in functions that allow
19063 more direct access to the VSX instruction set.
19068 @opindex mno-crypto
19069 Enable the use (disable) of the built-in functions that allow direct
19070 access to the cryptographic instructions that were added in version
19071 2.07 of the PowerPC ISA.
19073 @item -mdirect-move
19074 @itemx -mno-direct-move
19075 @opindex mdirect-move
19076 @opindex mno-direct-move
19077 Generate code that uses (does not use) the instructions to move data
19078 between the general purpose registers and the vector/scalar (VSX)
19079 registers that were added in version 2.07 of the PowerPC ISA.
19081 @item -mpower8-fusion
19082 @itemx -mno-power8-fusion
19083 @opindex mpower8-fusion
19084 @opindex mno-power8-fusion
19085 Generate code that keeps (does not keeps) some integer operations
19086 adjacent so that the instructions can be fused together on power8 and
19089 @item -mpower8-vector
19090 @itemx -mno-power8-vector
19091 @opindex mpower8-vector
19092 @opindex mno-power8-vector
19093 Generate code that uses (does not use) the vector and scalar
19094 instructions that were added in version 2.07 of the PowerPC ISA. Also
19095 enable the use of built-in functions that allow more direct access to
19096 the vector instructions.
19098 @item -mquad-memory
19099 @itemx -mno-quad-memory
19100 @opindex mquad-memory
19101 @opindex mno-quad-memory
19102 Generate code that uses (does not use) the non-atomic quad word memory
19103 instructions. The @option{-mquad-memory} option requires use of
19106 @item -mquad-memory-atomic
19107 @itemx -mno-quad-memory-atomic
19108 @opindex mquad-memory-atomic
19109 @opindex mno-quad-memory-atomic
19110 Generate code that uses (does not use) the atomic quad word memory
19111 instructions. The @option{-mquad-memory-atomic} option requires use of
19114 @item -mfloat-gprs=@var{yes/single/double/no}
19115 @itemx -mfloat-gprs
19116 @opindex mfloat-gprs
19117 This switch enables or disables the generation of floating-point
19118 operations on the general-purpose registers for architectures that
19121 The argument @var{yes} or @var{single} enables the use of
19122 single-precision floating-point operations.
19124 The argument @var{double} enables the use of single and
19125 double-precision floating-point operations.
19127 The argument @var{no} disables floating-point operations on the
19128 general-purpose registers.
19130 This option is currently only available on the MPC854x.
19136 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19137 targets (including GNU/Linux). The 32-bit environment sets int, long
19138 and pointer to 32 bits and generates code that runs on any PowerPC
19139 variant. The 64-bit environment sets int to 32 bits and long and
19140 pointer to 64 bits, and generates code for PowerPC64, as for
19141 @option{-mpowerpc64}.
19144 @itemx -mno-fp-in-toc
19145 @itemx -mno-sum-in-toc
19146 @itemx -mminimal-toc
19148 @opindex mno-fp-in-toc
19149 @opindex mno-sum-in-toc
19150 @opindex mminimal-toc
19151 Modify generation of the TOC (Table Of Contents), which is created for
19152 every executable file. The @option{-mfull-toc} option is selected by
19153 default. In that case, GCC allocates at least one TOC entry for
19154 each unique non-automatic variable reference in your program. GCC
19155 also places floating-point constants in the TOC@. However, only
19156 16,384 entries are available in the TOC@.
19158 If you receive a linker error message that saying you have overflowed
19159 the available TOC space, you can reduce the amount of TOC space used
19160 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19161 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19162 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19163 generate code to calculate the sum of an address and a constant at
19164 run time instead of putting that sum into the TOC@. You may specify one
19165 or both of these options. Each causes GCC to produce very slightly
19166 slower and larger code at the expense of conserving TOC space.
19168 If you still run out of space in the TOC even when you specify both of
19169 these options, specify @option{-mminimal-toc} instead. This option causes
19170 GCC to make only one TOC entry for every file. When you specify this
19171 option, GCC produces code that is slower and larger but which
19172 uses extremely little TOC space. You may wish to use this option
19173 only on files that contain less frequently-executed code.
19179 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19180 @code{long} type, and the infrastructure needed to support them.
19181 Specifying @option{-maix64} implies @option{-mpowerpc64},
19182 while @option{-maix32} disables the 64-bit ABI and
19183 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19186 @itemx -mno-xl-compat
19187 @opindex mxl-compat
19188 @opindex mno-xl-compat
19189 Produce code that conforms more closely to IBM XL compiler semantics
19190 when using AIX-compatible ABI@. Pass floating-point arguments to
19191 prototyped functions beyond the register save area (RSA) on the stack
19192 in addition to argument FPRs. Do not assume that most significant
19193 double in 128-bit long double value is properly rounded when comparing
19194 values and converting to double. Use XL symbol names for long double
19197 The AIX calling convention was extended but not initially documented to
19198 handle an obscure K&R C case of calling a function that takes the
19199 address of its arguments with fewer arguments than declared. IBM XL
19200 compilers access floating-point arguments that do not fit in the
19201 RSA from the stack when a subroutine is compiled without
19202 optimization. Because always storing floating-point arguments on the
19203 stack is inefficient and rarely needed, this option is not enabled by
19204 default and only is necessary when calling subroutines compiled by IBM
19205 XL compilers without optimization.
19209 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19210 application written to use message passing with special startup code to
19211 enable the application to run. The system must have PE installed in the
19212 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19213 must be overridden with the @option{-specs=} option to specify the
19214 appropriate directory location. The Parallel Environment does not
19215 support threads, so the @option{-mpe} option and the @option{-pthread}
19216 option are incompatible.
19218 @item -malign-natural
19219 @itemx -malign-power
19220 @opindex malign-natural
19221 @opindex malign-power
19222 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19223 @option{-malign-natural} overrides the ABI-defined alignment of larger
19224 types, such as floating-point doubles, on their natural size-based boundary.
19225 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19226 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19228 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19232 @itemx -mhard-float
19233 @opindex msoft-float
19234 @opindex mhard-float
19235 Generate code that does not use (uses) the floating-point register set.
19236 Software floating-point emulation is provided if you use the
19237 @option{-msoft-float} option, and pass the option to GCC when linking.
19239 @item -msingle-float
19240 @itemx -mdouble-float
19241 @opindex msingle-float
19242 @opindex mdouble-float
19243 Generate code for single- or double-precision floating-point operations.
19244 @option{-mdouble-float} implies @option{-msingle-float}.
19247 @opindex msimple-fpu
19248 Do not generate @code{sqrt} and @code{div} instructions for hardware
19249 floating-point unit.
19251 @item -mfpu=@var{name}
19253 Specify type of floating-point unit. Valid values for @var{name} are
19254 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19255 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19256 @samp{sp_full} (equivalent to @option{-msingle-float}),
19257 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19260 @opindex mxilinx-fpu
19261 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19264 @itemx -mno-multiple
19266 @opindex mno-multiple
19267 Generate code that uses (does not use) the load multiple word
19268 instructions and the store multiple word instructions. These
19269 instructions are generated by default on POWER systems, and not
19270 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19271 PowerPC systems, since those instructions do not work when the
19272 processor is in little-endian mode. The exceptions are PPC740 and
19273 PPC750 which permit these instructions in little-endian mode.
19278 @opindex mno-string
19279 Generate code that uses (does not use) the load string instructions
19280 and the store string word instructions to save multiple registers and
19281 do small block moves. These instructions are generated by default on
19282 POWER systems, and not generated on PowerPC systems. Do not use
19283 @option{-mstring} on little-endian PowerPC systems, since those
19284 instructions do not work when the processor is in little-endian mode.
19285 The exceptions are PPC740 and PPC750 which permit these instructions
19286 in little-endian mode.
19291 @opindex mno-update
19292 Generate code that uses (does not use) the load or store instructions
19293 that update the base register to the address of the calculated memory
19294 location. These instructions are generated by default. If you use
19295 @option{-mno-update}, there is a small window between the time that the
19296 stack pointer is updated and the address of the previous frame is
19297 stored, which means code that walks the stack frame across interrupts or
19298 signals may get corrupted data.
19300 @item -mavoid-indexed-addresses
19301 @itemx -mno-avoid-indexed-addresses
19302 @opindex mavoid-indexed-addresses
19303 @opindex mno-avoid-indexed-addresses
19304 Generate code that tries to avoid (not avoid) the use of indexed load
19305 or store instructions. These instructions can incur a performance
19306 penalty on Power6 processors in certain situations, such as when
19307 stepping through large arrays that cross a 16M boundary. This option
19308 is enabled by default when targeting Power6 and disabled otherwise.
19311 @itemx -mno-fused-madd
19312 @opindex mfused-madd
19313 @opindex mno-fused-madd
19314 Generate code that uses (does not use) the floating-point multiply and
19315 accumulate instructions. These instructions are generated by default
19316 if hardware floating point is used. The machine-dependent
19317 @option{-mfused-madd} option is now mapped to the machine-independent
19318 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19319 mapped to @option{-ffp-contract=off}.
19325 Generate code that uses (does not use) the half-word multiply and
19326 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19327 These instructions are generated by default when targeting those
19334 Generate code that uses (does not use) the string-search @samp{dlmzb}
19335 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19336 generated by default when targeting those processors.
19338 @item -mno-bit-align
19340 @opindex mno-bit-align
19341 @opindex mbit-align
19342 On System V.4 and embedded PowerPC systems do not (do) force structures
19343 and unions that contain bit-fields to be aligned to the base type of the
19346 For example, by default a structure containing nothing but 8
19347 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19348 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19349 the structure is aligned to a 1-byte boundary and is 1 byte in
19352 @item -mno-strict-align
19353 @itemx -mstrict-align
19354 @opindex mno-strict-align
19355 @opindex mstrict-align
19356 On System V.4 and embedded PowerPC systems do not (do) assume that
19357 unaligned memory references are handled by the system.
19359 @item -mrelocatable
19360 @itemx -mno-relocatable
19361 @opindex mrelocatable
19362 @opindex mno-relocatable
19363 Generate code that allows (does not allow) a static executable to be
19364 relocated to a different address at run time. A simple embedded
19365 PowerPC system loader should relocate the entire contents of
19366 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19367 a table of 32-bit addresses generated by this option. For this to
19368 work, all objects linked together must be compiled with
19369 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19370 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19372 @item -mrelocatable-lib
19373 @itemx -mno-relocatable-lib
19374 @opindex mrelocatable-lib
19375 @opindex mno-relocatable-lib
19376 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19377 @code{.fixup} section to allow static executables to be relocated at
19378 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19379 alignment of @option{-mrelocatable}. Objects compiled with
19380 @option{-mrelocatable-lib} may be linked with objects compiled with
19381 any combination of the @option{-mrelocatable} options.
19387 On System V.4 and embedded PowerPC systems do not (do) assume that
19388 register 2 contains a pointer to a global area pointing to the addresses
19389 used in the program.
19392 @itemx -mlittle-endian
19394 @opindex mlittle-endian
19395 On System V.4 and embedded PowerPC systems compile code for the
19396 processor in little-endian mode. The @option{-mlittle-endian} option is
19397 the same as @option{-mlittle}.
19400 @itemx -mbig-endian
19402 @opindex mbig-endian
19403 On System V.4 and embedded PowerPC systems compile code for the
19404 processor in big-endian mode. The @option{-mbig-endian} option is
19405 the same as @option{-mbig}.
19407 @item -mdynamic-no-pic
19408 @opindex mdynamic-no-pic
19409 On Darwin and Mac OS X systems, compile code so that it is not
19410 relocatable, but that its external references are relocatable. The
19411 resulting code is suitable for applications, but not shared
19414 @item -msingle-pic-base
19415 @opindex msingle-pic-base
19416 Treat the register used for PIC addressing as read-only, rather than
19417 loading it in the prologue for each function. The runtime system is
19418 responsible for initializing this register with an appropriate value
19419 before execution begins.
19421 @item -mprioritize-restricted-insns=@var{priority}
19422 @opindex mprioritize-restricted-insns
19423 This option controls the priority that is assigned to
19424 dispatch-slot restricted instructions during the second scheduling
19425 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19426 or @samp{2} to assign no, highest, or second-highest (respectively)
19427 priority to dispatch-slot restricted
19430 @item -msched-costly-dep=@var{dependence_type}
19431 @opindex msched-costly-dep
19432 This option controls which dependences are considered costly
19433 by the target during instruction scheduling. The argument
19434 @var{dependence_type} takes one of the following values:
19438 No dependence is costly.
19441 All dependences are costly.
19443 @item @samp{true_store_to_load}
19444 A true dependence from store to load is costly.
19446 @item @samp{store_to_load}
19447 Any dependence from store to load is costly.
19450 Any dependence for which the latency is greater than or equal to
19451 @var{number} is costly.
19454 @item -minsert-sched-nops=@var{scheme}
19455 @opindex minsert-sched-nops
19456 This option controls which NOP insertion scheme is used during
19457 the second scheduling pass. The argument @var{scheme} takes one of the
19465 Pad with NOPs any dispatch group that has vacant issue slots,
19466 according to the scheduler's grouping.
19468 @item @samp{regroup_exact}
19469 Insert NOPs to force costly dependent insns into
19470 separate groups. Insert exactly as many NOPs as needed to force an insn
19471 to a new group, according to the estimated processor grouping.
19474 Insert NOPs to force costly dependent insns into
19475 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19479 @opindex mcall-sysv
19480 On System V.4 and embedded PowerPC systems compile code using calling
19481 conventions that adhere to the March 1995 draft of the System V
19482 Application Binary Interface, PowerPC processor supplement. This is the
19483 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19485 @item -mcall-sysv-eabi
19487 @opindex mcall-sysv-eabi
19488 @opindex mcall-eabi
19489 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19491 @item -mcall-sysv-noeabi
19492 @opindex mcall-sysv-noeabi
19493 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19495 @item -mcall-aixdesc
19497 On System V.4 and embedded PowerPC systems compile code for the AIX
19501 @opindex mcall-linux
19502 On System V.4 and embedded PowerPC systems compile code for the
19503 Linux-based GNU system.
19505 @item -mcall-freebsd
19506 @opindex mcall-freebsd
19507 On System V.4 and embedded PowerPC systems compile code for the
19508 FreeBSD operating system.
19510 @item -mcall-netbsd
19511 @opindex mcall-netbsd
19512 On System V.4 and embedded PowerPC systems compile code for the
19513 NetBSD operating system.
19515 @item -mcall-openbsd
19516 @opindex mcall-netbsd
19517 On System V.4 and embedded PowerPC systems compile code for the
19518 OpenBSD operating system.
19520 @item -maix-struct-return
19521 @opindex maix-struct-return
19522 Return all structures in memory (as specified by the AIX ABI)@.
19524 @item -msvr4-struct-return
19525 @opindex msvr4-struct-return
19526 Return structures smaller than 8 bytes in registers (as specified by the
19529 @item -mabi=@var{abi-type}
19531 Extend the current ABI with a particular extension, or remove such extension.
19532 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
19533 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble},
19534 @var{elfv1}, @var{elfv2}@.
19538 Extend the current ABI with SPE ABI extensions. This does not change
19539 the default ABI, instead it adds the SPE ABI extensions to the current
19543 @opindex mabi=no-spe
19544 Disable Book-E SPE ABI extensions for the current ABI@.
19546 @item -mabi=ibmlongdouble
19547 @opindex mabi=ibmlongdouble
19548 Change the current ABI to use IBM extended-precision long double.
19549 This is a PowerPC 32-bit SYSV ABI option.
19551 @item -mabi=ieeelongdouble
19552 @opindex mabi=ieeelongdouble
19553 Change the current ABI to use IEEE extended-precision long double.
19554 This is a PowerPC 32-bit Linux ABI option.
19557 @opindex mabi=elfv1
19558 Change the current ABI to use the ELFv1 ABI.
19559 This is the default ABI for big-endian PowerPC 64-bit Linux.
19560 Overriding the default ABI requires special system support and is
19561 likely to fail in spectacular ways.
19564 @opindex mabi=elfv2
19565 Change the current ABI to use the ELFv2 ABI.
19566 This is the default ABI for little-endian PowerPC 64-bit Linux.
19567 Overriding the default ABI requires special system support and is
19568 likely to fail in spectacular ways.
19571 @itemx -mno-prototype
19572 @opindex mprototype
19573 @opindex mno-prototype
19574 On System V.4 and embedded PowerPC systems assume that all calls to
19575 variable argument functions are properly prototyped. Otherwise, the
19576 compiler must insert an instruction before every non-prototyped call to
19577 set or clear bit 6 of the condition code register (@var{CR}) to
19578 indicate whether floating-point values are passed in the floating-point
19579 registers in case the function takes variable arguments. With
19580 @option{-mprototype}, only calls to prototyped variable argument functions
19581 set or clear the bit.
19585 On embedded PowerPC systems, assume that the startup module is called
19586 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19587 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19592 On embedded PowerPC systems, assume that the startup module is called
19593 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19598 On embedded PowerPC systems, assume that the startup module is called
19599 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19602 @item -myellowknife
19603 @opindex myellowknife
19604 On embedded PowerPC systems, assume that the startup module is called
19605 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19610 On System V.4 and embedded PowerPC systems, specify that you are
19611 compiling for a VxWorks system.
19615 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
19616 header to indicate that @samp{eabi} extended relocations are used.
19622 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19623 Embedded Applications Binary Interface (EABI), which is a set of
19624 modifications to the System V.4 specifications. Selecting @option{-meabi}
19625 means that the stack is aligned to an 8-byte boundary, a function
19626 @code{__eabi} is called from @code{main} to set up the EABI
19627 environment, and the @option{-msdata} option can use both @code{r2} and
19628 @code{r13} to point to two separate small data areas. Selecting
19629 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19630 no EABI initialization function is called from @code{main}, and the
19631 @option{-msdata} option only uses @code{r13} to point to a single
19632 small data area. The @option{-meabi} option is on by default if you
19633 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19636 @opindex msdata=eabi
19637 On System V.4 and embedded PowerPC systems, put small initialized
19638 @code{const} global and static data in the @samp{.sdata2} section, which
19639 is pointed to by register @code{r2}. Put small initialized
19640 non-@code{const} global and static data in the @samp{.sdata} section,
19641 which is pointed to by register @code{r13}. Put small uninitialized
19642 global and static data in the @samp{.sbss} section, which is adjacent to
19643 the @samp{.sdata} section. The @option{-msdata=eabi} option is
19644 incompatible with the @option{-mrelocatable} option. The
19645 @option{-msdata=eabi} option also sets the @option{-memb} option.
19648 @opindex msdata=sysv
19649 On System V.4 and embedded PowerPC systems, put small global and static
19650 data in the @samp{.sdata} section, which is pointed to by register
19651 @code{r13}. Put small uninitialized global and static data in the
19652 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
19653 The @option{-msdata=sysv} option is incompatible with the
19654 @option{-mrelocatable} option.
19656 @item -msdata=default
19658 @opindex msdata=default
19660 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19661 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19662 same as @option{-msdata=sysv}.
19665 @opindex msdata=data
19666 On System V.4 and embedded PowerPC systems, put small global
19667 data in the @samp{.sdata} section. Put small uninitialized global
19668 data in the @samp{.sbss} section. Do not use register @code{r13}
19669 to address small data however. This is the default behavior unless
19670 other @option{-msdata} options are used.
19674 @opindex msdata=none
19676 On embedded PowerPC systems, put all initialized global and static data
19677 in the @samp{.data} section, and all uninitialized data in the
19678 @samp{.bss} section.
19680 @item -mblock-move-inline-limit=@var{num}
19681 @opindex mblock-move-inline-limit
19682 Inline all block moves (such as calls to @code{memcpy} or structure
19683 copies) less than or equal to @var{num} bytes. The minimum value for
19684 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19685 targets. The default value is target-specific.
19689 @cindex smaller data references (PowerPC)
19690 @cindex .sdata/.sdata2 references (PowerPC)
19691 On embedded PowerPC systems, put global and static items less than or
19692 equal to @var{num} bytes into the small data or BSS sections instead of
19693 the normal data or BSS section. By default, @var{num} is 8. The
19694 @option{-G @var{num}} switch is also passed to the linker.
19695 All modules should be compiled with the same @option{-G @var{num}} value.
19698 @itemx -mno-regnames
19700 @opindex mno-regnames
19701 On System V.4 and embedded PowerPC systems do (do not) emit register
19702 names in the assembly language output using symbolic forms.
19705 @itemx -mno-longcall
19707 @opindex mno-longcall
19708 By default assume that all calls are far away so that a longer and more
19709 expensive calling sequence is required. This is required for calls
19710 farther than 32 megabytes (33,554,432 bytes) from the current location.
19711 A short call is generated if the compiler knows
19712 the call cannot be that far away. This setting can be overridden by
19713 the @code{shortcall} function attribute, or by @code{#pragma
19716 Some linkers are capable of detecting out-of-range calls and generating
19717 glue code on the fly. On these systems, long calls are unnecessary and
19718 generate slower code. As of this writing, the AIX linker can do this,
19719 as can the GNU linker for PowerPC/64. It is planned to add this feature
19720 to the GNU linker for 32-bit PowerPC systems as well.
19722 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
19723 callee, L42}, plus a @dfn{branch island} (glue code). The two target
19724 addresses represent the callee and the branch island. The
19725 Darwin/PPC linker prefers the first address and generates a @code{bl
19726 callee} if the PPC @code{bl} instruction reaches the callee directly;
19727 otherwise, the linker generates @code{bl L42} to call the branch
19728 island. The branch island is appended to the body of the
19729 calling function; it computes the full 32-bit address of the callee
19732 On Mach-O (Darwin) systems, this option directs the compiler emit to
19733 the glue for every direct call, and the Darwin linker decides whether
19734 to use or discard it.
19736 In the future, GCC may ignore all longcall specifications
19737 when the linker is known to generate glue.
19739 @item -mtls-markers
19740 @itemx -mno-tls-markers
19741 @opindex mtls-markers
19742 @opindex mno-tls-markers
19743 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
19744 specifying the function argument. The relocation allows the linker to
19745 reliably associate function call with argument setup instructions for
19746 TLS optimization, which in turn allows GCC to better schedule the
19751 Adds support for multithreading with the @dfn{pthreads} library.
19752 This option sets flags for both the preprocessor and linker.
19757 This option enables use of the reciprocal estimate and
19758 reciprocal square root estimate instructions with additional
19759 Newton-Raphson steps to increase precision instead of doing a divide or
19760 square root and divide for floating-point arguments. You should use
19761 the @option{-ffast-math} option when using @option{-mrecip} (or at
19762 least @option{-funsafe-math-optimizations},
19763 @option{-finite-math-only}, @option{-freciprocal-math} and
19764 @option{-fno-trapping-math}). Note that while the throughput of the
19765 sequence is generally higher than the throughput of the non-reciprocal
19766 instruction, the precision of the sequence can be decreased by up to 2
19767 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
19770 @item -mrecip=@var{opt}
19771 @opindex mrecip=opt
19772 This option controls which reciprocal estimate instructions
19773 may be used. @var{opt} is a comma-separated list of options, which may
19774 be preceded by a @code{!} to invert the option:
19775 @code{all}: enable all estimate instructions,
19776 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
19777 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
19778 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
19779 @code{divf}: enable the single-precision reciprocal approximation instructions;
19780 @code{divd}: enable the double-precision reciprocal approximation instructions;
19781 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
19782 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
19783 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
19785 So, for example, @option{-mrecip=all,!rsqrtd} enables
19786 all of the reciprocal estimate instructions, except for the
19787 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
19788 which handle the double-precision reciprocal square root calculations.
19790 @item -mrecip-precision
19791 @itemx -mno-recip-precision
19792 @opindex mrecip-precision
19793 Assume (do not assume) that the reciprocal estimate instructions
19794 provide higher-precision estimates than is mandated by the PowerPC
19795 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
19796 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
19797 The double-precision square root estimate instructions are not generated by
19798 default on low-precision machines, since they do not provide an
19799 estimate that converges after three steps.
19801 @item -mveclibabi=@var{type}
19802 @opindex mveclibabi
19803 Specifies the ABI type to use for vectorizing intrinsics using an
19804 external library. The only type supported at present is @code{mass},
19805 which specifies to use IBM's Mathematical Acceleration Subsystem
19806 (MASS) libraries for vectorizing intrinsics using external libraries.
19807 GCC currently emits calls to @code{acosd2}, @code{acosf4},
19808 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
19809 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
19810 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
19811 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
19812 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
19813 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
19814 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
19815 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
19816 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
19817 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
19818 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
19819 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
19820 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
19821 for power7. Both @option{-ftree-vectorize} and
19822 @option{-funsafe-math-optimizations} must also be enabled. The MASS
19823 libraries must be specified at link time.
19828 Generate (do not generate) the @code{friz} instruction when the
19829 @option{-funsafe-math-optimizations} option is used to optimize
19830 rounding of floating-point values to 64-bit integer and back to floating
19831 point. The @code{friz} instruction does not return the same value if
19832 the floating-point number is too large to fit in an integer.
19834 @item -mpointers-to-nested-functions
19835 @itemx -mno-pointers-to-nested-functions
19836 @opindex mpointers-to-nested-functions
19837 Generate (do not generate) code to load up the static chain register
19838 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
19839 systems where a function pointer points to a 3-word descriptor giving
19840 the function address, TOC value to be loaded in register @var{r2}, and
19841 static chain value to be loaded in register @var{r11}. The
19842 @option{-mpointers-to-nested-functions} is on by default. You cannot
19843 call through pointers to nested functions or pointers
19844 to functions compiled in other languages that use the static chain if
19845 you use the @option{-mno-pointers-to-nested-functions}.
19847 @item -msave-toc-indirect
19848 @itemx -mno-save-toc-indirect
19849 @opindex msave-toc-indirect
19850 Generate (do not generate) code to save the TOC value in the reserved
19851 stack location in the function prologue if the function calls through
19852 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
19853 saved in the prologue, it is saved just before the call through the
19854 pointer. The @option{-mno-save-toc-indirect} option is the default.
19856 @item -mcompat-align-parm
19857 @itemx -mno-compat-align-parm
19858 @opindex mcompat-align-parm
19859 Generate (do not generate) code to pass structure parameters with a
19860 maximum alignment of 64 bits, for compatibility with older versions
19863 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
19864 structure parameter on a 128-bit boundary when that structure contained
19865 a member requiring 128-bit alignment. This is corrected in more
19866 recent versions of GCC. This option may be used to generate code
19867 that is compatible with functions compiled with older versions of
19870 The @option{-mno-compat-align-parm} option is the default.
19874 @subsection RX Options
19877 These command-line options are defined for RX targets:
19880 @item -m64bit-doubles
19881 @itemx -m32bit-doubles
19882 @opindex m64bit-doubles
19883 @opindex m32bit-doubles
19884 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19885 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19886 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
19887 works on 32-bit values, which is why the default is
19888 @option{-m32bit-doubles}.
19894 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
19895 floating-point hardware. The default is enabled for the @var{RX600}
19896 series and disabled for the @var{RX200} series.
19898 Floating-point instructions are only generated for 32-bit floating-point
19899 values, however, so the FPU hardware is not used for doubles if the
19900 @option{-m64bit-doubles} option is used.
19902 @emph{Note} If the @option{-fpu} option is enabled then
19903 @option{-funsafe-math-optimizations} is also enabled automatically.
19904 This is because the RX FPU instructions are themselves unsafe.
19906 @item -mcpu=@var{name}
19908 Selects the type of RX CPU to be targeted. Currently three types are
19909 supported, the generic @var{RX600} and @var{RX200} series hardware and
19910 the specific @var{RX610} CPU. The default is @var{RX600}.
19912 The only difference between @var{RX600} and @var{RX610} is that the
19913 @var{RX610} does not support the @code{MVTIPL} instruction.
19915 The @var{RX200} series does not have a hardware floating-point unit
19916 and so @option{-nofpu} is enabled by default when this type is
19919 @item -mbig-endian-data
19920 @itemx -mlittle-endian-data
19921 @opindex mbig-endian-data
19922 @opindex mlittle-endian-data
19923 Store data (but not code) in the big-endian format. The default is
19924 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
19927 @item -msmall-data-limit=@var{N}
19928 @opindex msmall-data-limit
19929 Specifies the maximum size in bytes of global and static variables
19930 which can be placed into the small data area. Using the small data
19931 area can lead to smaller and faster code, but the size of area is
19932 limited and it is up to the programmer to ensure that the area does
19933 not overflow. Also when the small data area is used one of the RX's
19934 registers (usually @code{r13}) is reserved for use pointing to this
19935 area, so it is no longer available for use by the compiler. This
19936 could result in slower and/or larger code if variables are pushed onto
19937 the stack instead of being held in this register.
19939 Note, common variables (variables that have not been initialized) and
19940 constants are not placed into the small data area as they are assigned
19941 to other sections in the output executable.
19943 The default value is zero, which disables this feature. Note, this
19944 feature is not enabled by default with higher optimization levels
19945 (@option{-O2} etc) because of the potentially detrimental effects of
19946 reserving a register. It is up to the programmer to experiment and
19947 discover whether this feature is of benefit to their program. See the
19948 description of the @option{-mpid} option for a description of how the
19949 actual register to hold the small data area pointer is chosen.
19955 Use the simulator runtime. The default is to use the libgloss
19956 board-specific runtime.
19958 @item -mas100-syntax
19959 @itemx -mno-as100-syntax
19960 @opindex mas100-syntax
19961 @opindex mno-as100-syntax
19962 When generating assembler output use a syntax that is compatible with
19963 Renesas's AS100 assembler. This syntax can also be handled by the GAS
19964 assembler, but it has some restrictions so it is not generated by default.
19966 @item -mmax-constant-size=@var{N}
19967 @opindex mmax-constant-size
19968 Specifies the maximum size, in bytes, of a constant that can be used as
19969 an operand in a RX instruction. Although the RX instruction set does
19970 allow constants of up to 4 bytes in length to be used in instructions,
19971 a longer value equates to a longer instruction. Thus in some
19972 circumstances it can be beneficial to restrict the size of constants
19973 that are used in instructions. Constants that are too big are instead
19974 placed into a constant pool and referenced via register indirection.
19976 The value @var{N} can be between 0 and 4. A value of 0 (the default)
19977 or 4 means that constants of any size are allowed.
19981 Enable linker relaxation. Linker relaxation is a process whereby the
19982 linker attempts to reduce the size of a program by finding shorter
19983 versions of various instructions. Disabled by default.
19985 @item -mint-register=@var{N}
19986 @opindex mint-register
19987 Specify the number of registers to reserve for fast interrupt handler
19988 functions. The value @var{N} can be between 0 and 4. A value of 1
19989 means that register @code{r13} is reserved for the exclusive use
19990 of fast interrupt handlers. A value of 2 reserves @code{r13} and
19991 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
19992 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
19993 A value of 0, the default, does not reserve any registers.
19995 @item -msave-acc-in-interrupts
19996 @opindex msave-acc-in-interrupts
19997 Specifies that interrupt handler functions should preserve the
19998 accumulator register. This is only necessary if normal code might use
19999 the accumulator register, for example because it performs 64-bit
20000 multiplications. The default is to ignore the accumulator as this
20001 makes the interrupt handlers faster.
20007 Enables the generation of position independent data. When enabled any
20008 access to constant data is done via an offset from a base address
20009 held in a register. This allows the location of constant data to be
20010 determined at run time without requiring the executable to be
20011 relocated, which is a benefit to embedded applications with tight
20012 memory constraints. Data that can be modified is not affected by this
20015 Note, using this feature reserves a register, usually @code{r13}, for
20016 the constant data base address. This can result in slower and/or
20017 larger code, especially in complicated functions.
20019 The actual register chosen to hold the constant data base address
20020 depends upon whether the @option{-msmall-data-limit} and/or the
20021 @option{-mint-register} command-line options are enabled. Starting
20022 with register @code{r13} and proceeding downwards, registers are
20023 allocated first to satisfy the requirements of @option{-mint-register},
20024 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20025 is possible for the small data area register to be @code{r8} if both
20026 @option{-mint-register=4} and @option{-mpid} are specified on the
20029 By default this feature is not enabled. The default can be restored
20030 via the @option{-mno-pid} command-line option.
20032 @item -mno-warn-multiple-fast-interrupts
20033 @itemx -mwarn-multiple-fast-interrupts
20034 @opindex mno-warn-multiple-fast-interrupts
20035 @opindex mwarn-multiple-fast-interrupts
20036 Prevents GCC from issuing a warning message if it finds more than one
20037 fast interrupt handler when it is compiling a file. The default is to
20038 issue a warning for each extra fast interrupt handler found, as the RX
20039 only supports one such interrupt.
20043 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20044 has special significance to the RX port when used with the
20045 @code{interrupt} function attribute. This attribute indicates a
20046 function intended to process fast interrupts. GCC ensures
20047 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20048 and/or @code{r13} and only provided that the normal use of the
20049 corresponding registers have been restricted via the
20050 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20053 @node S/390 and zSeries Options
20054 @subsection S/390 and zSeries Options
20055 @cindex S/390 and zSeries Options
20057 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20061 @itemx -msoft-float
20062 @opindex mhard-float
20063 @opindex msoft-float
20064 Use (do not use) the hardware floating-point instructions and registers
20065 for floating-point operations. When @option{-msoft-float} is specified,
20066 functions in @file{libgcc.a} are used to perform floating-point
20067 operations. When @option{-mhard-float} is specified, the compiler
20068 generates IEEE floating-point instructions. This is the default.
20071 @itemx -mno-hard-dfp
20073 @opindex mno-hard-dfp
20074 Use (do not use) the hardware decimal-floating-point instructions for
20075 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20076 specified, functions in @file{libgcc.a} are used to perform
20077 decimal-floating-point operations. When @option{-mhard-dfp} is
20078 specified, the compiler generates decimal-floating-point hardware
20079 instructions. This is the default for @option{-march=z9-ec} or higher.
20081 @item -mlong-double-64
20082 @itemx -mlong-double-128
20083 @opindex mlong-double-64
20084 @opindex mlong-double-128
20085 These switches control the size of @code{long double} type. A size
20086 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20087 type. This is the default.
20090 @itemx -mno-backchain
20091 @opindex mbackchain
20092 @opindex mno-backchain
20093 Store (do not store) the address of the caller's frame as backchain pointer
20094 into the callee's stack frame.
20095 A backchain may be needed to allow debugging using tools that do not understand
20096 DWARF 2 call frame information.
20097 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20098 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20099 the backchain is placed into the topmost word of the 96/160 byte register
20102 In general, code compiled with @option{-mbackchain} is call-compatible with
20103 code compiled with @option{-mmo-backchain}; however, use of the backchain
20104 for debugging purposes usually requires that the whole binary is built with
20105 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20106 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20107 to build a linux kernel use @option{-msoft-float}.
20109 The default is to not maintain the backchain.
20111 @item -mpacked-stack
20112 @itemx -mno-packed-stack
20113 @opindex mpacked-stack
20114 @opindex mno-packed-stack
20115 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20116 specified, the compiler uses the all fields of the 96/160 byte register save
20117 area only for their default purpose; unused fields still take up stack space.
20118 When @option{-mpacked-stack} is specified, register save slots are densely
20119 packed at the top of the register save area; unused space is reused for other
20120 purposes, allowing for more efficient use of the available stack space.
20121 However, when @option{-mbackchain} is also in effect, the topmost word of
20122 the save area is always used to store the backchain, and the return address
20123 register is always saved two words below the backchain.
20125 As long as the stack frame backchain is not used, code generated with
20126 @option{-mpacked-stack} is call-compatible with code generated with
20127 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20128 S/390 or zSeries generated code that uses the stack frame backchain at run
20129 time, not just for debugging purposes. Such code is not call-compatible
20130 with code compiled with @option{-mpacked-stack}. Also, note that the
20131 combination of @option{-mbackchain},
20132 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20133 to build a linux kernel use @option{-msoft-float}.
20135 The default is to not use the packed stack layout.
20138 @itemx -mno-small-exec
20139 @opindex msmall-exec
20140 @opindex mno-small-exec
20141 Generate (or do not generate) code using the @code{bras} instruction
20142 to do subroutine calls.
20143 This only works reliably if the total executable size does not
20144 exceed 64k. The default is to use the @code{basr} instruction instead,
20145 which does not have this limitation.
20151 When @option{-m31} is specified, generate code compliant to the
20152 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20153 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20154 particular to generate 64-bit instructions. For the @samp{s390}
20155 targets, the default is @option{-m31}, while the @samp{s390x}
20156 targets default to @option{-m64}.
20162 When @option{-mzarch} is specified, generate code using the
20163 instructions available on z/Architecture.
20164 When @option{-mesa} is specified, generate code using the
20165 instructions available on ESA/390. Note that @option{-mesa} is
20166 not possible with @option{-m64}.
20167 When generating code compliant to the GNU/Linux for S/390 ABI,
20168 the default is @option{-mesa}. When generating code compliant
20169 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20175 Generate (or do not generate) code using the @code{mvcle} instruction
20176 to perform block moves. When @option{-mno-mvcle} is specified,
20177 use a @code{mvc} loop instead. This is the default unless optimizing for
20184 Print (or do not print) additional debug information when compiling.
20185 The default is to not print debug information.
20187 @item -march=@var{cpu-type}
20189 Generate code that runs on @var{cpu-type}, which is the name of a system
20190 representing a certain processor type. Possible values for
20191 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20192 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
20193 When generating code using the instructions available on z/Architecture,
20194 the default is @option{-march=z900}. Otherwise, the default is
20195 @option{-march=g5}.
20197 @item -mtune=@var{cpu-type}
20199 Tune to @var{cpu-type} everything applicable about the generated code,
20200 except for the ABI and the set of available instructions.
20201 The list of @var{cpu-type} values is the same as for @option{-march}.
20202 The default is the value used for @option{-march}.
20205 @itemx -mno-tpf-trace
20206 @opindex mtpf-trace
20207 @opindex mno-tpf-trace
20208 Generate code that adds (does not add) in TPF OS specific branches to trace
20209 routines in the operating system. This option is off by default, even
20210 when compiling for the TPF OS@.
20213 @itemx -mno-fused-madd
20214 @opindex mfused-madd
20215 @opindex mno-fused-madd
20216 Generate code that uses (does not use) the floating-point multiply and
20217 accumulate instructions. These instructions are generated by default if
20218 hardware floating point is used.
20220 @item -mwarn-framesize=@var{framesize}
20221 @opindex mwarn-framesize
20222 Emit a warning if the current function exceeds the given frame size. Because
20223 this is a compile-time check it doesn't need to be a real problem when the program
20224 runs. It is intended to identify functions that most probably cause
20225 a stack overflow. It is useful to be used in an environment with limited stack
20226 size e.g.@: the linux kernel.
20228 @item -mwarn-dynamicstack
20229 @opindex mwarn-dynamicstack
20230 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20231 arrays. This is generally a bad idea with a limited stack size.
20233 @item -mstack-guard=@var{stack-guard}
20234 @itemx -mstack-size=@var{stack-size}
20235 @opindex mstack-guard
20236 @opindex mstack-size
20237 If these options are provided the S/390 back end emits additional instructions in
20238 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20239 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20240 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20241 the frame size of the compiled function is chosen.
20242 These options are intended to be used to help debugging stack overflow problems.
20243 The additionally emitted code causes only little overhead and hence can also be
20244 used in production-like systems without greater performance degradation. The given
20245 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20246 @var{stack-guard} without exceeding 64k.
20247 In order to be efficient the extra code makes the assumption that the stack starts
20248 at an address aligned to the value given by @var{stack-size}.
20249 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20251 @item -mhotpatch[=@var{halfwords}]
20252 @itemx -mno-hotpatch
20254 If the hotpatch option is enabled, a ``hot-patching'' function
20255 prologue is generated for all functions in the compilation unit.
20256 The funtion label is prepended with the given number of two-byte
20257 Nop instructions (@var{halfwords}, maximum 1000000) or 12 Nop
20258 instructions if no argument is present. Functions with a
20259 hot-patching prologue are never inlined automatically, and a
20260 hot-patching prologue is never generated for functions functions
20261 that are explicitly inline.
20263 This option can be overridden for individual functions with the
20264 @code{hotpatch} attribute.
20267 @node Score Options
20268 @subsection Score Options
20269 @cindex Score Options
20271 These options are defined for Score implementations:
20276 Compile code for big-endian mode. This is the default.
20280 Compile code for little-endian mode.
20284 Disable generation of @code{bcnz} instructions.
20288 Enable generation of unaligned load and store instructions.
20292 Enable the use of multiply-accumulate instructions. Disabled by default.
20296 Specify the SCORE5 as the target architecture.
20300 Specify the SCORE5U of the target architecture.
20304 Specify the SCORE7 as the target architecture. This is the default.
20308 Specify the SCORE7D as the target architecture.
20312 @subsection SH Options
20314 These @samp{-m} options are defined for the SH implementations:
20319 Generate code for the SH1.
20323 Generate code for the SH2.
20326 Generate code for the SH2e.
20330 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20331 that the floating-point unit is not used.
20333 @item -m2a-single-only
20334 @opindex m2a-single-only
20335 Generate code for the SH2a-FPU, in such a way that no double-precision
20336 floating-point operations are used.
20339 @opindex m2a-single
20340 Generate code for the SH2a-FPU assuming the floating-point unit is in
20341 single-precision mode by default.
20345 Generate code for the SH2a-FPU assuming the floating-point unit is in
20346 double-precision mode by default.
20350 Generate code for the SH3.
20354 Generate code for the SH3e.
20358 Generate code for the SH4 without a floating-point unit.
20360 @item -m4-single-only
20361 @opindex m4-single-only
20362 Generate code for the SH4 with a floating-point unit that only
20363 supports single-precision arithmetic.
20367 Generate code for the SH4 assuming the floating-point unit is in
20368 single-precision mode by default.
20372 Generate code for the SH4.
20376 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20377 floating-point unit is not used.
20379 @item -m4a-single-only
20380 @opindex m4a-single-only
20381 Generate code for the SH4a, in such a way that no double-precision
20382 floating-point operations are used.
20385 @opindex m4a-single
20386 Generate code for the SH4a assuming the floating-point unit is in
20387 single-precision mode by default.
20391 Generate code for the SH4a.
20395 Same as @option{-m4a-nofpu}, except that it implicitly passes
20396 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20397 instructions at the moment.
20401 Compile code for the processor in big-endian mode.
20405 Compile code for the processor in little-endian mode.
20409 Align doubles at 64-bit boundaries. Note that this changes the calling
20410 conventions, and thus some functions from the standard C library do
20411 not work unless you recompile it first with @option{-mdalign}.
20415 Shorten some address references at link time, when possible; uses the
20416 linker option @option{-relax}.
20420 Use 32-bit offsets in @code{switch} tables. The default is to use
20425 Enable the use of bit manipulation instructions on SH2A.
20429 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20430 alignment constraints.
20434 Comply with the calling conventions defined by Renesas.
20438 Comply with the calling conventions defined by Renesas.
20442 Comply with the calling conventions defined for GCC before the Renesas
20443 conventions were available. This option is the default for all
20444 targets of the SH toolchain.
20447 @opindex mnomacsave
20448 Mark the @code{MAC} register as call-clobbered, even if
20449 @option{-mhitachi} is given.
20455 Control the IEEE compliance of floating-point comparisons, which affects the
20456 handling of cases where the result of a comparison is unordered. By default
20457 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20458 enabled @option{-mno-ieee} is implicitly set, which results in faster
20459 floating-point greater-equal and less-equal comparisons. The implcit settings
20460 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20462 @item -minline-ic_invalidate
20463 @opindex minline-ic_invalidate
20464 Inline code to invalidate instruction cache entries after setting up
20465 nested function trampolines.
20466 This option has no effect if @option{-musermode} is in effect and the selected
20467 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20469 If the selected code generation option does not allow the use of the @code{icbi}
20470 instruction, and @option{-musermode} is not in effect, the inlined code
20471 manipulates the instruction cache address array directly with an associative
20472 write. This not only requires privileged mode at run time, but it also
20473 fails if the cache line had been mapped via the TLB and has become unmapped.
20477 Dump instruction size and location in the assembly code.
20480 @opindex mpadstruct
20481 This option is deprecated. It pads structures to multiple of 4 bytes,
20482 which is incompatible with the SH ABI@.
20484 @item -matomic-model=@var{model}
20485 @opindex matomic-model=@var{model}
20486 Sets the model of atomic operations and additional parameters as a comma
20487 separated list. For details on the atomic built-in functions see
20488 @ref{__atomic Builtins}. The following models and parameters are supported:
20493 Disable compiler generated atomic sequences and emit library calls for atomic
20494 operations. This is the default if the target is not @code{sh-*-linux*}.
20497 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20498 built-in functions. The generated atomic sequences require additional support
20499 from the interrupt/exception handling code of the system and are only suitable
20500 for SH3* and SH4* single-core systems. This option is enabled by default when
20501 the target is @code{sh-*-linux*} and SH3* or SH4*. When the target is SH4A,
20502 this option will also partially utilize the hardware atomic instructions
20503 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20504 @samp{strict} is specified.
20507 Generate software atomic sequences that use a variable in the thread control
20508 block. This is a variation of the gUSA sequences which can also be used on
20509 SH1* and SH2* targets. The generated atomic sequences require additional
20510 support from the interrupt/exception handling code of the system and are only
20511 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20512 parameter has to be specified as well.
20515 Generate software atomic sequences that temporarily disable interrupts by
20516 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20517 in privileged mode and is only suitable for single-core systems. Additional
20518 support from the interrupt/exception handling code of the system is not
20519 required. This model is enabled by default when the target is
20520 @code{sh-*-linux*} and SH1* or SH2*.
20523 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20524 instructions only. This is only available on SH4A and is suitable for
20525 multi-core systems. Since the hardware instructions support only 32 bit atomic
20526 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20527 Code compiled with this option will also be compatible with other software
20528 atomic model interrupt/exception handling systems if executed on an SH4A
20529 system. Additional support from the interrupt/exception handling code of the
20530 system is not required for this model.
20533 This parameter specifies the offset in bytes of the variable in the thread
20534 control block structure that should be used by the generated atomic sequences
20535 when the @samp{soft-tcb} model has been selected. For other models this
20536 parameter is ignored. The specified value must be an integer multiple of four
20537 and in the range 0-1020.
20540 This parameter prevents mixed usage of multiple atomic models, even though they
20541 would be compatible, and will make the compiler generate atomic sequences of the
20542 specified model only.
20548 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20549 Notice that depending on the particular hardware and software configuration
20550 this can degrade overall performance due to the operand cache line flushes
20551 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20552 processors the @code{tas.b} instruction must be used with caution since it
20553 can result in data corruption for certain cache configurations.
20557 Optimize for space instead of speed. Implied by @option{-Os}.
20560 @opindex mprefergot
20561 When generating position-independent code, emit function calls using
20562 the Global Offset Table instead of the Procedure Linkage Table.
20566 Don't generate privileged mode only code. This option
20567 implies @option{-mno-inline-ic_invalidate}
20568 if the inlined code would not work in user mode.
20569 This is the default when the target is @code{sh-*-linux*}.
20571 @item -multcost=@var{number}
20572 @opindex multcost=@var{number}
20573 Set the cost to assume for a multiply insn.
20575 @item -mdiv=@var{strategy}
20576 @opindex mdiv=@var{strategy}
20577 Set the division strategy to be used for integer division operations.
20578 For SHmedia @var{strategy} can be one of:
20583 Performs the operation in floating point. This has a very high latency,
20584 but needs only a few instructions, so it might be a good choice if
20585 your code has enough easily-exploitable ILP to allow the compiler to
20586 schedule the floating-point instructions together with other instructions.
20587 Division by zero causes a floating-point exception.
20590 Uses integer operations to calculate the inverse of the divisor,
20591 and then multiplies the dividend with the inverse. This strategy allows
20592 CSE and hoisting of the inverse calculation. Division by zero calculates
20593 an unspecified result, but does not trap.
20596 A variant of @samp{inv} where, if no CSE or hoisting opportunities
20597 have been found, or if the entire operation has been hoisted to the same
20598 place, the last stages of the inverse calculation are intertwined with the
20599 final multiply to reduce the overall latency, at the expense of using a few
20600 more instructions, and thus offering fewer scheduling opportunities with
20604 Calls a library function that usually implements the @samp{inv:minlat}
20606 This gives high code density for @code{m5-*media-nofpu} compilations.
20609 Uses a different entry point of the same library function, where it
20610 assumes that a pointer to a lookup table has already been set up, which
20611 exposes the pointer load to CSE and code hoisting optimizations.
20616 Use the @samp{inv} algorithm for initial
20617 code generation, but if the code stays unoptimized, revert to the @samp{call},
20618 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
20619 potentially-trapping side effect of division by zero is carried by a
20620 separate instruction, so it is possible that all the integer instructions
20621 are hoisted out, but the marker for the side effect stays where it is.
20622 A recombination to floating-point operations or a call is not possible
20627 Variants of the @samp{inv:minlat} strategy. In the case
20628 that the inverse calculation is not separated from the multiply, they speed
20629 up division where the dividend fits into 20 bits (plus sign where applicable)
20630 by inserting a test to skip a number of operations in this case; this test
20631 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
20632 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
20636 For targets other than SHmedia @var{strategy} can be one of:
20641 Calls a library function that uses the single-step division instruction
20642 @code{div1} to perform the operation. Division by zero calculates an
20643 unspecified result and does not trap. This is the default except for SH4,
20644 SH2A and SHcompact.
20647 Calls a library function that performs the operation in double precision
20648 floating point. Division by zero causes a floating-point exception. This is
20649 the default for SHcompact with FPU. Specifying this for targets that do not
20650 have a double precision FPU will default to @code{call-div1}.
20653 Calls a library function that uses a lookup table for small divisors and
20654 the @code{div1} instruction with case distinction for larger divisors. Division
20655 by zero calculates an unspecified result and does not trap. This is the default
20656 for SH4. Specifying this for targets that do not have dynamic shift
20657 instructions will default to @code{call-div1}.
20661 When a division strategy has not been specified the default strategy will be
20662 selected based on the current target. For SH2A the default strategy is to
20663 use the @code{divs} and @code{divu} instructions instead of library function
20666 @item -maccumulate-outgoing-args
20667 @opindex maccumulate-outgoing-args
20668 Reserve space once for outgoing arguments in the function prologue rather
20669 than around each call. Generally beneficial for performance and size. Also
20670 needed for unwinding to avoid changing the stack frame around conditional code.
20672 @item -mdivsi3_libfunc=@var{name}
20673 @opindex mdivsi3_libfunc=@var{name}
20674 Set the name of the library function used for 32-bit signed division to
20676 This only affects the name used in the @samp{call} and @samp{inv:call}
20677 division strategies, and the compiler still expects the same
20678 sets of input/output/clobbered registers as if this option were not present.
20680 @item -mfixed-range=@var{register-range}
20681 @opindex mfixed-range
20682 Generate code treating the given register range as fixed registers.
20683 A fixed register is one that the register allocator can not use. This is
20684 useful when compiling kernel code. A register range is specified as
20685 two registers separated by a dash. Multiple register ranges can be
20686 specified separated by a comma.
20688 @item -mindexed-addressing
20689 @opindex mindexed-addressing
20690 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
20691 This is only safe if the hardware and/or OS implement 32-bit wrap-around
20692 semantics for the indexed addressing mode. The architecture allows the
20693 implementation of processors with 64-bit MMU, which the OS could use to
20694 get 32-bit addressing, but since no current hardware implementation supports
20695 this or any other way to make the indexed addressing mode safe to use in
20696 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
20698 @item -mgettrcost=@var{number}
20699 @opindex mgettrcost=@var{number}
20700 Set the cost assumed for the @code{gettr} instruction to @var{number}.
20701 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
20705 Assume @code{pt*} instructions won't trap. This generally generates
20706 better-scheduled code, but is unsafe on current hardware.
20707 The current architecture
20708 definition says that @code{ptabs} and @code{ptrel} trap when the target
20710 This has the unintentional effect of making it unsafe to schedule these
20711 instructions before a branch, or hoist them out of a loop. For example,
20712 @code{__do_global_ctors}, a part of @file{libgcc}
20713 that runs constructors at program
20714 startup, calls functions in a list which is delimited by @minus{}1. With the
20715 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
20716 That means that all the constructors run a bit more quickly, but when
20717 the loop comes to the end of the list, the program crashes because @code{ptabs}
20718 loads @minus{}1 into a target register.
20720 Since this option is unsafe for any
20721 hardware implementing the current architecture specification, the default
20722 is @option{-mno-pt-fixed}. Unless specified explicitly with
20723 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
20724 this deters register allocation from using target registers for storing
20727 @item -minvalid-symbols
20728 @opindex minvalid-symbols
20729 Assume symbols might be invalid. Ordinary function symbols generated by
20730 the compiler are always valid to load with
20731 @code{movi}/@code{shori}/@code{ptabs} or
20732 @code{movi}/@code{shori}/@code{ptrel},
20733 but with assembler and/or linker tricks it is possible
20734 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
20735 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
20736 It prevents cross-basic-block CSE, hoisting and most scheduling
20737 of symbol loads. The default is @option{-mno-invalid-symbols}.
20739 @item -mbranch-cost=@var{num}
20740 @opindex mbranch-cost=@var{num}
20741 Assume @var{num} to be the cost for a branch instruction. Higher numbers
20742 make the compiler try to generate more branch-free code if possible.
20743 If not specified the value is selected depending on the processor type that
20744 is being compiled for.
20747 @itemx -mno-zdcbranch
20748 @opindex mzdcbranch
20749 @opindex mno-zdcbranch
20750 Assume (do not assume) that zero displacement conditional branch instructions
20751 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
20752 compiler will try to prefer zero displacement branch code sequences. This is
20753 enabled by default when generating code for SH4 and SH4A. It can be explicitly
20754 disabled by specifying @option{-mno-zdcbranch}.
20757 @itemx -mno-fused-madd
20758 @opindex mfused-madd
20759 @opindex mno-fused-madd
20760 Generate code that uses (does not use) the floating-point multiply and
20761 accumulate instructions. These instructions are generated by default
20762 if hardware floating point is used. The machine-dependent
20763 @option{-mfused-madd} option is now mapped to the machine-independent
20764 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20765 mapped to @option{-ffp-contract=off}.
20771 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
20772 and cosine approximations. The option @code{-mfsca} must be used in
20773 combination with @code{-funsafe-math-optimizations}. It is enabled by default
20774 when generating code for SH4A. Using @code{-mno-fsca} disables sine and cosine
20775 approximations even if @code{-funsafe-math-optimizations} is in effect.
20781 Allow or disallow the compiler to emit the @code{fsrra} instruction for
20782 reciprocal square root approximations. The option @code{-mfsrra} must be used
20783 in combination with @code{-funsafe-math-optimizations} and
20784 @code{-ffinite-math-only}. It is enabled by default when generating code for
20785 SH4A. Using @code{-mno-fsrra} disables reciprocal square root approximations
20786 even if @code{-funsafe-math-optimizations} and @code{-ffinite-math-only} are
20789 @item -mpretend-cmove
20790 @opindex mpretend-cmove
20791 Prefer zero-displacement conditional branches for conditional move instruction
20792 patterns. This can result in faster code on the SH4 processor.
20796 @node Solaris 2 Options
20797 @subsection Solaris 2 Options
20798 @cindex Solaris 2 options
20800 These @samp{-m} options are supported on Solaris 2:
20803 @item -mimpure-text
20804 @opindex mimpure-text
20805 @option{-mimpure-text}, used in addition to @option{-shared}, tells
20806 the compiler to not pass @option{-z text} to the linker when linking a
20807 shared object. Using this option, you can link position-dependent
20808 code into a shared object.
20810 @option{-mimpure-text} suppresses the ``relocations remain against
20811 allocatable but non-writable sections'' linker error message.
20812 However, the necessary relocations trigger copy-on-write, and the
20813 shared object is not actually shared across processes. Instead of
20814 using @option{-mimpure-text}, you should compile all source code with
20815 @option{-fpic} or @option{-fPIC}.
20819 These switches are supported in addition to the above on Solaris 2:
20824 Add support for multithreading using the POSIX threads library. This
20825 option sets flags for both the preprocessor and linker. This option does
20826 not affect the thread safety of object code produced by the compiler or
20827 that of libraries supplied with it.
20831 This is a synonym for @option{-pthreads}.
20834 @node SPARC Options
20835 @subsection SPARC Options
20836 @cindex SPARC options
20838 These @samp{-m} options are supported on the SPARC:
20841 @item -mno-app-regs
20843 @opindex mno-app-regs
20845 Specify @option{-mapp-regs} to generate output using the global registers
20846 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
20847 global register 1, each global register 2 through 4 is then treated as an
20848 allocable register that is clobbered by function calls. This is the default.
20850 To be fully SVR4 ABI-compliant at the cost of some performance loss,
20851 specify @option{-mno-app-regs}. You should compile libraries and system
20852 software with this option.
20858 With @option{-mflat}, the compiler does not generate save/restore instructions
20859 and uses a ``flat'' or single register window model. This model is compatible
20860 with the regular register window model. The local registers and the input
20861 registers (0--5) are still treated as ``call-saved'' registers and are
20862 saved on the stack as needed.
20864 With @option{-mno-flat} (the default), the compiler generates save/restore
20865 instructions (except for leaf functions). This is the normal operating mode.
20868 @itemx -mhard-float
20870 @opindex mhard-float
20871 Generate output containing floating-point instructions. This is the
20875 @itemx -msoft-float
20877 @opindex msoft-float
20878 Generate output containing library calls for floating point.
20879 @strong{Warning:} the requisite libraries are not available for all SPARC
20880 targets. Normally the facilities of the machine's usual C compiler are
20881 used, but this cannot be done directly in cross-compilation. You must make
20882 your own arrangements to provide suitable library functions for
20883 cross-compilation. The embedded targets @samp{sparc-*-aout} and
20884 @samp{sparclite-*-*} do provide software floating-point support.
20886 @option{-msoft-float} changes the calling convention in the output file;
20887 therefore, it is only useful if you compile @emph{all} of a program with
20888 this option. In particular, you need to compile @file{libgcc.a}, the
20889 library that comes with GCC, with @option{-msoft-float} in order for
20892 @item -mhard-quad-float
20893 @opindex mhard-quad-float
20894 Generate output containing quad-word (long double) floating-point
20897 @item -msoft-quad-float
20898 @opindex msoft-quad-float
20899 Generate output containing library calls for quad-word (long double)
20900 floating-point instructions. The functions called are those specified
20901 in the SPARC ABI@. This is the default.
20903 As of this writing, there are no SPARC implementations that have hardware
20904 support for the quad-word floating-point instructions. They all invoke
20905 a trap handler for one of these instructions, and then the trap handler
20906 emulates the effect of the instruction. Because of the trap handler overhead,
20907 this is much slower than calling the ABI library routines. Thus the
20908 @option{-msoft-quad-float} option is the default.
20910 @item -mno-unaligned-doubles
20911 @itemx -munaligned-doubles
20912 @opindex mno-unaligned-doubles
20913 @opindex munaligned-doubles
20914 Assume that doubles have 8-byte alignment. This is the default.
20916 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
20917 alignment only if they are contained in another type, or if they have an
20918 absolute address. Otherwise, it assumes they have 4-byte alignment.
20919 Specifying this option avoids some rare compatibility problems with code
20920 generated by other compilers. It is not the default because it results
20921 in a performance loss, especially for floating-point code.
20923 @item -mno-faster-structs
20924 @itemx -mfaster-structs
20925 @opindex mno-faster-structs
20926 @opindex mfaster-structs
20927 With @option{-mfaster-structs}, the compiler assumes that structures
20928 should have 8-byte alignment. This enables the use of pairs of
20929 @code{ldd} and @code{std} instructions for copies in structure
20930 assignment, in place of twice as many @code{ld} and @code{st} pairs.
20931 However, the use of this changed alignment directly violates the SPARC
20932 ABI@. Thus, it's intended only for use on targets where the developer
20933 acknowledges that their resulting code is not directly in line with
20934 the rules of the ABI@.
20936 @item -mcpu=@var{cpu_type}
20938 Set the instruction set, register set, and instruction scheduling parameters
20939 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
20940 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
20941 @samp{leon}, @samp{leon3}, @samp{sparclite}, @samp{f930}, @samp{f934},
20942 @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
20943 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
20944 @samp{niagara3} and @samp{niagara4}.
20946 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
20947 which selects the best architecture option for the host processor.
20948 @option{-mcpu=native} has no effect if GCC does not recognize
20951 Default instruction scheduling parameters are used for values that select
20952 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
20953 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
20955 Here is a list of each supported architecture and their supported
20963 supersparc, hypersparc, leon, leon3
20966 f930, f934, sparclite86x
20972 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
20975 By default (unless configured otherwise), GCC generates code for the V7
20976 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
20977 additionally optimizes it for the Cypress CY7C602 chip, as used in the
20978 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
20979 SPARCStation 1, 2, IPX etc.
20981 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
20982 architecture. The only difference from V7 code is that the compiler emits
20983 the integer multiply and integer divide instructions which exist in SPARC-V8
20984 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
20985 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
20988 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
20989 the SPARC architecture. This adds the integer multiply, integer divide step
20990 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
20991 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
20992 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
20993 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
20994 MB86934 chip, which is the more recent SPARClite with FPU@.
20996 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
20997 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
20998 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
20999 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21000 optimizes it for the TEMIC SPARClet chip.
21002 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21003 architecture. This adds 64-bit integer and floating-point move instructions,
21004 3 additional floating-point condition code registers and conditional move
21005 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21006 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21007 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21008 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21009 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21010 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21011 additionally optimizes it for Sun UltraSPARC T2 chips. With
21012 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21013 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21014 additionally optimizes it for Sun UltraSPARC T4 chips.
21016 @item -mtune=@var{cpu_type}
21018 Set the instruction scheduling parameters for machine type
21019 @var{cpu_type}, but do not set the instruction set or register set that the
21020 option @option{-mcpu=@var{cpu_type}} does.
21022 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21023 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21024 that select a particular CPU implementation. Those are @samp{cypress},
21025 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3}, @samp{f930},
21026 @samp{f934}, @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
21027 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3} and
21028 @samp{niagara4}. With native Solaris and GNU/Linux toolchains, @samp{native}
21034 @opindex mno-v8plus
21035 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21036 difference from the V8 ABI is that the global and out registers are
21037 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21038 mode for all SPARC-V9 processors.
21044 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21045 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21051 With @option{-mvis2}, GCC generates code that takes advantage of
21052 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21053 default is @option{-mvis2} when targeting a cpu that supports such
21054 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21055 also sets @option{-mvis}.
21061 With @option{-mvis3}, GCC generates code that takes advantage of
21062 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21063 default is @option{-mvis3} when targeting a cpu that supports such
21064 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21065 also sets @option{-mvis2} and @option{-mvis}.
21070 @opindex mno-cbcond
21071 With @option{-mcbcond}, GCC generates code that takes advantage of
21072 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21073 The default is @option{-mcbcond} when targeting a cpu that supports such
21074 instructions, such as niagara-4 and later.
21080 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21081 population count instruction. The default is @option{-mpopc}
21082 when targeting a cpu that supports such instructions, such as Niagara-2 and
21089 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21090 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21091 when targeting a cpu that supports such instructions, such as Niagara-3 and
21095 @opindex mfix-at697f
21096 Enable the documented workaround for the single erratum of the Atmel AT697F
21097 processor (which corresponds to erratum #13 of the AT697E processor).
21100 @opindex mfix-ut699
21101 Enable the documented workarounds for the floating-point errata and the data
21102 cache nullify errata of the UT699 processor.
21105 These @samp{-m} options are supported in addition to the above
21106 on SPARC-V9 processors in 64-bit environments:
21113 Generate code for a 32-bit or 64-bit environment.
21114 The 32-bit environment sets int, long and pointer to 32 bits.
21115 The 64-bit environment sets int to 32 bits and long and pointer
21118 @item -mcmodel=@var{which}
21120 Set the code model to one of
21124 The Medium/Low code model: 64-bit addresses, programs
21125 must be linked in the low 32 bits of memory. Programs can be statically
21126 or dynamically linked.
21129 The Medium/Middle code model: 64-bit addresses, programs
21130 must be linked in the low 44 bits of memory, the text and data segments must
21131 be less than 2GB in size and the data segment must be located within 2GB of
21135 The Medium/Anywhere code model: 64-bit addresses, programs
21136 may be linked anywhere in memory, the text and data segments must be less
21137 than 2GB in size and the data segment must be located within 2GB of the
21141 The Medium/Anywhere code model for embedded systems:
21142 64-bit addresses, the text and data segments must be less than 2GB in
21143 size, both starting anywhere in memory (determined at link time). The
21144 global register %g4 points to the base of the data segment. Programs
21145 are statically linked and PIC is not supported.
21148 @item -mmemory-model=@var{mem-model}
21149 @opindex mmemory-model
21150 Set the memory model in force on the processor to one of
21154 The default memory model for the processor and operating system.
21157 Relaxed Memory Order
21160 Partial Store Order
21166 Sequential Consistency
21169 These memory models are formally defined in Appendix D of the Sparc V9
21170 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21173 @itemx -mno-stack-bias
21174 @opindex mstack-bias
21175 @opindex mno-stack-bias
21176 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21177 frame pointer if present, are offset by @minus{}2047 which must be added back
21178 when making stack frame references. This is the default in 64-bit mode.
21179 Otherwise, assume no such offset is present.
21183 @subsection SPU Options
21184 @cindex SPU options
21186 These @samp{-m} options are supported on the SPU:
21190 @itemx -merror-reloc
21191 @opindex mwarn-reloc
21192 @opindex merror-reloc
21194 The loader for SPU does not handle dynamic relocations. By default, GCC
21195 gives an error when it generates code that requires a dynamic
21196 relocation. @option{-mno-error-reloc} disables the error,
21197 @option{-mwarn-reloc} generates a warning instead.
21200 @itemx -munsafe-dma
21202 @opindex munsafe-dma
21204 Instructions that initiate or test completion of DMA must not be
21205 reordered with respect to loads and stores of the memory that is being
21207 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21208 memory accesses, but that can lead to inefficient code in places where the
21209 memory is known to not change. Rather than mark the memory as volatile,
21210 you can use @option{-msafe-dma} to tell the compiler to treat
21211 the DMA instructions as potentially affecting all memory.
21213 @item -mbranch-hints
21214 @opindex mbranch-hints
21216 By default, GCC generates a branch hint instruction to avoid
21217 pipeline stalls for always-taken or probably-taken branches. A hint
21218 is not generated closer than 8 instructions away from its branch.
21219 There is little reason to disable them, except for debugging purposes,
21220 or to make an object a little bit smaller.
21224 @opindex msmall-mem
21225 @opindex mlarge-mem
21227 By default, GCC generates code assuming that addresses are never larger
21228 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21229 a full 32-bit address.
21234 By default, GCC links against startup code that assumes the SPU-style
21235 main function interface (which has an unconventional parameter list).
21236 With @option{-mstdmain}, GCC links your program against startup
21237 code that assumes a C99-style interface to @code{main}, including a
21238 local copy of @code{argv} strings.
21240 @item -mfixed-range=@var{register-range}
21241 @opindex mfixed-range
21242 Generate code treating the given register range as fixed registers.
21243 A fixed register is one that the register allocator cannot use. This is
21244 useful when compiling kernel code. A register range is specified as
21245 two registers separated by a dash. Multiple register ranges can be
21246 specified separated by a comma.
21252 Compile code assuming that pointers to the PPU address space accessed
21253 via the @code{__ea} named address space qualifier are either 32 or 64
21254 bits wide. The default is 32 bits. As this is an ABI-changing option,
21255 all object code in an executable must be compiled with the same setting.
21257 @item -maddress-space-conversion
21258 @itemx -mno-address-space-conversion
21259 @opindex maddress-space-conversion
21260 @opindex mno-address-space-conversion
21261 Allow/disallow treating the @code{__ea} address space as superset
21262 of the generic address space. This enables explicit type casts
21263 between @code{__ea} and generic pointer as well as implicit
21264 conversions of generic pointers to @code{__ea} pointers. The
21265 default is to allow address space pointer conversions.
21267 @item -mcache-size=@var{cache-size}
21268 @opindex mcache-size
21269 This option controls the version of libgcc that the compiler links to an
21270 executable and selects a software-managed cache for accessing variables
21271 in the @code{__ea} address space with a particular cache size. Possible
21272 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21273 and @samp{128}. The default cache size is 64KB.
21275 @item -matomic-updates
21276 @itemx -mno-atomic-updates
21277 @opindex matomic-updates
21278 @opindex mno-atomic-updates
21279 This option controls the version of libgcc that the compiler links to an
21280 executable and selects whether atomic updates to the software-managed
21281 cache of PPU-side variables are used. If you use atomic updates, changes
21282 to a PPU variable from SPU code using the @code{__ea} named address space
21283 qualifier do not interfere with changes to other PPU variables residing
21284 in the same cache line from PPU code. If you do not use atomic updates,
21285 such interference may occur; however, writing back cache lines is
21286 more efficient. The default behavior is to use atomic updates.
21289 @itemx -mdual-nops=@var{n}
21290 @opindex mdual-nops
21291 By default, GCC inserts nops to increase dual issue when it expects
21292 it to increase performance. @var{n} can be a value from 0 to 10. A
21293 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21294 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21296 @item -mhint-max-nops=@var{n}
21297 @opindex mhint-max-nops
21298 Maximum number of nops to insert for a branch hint. A branch hint must
21299 be at least 8 instructions away from the branch it is affecting. GCC
21300 inserts up to @var{n} nops to enforce this, otherwise it does not
21301 generate the branch hint.
21303 @item -mhint-max-distance=@var{n}
21304 @opindex mhint-max-distance
21305 The encoding of the branch hint instruction limits the hint to be within
21306 256 instructions of the branch it is affecting. By default, GCC makes
21307 sure it is within 125.
21310 @opindex msafe-hints
21311 Work around a hardware bug that causes the SPU to stall indefinitely.
21312 By default, GCC inserts the @code{hbrp} instruction to make sure
21313 this stall won't happen.
21317 @node System V Options
21318 @subsection Options for System V
21320 These additional options are available on System V Release 4 for
21321 compatibility with other compilers on those systems:
21326 Create a shared object.
21327 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21331 Identify the versions of each tool used by the compiler, in a
21332 @code{.ident} assembler directive in the output.
21336 Refrain from adding @code{.ident} directives to the output file (this is
21339 @item -YP,@var{dirs}
21341 Search the directories @var{dirs}, and no others, for libraries
21342 specified with @option{-l}.
21344 @item -Ym,@var{dir}
21346 Look in the directory @var{dir} to find the M4 preprocessor.
21347 The assembler uses this option.
21348 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21349 @c the generic assembler that comes with Solaris takes just -Ym.
21352 @node TILE-Gx Options
21353 @subsection TILE-Gx Options
21354 @cindex TILE-Gx options
21356 These @samp{-m} options are supported on the TILE-Gx:
21359 @item -mcmodel=small
21360 @opindex mcmodel=small
21361 Generate code for the small model. The distance for direct calls is
21362 limited to 500M in either direction. PC-relative addresses are 32
21363 bits. Absolute addresses support the full address range.
21365 @item -mcmodel=large
21366 @opindex mcmodel=large
21367 Generate code for the large model. There is no limitation on call
21368 distance, pc-relative addresses, or absolute addresses.
21370 @item -mcpu=@var{name}
21372 Selects the type of CPU to be targeted. Currently the only supported
21373 type is @samp{tilegx}.
21379 Generate code for a 32-bit or 64-bit environment. The 32-bit
21380 environment sets int, long, and pointer to 32 bits. The 64-bit
21381 environment sets int to 32 bits and long and pointer to 64 bits.
21384 @itemx -mlittle-endian
21385 @opindex mbig-endian
21386 @opindex mlittle-endian
21387 Generate code in big/little endian mode, respectively.
21390 @node TILEPro Options
21391 @subsection TILEPro Options
21392 @cindex TILEPro options
21394 These @samp{-m} options are supported on the TILEPro:
21397 @item -mcpu=@var{name}
21399 Selects the type of CPU to be targeted. Currently the only supported
21400 type is @samp{tilepro}.
21404 Generate code for a 32-bit environment, which sets int, long, and
21405 pointer to 32 bits. This is the only supported behavior so the flag
21406 is essentially ignored.
21410 @subsection V850 Options
21411 @cindex V850 Options
21413 These @samp{-m} options are defined for V850 implementations:
21417 @itemx -mno-long-calls
21418 @opindex mlong-calls
21419 @opindex mno-long-calls
21420 Treat all calls as being far away (near). If calls are assumed to be
21421 far away, the compiler always loads the function's address into a
21422 register, and calls indirect through the pointer.
21428 Do not optimize (do optimize) basic blocks that use the same index
21429 pointer 4 or more times to copy pointer into the @code{ep} register, and
21430 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21431 option is on by default if you optimize.
21433 @item -mno-prolog-function
21434 @itemx -mprolog-function
21435 @opindex mno-prolog-function
21436 @opindex mprolog-function
21437 Do not use (do use) external functions to save and restore registers
21438 at the prologue and epilogue of a function. The external functions
21439 are slower, but use less code space if more than one function saves
21440 the same number of registers. The @option{-mprolog-function} option
21441 is on by default if you optimize.
21445 Try to make the code as small as possible. At present, this just turns
21446 on the @option{-mep} and @option{-mprolog-function} options.
21448 @item -mtda=@var{n}
21450 Put static or global variables whose size is @var{n} bytes or less into
21451 the tiny data area that register @code{ep} points to. The tiny data
21452 area can hold up to 256 bytes in total (128 bytes for byte references).
21454 @item -msda=@var{n}
21456 Put static or global variables whose size is @var{n} bytes or less into
21457 the small data area that register @code{gp} points to. The small data
21458 area can hold up to 64 kilobytes.
21460 @item -mzda=@var{n}
21462 Put static or global variables whose size is @var{n} bytes or less into
21463 the first 32 kilobytes of memory.
21467 Specify that the target processor is the V850.
21471 Specify that the target processor is the V850E3V5. The preprocessor
21472 constant @samp{__v850e3v5__} is defined if this option is used.
21476 Specify that the target processor is the V850E3V5. This is an alias for
21477 the @option{-mv850e3v5} option.
21481 Specify that the target processor is the V850E2V3. The preprocessor
21482 constant @samp{__v850e2v3__} is defined if this option is used.
21486 Specify that the target processor is the V850E2. The preprocessor
21487 constant @samp{__v850e2__} is defined if this option is used.
21491 Specify that the target processor is the V850E1. The preprocessor
21492 constants @samp{__v850e1__} and @samp{__v850e__} are defined if
21493 this option is used.
21497 Specify that the target processor is the V850ES. This is an alias for
21498 the @option{-mv850e1} option.
21502 Specify that the target processor is the V850E@. The preprocessor
21503 constant @samp{__v850e__} is defined if this option is used.
21505 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21506 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21507 are defined then a default target processor is chosen and the
21508 relevant @samp{__v850*__} preprocessor constant is defined.
21510 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
21511 defined, regardless of which processor variant is the target.
21513 @item -mdisable-callt
21514 @itemx -mno-disable-callt
21515 @opindex mdisable-callt
21516 @opindex mno-disable-callt
21517 This option suppresses generation of the @code{CALLT} instruction for the
21518 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21521 This option is enabled by default when the RH850 ABI is
21522 in use (see @option{-mrh850-abi}), and disabled by default when the
21523 GCC ABI is in use. If @code{CALLT} instructions are being generated
21524 then the C preprocessor symbol @code{__V850_CALLT__} will be defined.
21530 Pass on (or do not pass on) the @option{-mrelax} command line option
21534 @itemx -mno-long-jumps
21535 @opindex mlong-jumps
21536 @opindex mno-long-jumps
21537 Disable (or re-enable) the generation of PC-relative jump instructions.
21540 @itemx -mhard-float
21541 @opindex msoft-float
21542 @opindex mhard-float
21543 Disable (or re-enable) the generation of hardware floating point
21544 instructions. This option is only significant when the target
21545 architecture is @samp{V850E2V3} or higher. If hardware floating point
21546 instructions are being generated then the C preprocessor symbol
21547 @code{__FPU_OK__} will be defined, otherwise the symbol
21548 @code{__NO_FPU__} will be defined.
21552 Enables the use of the e3v5 LOOP instruction. The use of this
21553 instruction is not enabled by default when the e3v5 architecture is
21554 selected because its use is still experimental.
21558 @opindex mrh850-abi
21560 Enables support for the RH850 version of the V850 ABI. This is the
21561 default. With this version of the ABI the following rules apply:
21565 Integer sized structures and unions are returned via a memory pointer
21566 rather than a register.
21569 Large structures and unions (more than 8 bytes in size) are passed by
21573 Functions are aligned to 16-bit boundaries.
21576 The @option{-m8byte-align} command line option is supported.
21579 The @option{-mdisable-callt} command line option is enabled by
21580 default. The @option{-mno-disable-callt} command line option is not
21584 When this version of the ABI is enabled the C preprocessor symbol
21585 @code{__V850_RH850_ABI__} is defined.
21589 Enables support for the old GCC version of the V850 ABI. With this
21590 version of the ABI the following rules apply:
21594 Integer sized structures and unions are returned in register @code{r10}.
21597 Large structures and unions (more than 8 bytes in size) are passed by
21601 Functions are aligned to 32-bit boundaries, unless optimizing for
21605 The @option{-m8byte-align} command line option is not supported.
21608 The @option{-mdisable-callt} command line option is supported but not
21609 enabled by default.
21612 When this version of the ABI is enabled the C preprocessor symbol
21613 @code{__V850_GCC_ABI__} is defined.
21615 @item -m8byte-align
21616 @itemx -mno-8byte-align
21617 @opindex m8byte-align
21618 @opindex mno-8byte-align
21619 Enables support for @code{doubles} and @code{long long} types to be
21620 aligned on 8-byte boundaries. The default is to restrict the
21621 alignment of all objects to at most 4-bytes. When
21622 @option{-m8byte-align} is in effect the C preprocessor symbol
21623 @code{__V850_8BYTE_ALIGN__} will be defined.
21626 @opindex mbig-switch
21627 Generate code suitable for big switch tables. Use this option only if
21628 the assembler/linker complain about out of range branches within a switch
21633 This option causes r2 and r5 to be used in the code generated by
21634 the compiler. This setting is the default.
21636 @item -mno-app-regs
21637 @opindex mno-app-regs
21638 This option causes r2 and r5 to be treated as fixed registers.
21643 @subsection VAX Options
21644 @cindex VAX options
21646 These @samp{-m} options are defined for the VAX:
21651 Do not output certain jump instructions (@code{aobleq} and so on)
21652 that the Unix assembler for the VAX cannot handle across long
21657 Do output those jump instructions, on the assumption that the
21658 GNU assembler is being used.
21662 Output code for G-format floating-point numbers instead of D-format.
21666 @subsection VMS Options
21668 These @samp{-m} options are defined for the VMS implementations:
21671 @item -mvms-return-codes
21672 @opindex mvms-return-codes
21673 Return VMS condition codes from @code{main}. The default is to return POSIX-style
21674 condition (e.g.@ error) codes.
21676 @item -mdebug-main=@var{prefix}
21677 @opindex mdebug-main=@var{prefix}
21678 Flag the first routine whose name starts with @var{prefix} as the main
21679 routine for the debugger.
21683 Default to 64-bit memory allocation routines.
21685 @item -mpointer-size=@var{size}
21686 @opindex -mpointer-size=@var{size}
21687 Set the default size of pointers. Possible options for @var{size} are
21688 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
21689 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
21690 The later option disables @code{pragma pointer_size}.
21693 @node VxWorks Options
21694 @subsection VxWorks Options
21695 @cindex VxWorks Options
21697 The options in this section are defined for all VxWorks targets.
21698 Options specific to the target hardware are listed with the other
21699 options for that target.
21704 GCC can generate code for both VxWorks kernels and real time processes
21705 (RTPs). This option switches from the former to the latter. It also
21706 defines the preprocessor macro @code{__RTP__}.
21709 @opindex non-static
21710 Link an RTP executable against shared libraries rather than static
21711 libraries. The options @option{-static} and @option{-shared} can
21712 also be used for RTPs (@pxref{Link Options}); @option{-static}
21719 These options are passed down to the linker. They are defined for
21720 compatibility with Diab.
21723 @opindex Xbind-lazy
21724 Enable lazy binding of function calls. This option is equivalent to
21725 @option{-Wl,-z,now} and is defined for compatibility with Diab.
21729 Disable lazy binding of function calls. This option is the default and
21730 is defined for compatibility with Diab.
21733 @node x86-64 Options
21734 @subsection x86-64 Options
21735 @cindex x86-64 options
21737 These are listed under @xref{i386 and x86-64 Options}.
21739 @node Xstormy16 Options
21740 @subsection Xstormy16 Options
21741 @cindex Xstormy16 Options
21743 These options are defined for Xstormy16:
21748 Choose startup files and linker script suitable for the simulator.
21751 @node Xtensa Options
21752 @subsection Xtensa Options
21753 @cindex Xtensa Options
21755 These options are supported for Xtensa targets:
21759 @itemx -mno-const16
21761 @opindex mno-const16
21762 Enable or disable use of @code{CONST16} instructions for loading
21763 constant values. The @code{CONST16} instruction is currently not a
21764 standard option from Tensilica. When enabled, @code{CONST16}
21765 instructions are always used in place of the standard @code{L32R}
21766 instructions. The use of @code{CONST16} is enabled by default only if
21767 the @code{L32R} instruction is not available.
21770 @itemx -mno-fused-madd
21771 @opindex mfused-madd
21772 @opindex mno-fused-madd
21773 Enable or disable use of fused multiply/add and multiply/subtract
21774 instructions in the floating-point option. This has no effect if the
21775 floating-point option is not also enabled. Disabling fused multiply/add
21776 and multiply/subtract instructions forces the compiler to use separate
21777 instructions for the multiply and add/subtract operations. This may be
21778 desirable in some cases where strict IEEE 754-compliant results are
21779 required: the fused multiply add/subtract instructions do not round the
21780 intermediate result, thereby producing results with @emph{more} bits of
21781 precision than specified by the IEEE standard. Disabling fused multiply
21782 add/subtract instructions also ensures that the program output is not
21783 sensitive to the compiler's ability to combine multiply and add/subtract
21786 @item -mserialize-volatile
21787 @itemx -mno-serialize-volatile
21788 @opindex mserialize-volatile
21789 @opindex mno-serialize-volatile
21790 When this option is enabled, GCC inserts @code{MEMW} instructions before
21791 @code{volatile} memory references to guarantee sequential consistency.
21792 The default is @option{-mserialize-volatile}. Use
21793 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
21795 @item -mforce-no-pic
21796 @opindex mforce-no-pic
21797 For targets, like GNU/Linux, where all user-mode Xtensa code must be
21798 position-independent code (PIC), this option disables PIC for compiling
21801 @item -mtext-section-literals
21802 @itemx -mno-text-section-literals
21803 @opindex mtext-section-literals
21804 @opindex mno-text-section-literals
21805 Control the treatment of literal pools. The default is
21806 @option{-mno-text-section-literals}, which places literals in a separate
21807 section in the output file. This allows the literal pool to be placed
21808 in a data RAM/ROM, and it also allows the linker to combine literal
21809 pools from separate object files to remove redundant literals and
21810 improve code size. With @option{-mtext-section-literals}, the literals
21811 are interspersed in the text section in order to keep them as close as
21812 possible to their references. This may be necessary for large assembly
21815 @item -mtarget-align
21816 @itemx -mno-target-align
21817 @opindex mtarget-align
21818 @opindex mno-target-align
21819 When this option is enabled, GCC instructs the assembler to
21820 automatically align instructions to reduce branch penalties at the
21821 expense of some code density. The assembler attempts to widen density
21822 instructions to align branch targets and the instructions following call
21823 instructions. If there are not enough preceding safe density
21824 instructions to align a target, no widening is performed. The
21825 default is @option{-mtarget-align}. These options do not affect the
21826 treatment of auto-aligned instructions like @code{LOOP}, which the
21827 assembler always aligns, either by widening density instructions or
21828 by inserting NOP instructions.
21831 @itemx -mno-longcalls
21832 @opindex mlongcalls
21833 @opindex mno-longcalls
21834 When this option is enabled, GCC instructs the assembler to translate
21835 direct calls to indirect calls unless it can determine that the target
21836 of a direct call is in the range allowed by the call instruction. This
21837 translation typically occurs for calls to functions in other source
21838 files. Specifically, the assembler translates a direct @code{CALL}
21839 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
21840 The default is @option{-mno-longcalls}. This option should be used in
21841 programs where the call target can potentially be out of range. This
21842 option is implemented in the assembler, not the compiler, so the
21843 assembly code generated by GCC still shows direct call
21844 instructions---look at the disassembled object code to see the actual
21845 instructions. Note that the assembler uses an indirect call for
21846 every cross-file call, not just those that really are out of range.
21849 @node zSeries Options
21850 @subsection zSeries Options
21851 @cindex zSeries options
21853 These are listed under @xref{S/390 and zSeries Options}.
21855 @node Code Gen Options
21856 @section Options for Code Generation Conventions
21857 @cindex code generation conventions
21858 @cindex options, code generation
21859 @cindex run-time options
21861 These machine-independent options control the interface conventions
21862 used in code generation.
21864 Most of them have both positive and negative forms; the negative form
21865 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
21866 one of the forms is listed---the one that is not the default. You
21867 can figure out the other form by either removing @samp{no-} or adding
21871 @item -fbounds-check
21872 @opindex fbounds-check
21873 For front ends that support it, generate additional code to check that
21874 indices used to access arrays are within the declared range. This is
21875 currently only supported by the Java and Fortran front ends, where
21876 this option defaults to true and false respectively.
21878 @item -fstack-reuse=@var{reuse-level}
21879 @opindex fstack_reuse
21880 This option controls stack space reuse for user declared local/auto variables
21881 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
21882 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
21883 local variables and temporaries, @samp{named_vars} enables the reuse only for
21884 user defined local variables with names, and @samp{none} disables stack reuse
21885 completely. The default value is @samp{all}. The option is needed when the
21886 program extends the lifetime of a scoped local variable or a compiler generated
21887 temporary beyond the end point defined by the language. When a lifetime of
21888 a variable ends, and if the variable lives in memory, the optimizing compiler
21889 has the freedom to reuse its stack space with other temporaries or scoped
21890 local variables whose live range does not overlap with it. Legacy code extending
21891 local lifetime will likely to break with the stack reuse optimization.
21910 if (*p == 10) // out of scope use of local1
21921 A(int k) : i(k), j(k) @{ @}
21928 void foo(const A& ar)
21935 foo(A(10)); // temp object's lifetime ends when foo returns
21941 ap->i+= 10; // ap references out of scope temp whose space
21942 // is reused with a. What is the value of ap->i?
21947 The lifetime of a compiler generated temporary is well defined by the C++
21948 standard. When a lifetime of a temporary ends, and if the temporary lives
21949 in memory, the optimizing compiler has the freedom to reuse its stack
21950 space with other temporaries or scoped local variables whose live range
21951 does not overlap with it. However some of the legacy code relies on
21952 the behavior of older compilers in which temporaries' stack space is
21953 not reused, the aggressive stack reuse can lead to runtime errors. This
21954 option is used to control the temporary stack reuse optimization.
21958 This option generates traps for signed overflow on addition, subtraction,
21959 multiplication operations.
21963 This option instructs the compiler to assume that signed arithmetic
21964 overflow of addition, subtraction and multiplication wraps around
21965 using twos-complement representation. This flag enables some optimizations
21966 and disables others. This option is enabled by default for the Java
21967 front end, as required by the Java language specification.
21970 @opindex fexceptions
21971 Enable exception handling. Generates extra code needed to propagate
21972 exceptions. For some targets, this implies GCC generates frame
21973 unwind information for all functions, which can produce significant data
21974 size overhead, although it does not affect execution. If you do not
21975 specify this option, GCC enables it by default for languages like
21976 C++ that normally require exception handling, and disables it for
21977 languages like C that do not normally require it. However, you may need
21978 to enable this option when compiling C code that needs to interoperate
21979 properly with exception handlers written in C++. You may also wish to
21980 disable this option if you are compiling older C++ programs that don't
21981 use exception handling.
21983 @item -fnon-call-exceptions
21984 @opindex fnon-call-exceptions
21985 Generate code that allows trapping instructions to throw exceptions.
21986 Note that this requires platform-specific runtime support that does
21987 not exist everywhere. Moreover, it only allows @emph{trapping}
21988 instructions to throw exceptions, i.e.@: memory references or floating-point
21989 instructions. It does not allow exceptions to be thrown from
21990 arbitrary signal handlers such as @code{SIGALRM}.
21992 @item -fdelete-dead-exceptions
21993 @opindex fdelete-dead-exceptions
21994 Consider that instructions that may throw exceptions but don't otherwise
21995 contribute to the execution of the program can be optimized away.
21996 This option is enabled by default for the Ada front end, as permitted by
21997 the Ada language specification.
21998 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
22000 @item -funwind-tables
22001 @opindex funwind-tables
22002 Similar to @option{-fexceptions}, except that it just generates any needed
22003 static data, but does not affect the generated code in any other way.
22004 You normally do not need to enable this option; instead, a language processor
22005 that needs this handling enables it on your behalf.
22007 @item -fasynchronous-unwind-tables
22008 @opindex fasynchronous-unwind-tables
22009 Generate unwind table in DWARF 2 format, if supported by target machine. The
22010 table is exact at each instruction boundary, so it can be used for stack
22011 unwinding from asynchronous events (such as debugger or garbage collector).
22013 @item -fpcc-struct-return
22014 @opindex fpcc-struct-return
22015 Return ``short'' @code{struct} and @code{union} values in memory like
22016 longer ones, rather than in registers. This convention is less
22017 efficient, but it has the advantage of allowing intercallability between
22018 GCC-compiled files and files compiled with other compilers, particularly
22019 the Portable C Compiler (pcc).
22021 The precise convention for returning structures in memory depends
22022 on the target configuration macros.
22024 Short structures and unions are those whose size and alignment match
22025 that of some integer type.
22027 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
22028 switch is not binary compatible with code compiled with the
22029 @option{-freg-struct-return} switch.
22030 Use it to conform to a non-default application binary interface.
22032 @item -freg-struct-return
22033 @opindex freg-struct-return
22034 Return @code{struct} and @code{union} values in registers when possible.
22035 This is more efficient for small structures than
22036 @option{-fpcc-struct-return}.
22038 If you specify neither @option{-fpcc-struct-return} nor
22039 @option{-freg-struct-return}, GCC defaults to whichever convention is
22040 standard for the target. If there is no standard convention, GCC
22041 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
22042 the principal compiler. In those cases, we can choose the standard, and
22043 we chose the more efficient register return alternative.
22045 @strong{Warning:} code compiled with the @option{-freg-struct-return}
22046 switch is not binary compatible with code compiled with the
22047 @option{-fpcc-struct-return} switch.
22048 Use it to conform to a non-default application binary interface.
22050 @item -fshort-enums
22051 @opindex fshort-enums
22052 Allocate to an @code{enum} type only as many bytes as it needs for the
22053 declared range of possible values. Specifically, the @code{enum} type
22054 is equivalent to the smallest integer type that has enough room.
22056 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
22057 code that is not binary compatible with code generated without that switch.
22058 Use it to conform to a non-default application binary interface.
22060 @item -fshort-double
22061 @opindex fshort-double
22062 Use the same size for @code{double} as for @code{float}.
22064 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
22065 code that is not binary compatible with code generated without that switch.
22066 Use it to conform to a non-default application binary interface.
22068 @item -fshort-wchar
22069 @opindex fshort-wchar
22070 Override the underlying type for @samp{wchar_t} to be @samp{short
22071 unsigned int} instead of the default for the target. This option is
22072 useful for building programs to run under WINE@.
22074 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
22075 code that is not binary compatible with code generated without that switch.
22076 Use it to conform to a non-default application binary interface.
22079 @opindex fno-common
22080 In C code, controls the placement of uninitialized global variables.
22081 Unix C compilers have traditionally permitted multiple definitions of
22082 such variables in different compilation units by placing the variables
22084 This is the behavior specified by @option{-fcommon}, and is the default
22085 for GCC on most targets.
22086 On the other hand, this behavior is not required by ISO C, and on some
22087 targets may carry a speed or code size penalty on variable references.
22088 The @option{-fno-common} option specifies that the compiler should place
22089 uninitialized global variables in the data section of the object file,
22090 rather than generating them as common blocks.
22091 This has the effect that if the same variable is declared
22092 (without @code{extern}) in two different compilations,
22093 you get a multiple-definition error when you link them.
22094 In this case, you must compile with @option{-fcommon} instead.
22095 Compiling with @option{-fno-common} is useful on targets for which
22096 it provides better performance, or if you wish to verify that the
22097 program will work on other systems that always treat uninitialized
22098 variable declarations this way.
22102 Ignore the @samp{#ident} directive.
22104 @item -finhibit-size-directive
22105 @opindex finhibit-size-directive
22106 Don't output a @code{.size} assembler directive, or anything else that
22107 would cause trouble if the function is split in the middle, and the
22108 two halves are placed at locations far apart in memory. This option is
22109 used when compiling @file{crtstuff.c}; you should not need to use it
22112 @item -fverbose-asm
22113 @opindex fverbose-asm
22114 Put extra commentary information in the generated assembly code to
22115 make it more readable. This option is generally only of use to those
22116 who actually need to read the generated assembly code (perhaps while
22117 debugging the compiler itself).
22119 @option{-fno-verbose-asm}, the default, causes the
22120 extra information to be omitted and is useful when comparing two assembler
22123 @item -frecord-gcc-switches
22124 @opindex frecord-gcc-switches
22125 This switch causes the command line used to invoke the
22126 compiler to be recorded into the object file that is being created.
22127 This switch is only implemented on some targets and the exact format
22128 of the recording is target and binary file format dependent, but it
22129 usually takes the form of a section containing ASCII text. This
22130 switch is related to the @option{-fverbose-asm} switch, but that
22131 switch only records information in the assembler output file as
22132 comments, so it never reaches the object file.
22133 See also @option{-grecord-gcc-switches} for another
22134 way of storing compiler options into the object file.
22138 @cindex global offset table
22140 Generate position-independent code (PIC) suitable for use in a shared
22141 library, if supported for the target machine. Such code accesses all
22142 constant addresses through a global offset table (GOT)@. The dynamic
22143 loader resolves the GOT entries when the program starts (the dynamic
22144 loader is not part of GCC; it is part of the operating system). If
22145 the GOT size for the linked executable exceeds a machine-specific
22146 maximum size, you get an error message from the linker indicating that
22147 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
22148 instead. (These maximums are 8k on the SPARC and 32k
22149 on the m68k and RS/6000. The 386 has no such limit.)
22151 Position-independent code requires special support, and therefore works
22152 only on certain machines. For the 386, GCC supports PIC for System V
22153 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
22154 position-independent.
22156 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22161 If supported for the target machine, emit position-independent code,
22162 suitable for dynamic linking and avoiding any limit on the size of the
22163 global offset table. This option makes a difference on the m68k,
22164 PowerPC and SPARC@.
22166 Position-independent code requires special support, and therefore works
22167 only on certain machines.
22169 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22176 These options are similar to @option{-fpic} and @option{-fPIC}, but
22177 generated position independent code can be only linked into executables.
22178 Usually these options are used when @option{-pie} GCC option is
22179 used during linking.
22181 @option{-fpie} and @option{-fPIE} both define the macros
22182 @code{__pie__} and @code{__PIE__}. The macros have the value 1
22183 for @option{-fpie} and 2 for @option{-fPIE}.
22185 @item -fno-jump-tables
22186 @opindex fno-jump-tables
22187 Do not use jump tables for switch statements even where it would be
22188 more efficient than other code generation strategies. This option is
22189 of use in conjunction with @option{-fpic} or @option{-fPIC} for
22190 building code that forms part of a dynamic linker and cannot
22191 reference the address of a jump table. On some targets, jump tables
22192 do not require a GOT and this option is not needed.
22194 @item -ffixed-@var{reg}
22196 Treat the register named @var{reg} as a fixed register; generated code
22197 should never refer to it (except perhaps as a stack pointer, frame
22198 pointer or in some other fixed role).
22200 @var{reg} must be the name of a register. The register names accepted
22201 are machine-specific and are defined in the @code{REGISTER_NAMES}
22202 macro in the machine description macro file.
22204 This flag does not have a negative form, because it specifies a
22207 @item -fcall-used-@var{reg}
22208 @opindex fcall-used
22209 Treat the register named @var{reg} as an allocable register that is
22210 clobbered by function calls. It may be allocated for temporaries or
22211 variables that do not live across a call. Functions compiled this way
22212 do not save and restore the register @var{reg}.
22214 It is an error to use this flag with the frame pointer or stack pointer.
22215 Use of this flag for other registers that have fixed pervasive roles in
22216 the machine's execution model produces disastrous results.
22218 This flag does not have a negative form, because it specifies a
22221 @item -fcall-saved-@var{reg}
22222 @opindex fcall-saved
22223 Treat the register named @var{reg} as an allocable register saved by
22224 functions. It may be allocated even for temporaries or variables that
22225 live across a call. Functions compiled this way save and restore
22226 the register @var{reg} if they use it.
22228 It is an error to use this flag with the frame pointer or stack pointer.
22229 Use of this flag for other registers that have fixed pervasive roles in
22230 the machine's execution model produces disastrous results.
22232 A different sort of disaster results from the use of this flag for
22233 a register in which function values may be returned.
22235 This flag does not have a negative form, because it specifies a
22238 @item -fpack-struct[=@var{n}]
22239 @opindex fpack-struct
22240 Without a value specified, pack all structure members together without
22241 holes. When a value is specified (which must be a small power of two), pack
22242 structure members according to this value, representing the maximum
22243 alignment (that is, objects with default alignment requirements larger than
22244 this are output potentially unaligned at the next fitting location.
22246 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
22247 code that is not binary compatible with code generated without that switch.
22248 Additionally, it makes the code suboptimal.
22249 Use it to conform to a non-default application binary interface.
22251 @item -finstrument-functions
22252 @opindex finstrument-functions
22253 Generate instrumentation calls for entry and exit to functions. Just
22254 after function entry and just before function exit, the following
22255 profiling functions are called with the address of the current
22256 function and its call site. (On some platforms,
22257 @code{__builtin_return_address} does not work beyond the current
22258 function, so the call site information may not be available to the
22259 profiling functions otherwise.)
22262 void __cyg_profile_func_enter (void *this_fn,
22264 void __cyg_profile_func_exit (void *this_fn,
22268 The first argument is the address of the start of the current function,
22269 which may be looked up exactly in the symbol table.
22271 This instrumentation is also done for functions expanded inline in other
22272 functions. The profiling calls indicate where, conceptually, the
22273 inline function is entered and exited. This means that addressable
22274 versions of such functions must be available. If all your uses of a
22275 function are expanded inline, this may mean an additional expansion of
22276 code size. If you use @samp{extern inline} in your C code, an
22277 addressable version of such functions must be provided. (This is
22278 normally the case anyway, but if you get lucky and the optimizer always
22279 expands the functions inline, you might have gotten away without
22280 providing static copies.)
22282 A function may be given the attribute @code{no_instrument_function}, in
22283 which case this instrumentation is not done. This can be used, for
22284 example, for the profiling functions listed above, high-priority
22285 interrupt routines, and any functions from which the profiling functions
22286 cannot safely be called (perhaps signal handlers, if the profiling
22287 routines generate output or allocate memory).
22289 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
22290 @opindex finstrument-functions-exclude-file-list
22292 Set the list of functions that are excluded from instrumentation (see
22293 the description of @code{-finstrument-functions}). If the file that
22294 contains a function definition matches with one of @var{file}, then
22295 that function is not instrumented. The match is done on substrings:
22296 if the @var{file} parameter is a substring of the file name, it is
22297 considered to be a match.
22302 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
22306 excludes any inline function defined in files whose pathnames
22307 contain @code{/bits/stl} or @code{include/sys}.
22309 If, for some reason, you want to include letter @code{','} in one of
22310 @var{sym}, write @code{'\,'}. For example,
22311 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
22312 (note the single quote surrounding the option).
22314 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
22315 @opindex finstrument-functions-exclude-function-list
22317 This is similar to @code{-finstrument-functions-exclude-file-list},
22318 but this option sets the list of function names to be excluded from
22319 instrumentation. The function name to be matched is its user-visible
22320 name, such as @code{vector<int> blah(const vector<int> &)}, not the
22321 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
22322 match is done on substrings: if the @var{sym} parameter is a substring
22323 of the function name, it is considered to be a match. For C99 and C++
22324 extended identifiers, the function name must be given in UTF-8, not
22325 using universal character names.
22327 @item -fstack-check
22328 @opindex fstack-check
22329 Generate code to verify that you do not go beyond the boundary of the
22330 stack. You should specify this flag if you are running in an
22331 environment with multiple threads, but you only rarely need to specify it in
22332 a single-threaded environment since stack overflow is automatically
22333 detected on nearly all systems if there is only one stack.
22335 Note that this switch does not actually cause checking to be done; the
22336 operating system or the language runtime must do that. The switch causes
22337 generation of code to ensure that they see the stack being extended.
22339 You can additionally specify a string parameter: @code{no} means no
22340 checking, @code{generic} means force the use of old-style checking,
22341 @code{specific} means use the best checking method and is equivalent
22342 to bare @option{-fstack-check}.
22344 Old-style checking is a generic mechanism that requires no specific
22345 target support in the compiler but comes with the following drawbacks:
22349 Modified allocation strategy for large objects: they are always
22350 allocated dynamically if their size exceeds a fixed threshold.
22353 Fixed limit on the size of the static frame of functions: when it is
22354 topped by a particular function, stack checking is not reliable and
22355 a warning is issued by the compiler.
22358 Inefficiency: because of both the modified allocation strategy and the
22359 generic implementation, code performance is hampered.
22362 Note that old-style stack checking is also the fallback method for
22363 @code{specific} if no target support has been added in the compiler.
22365 @item -fstack-limit-register=@var{reg}
22366 @itemx -fstack-limit-symbol=@var{sym}
22367 @itemx -fno-stack-limit
22368 @opindex fstack-limit-register
22369 @opindex fstack-limit-symbol
22370 @opindex fno-stack-limit
22371 Generate code to ensure that the stack does not grow beyond a certain value,
22372 either the value of a register or the address of a symbol. If a larger
22373 stack is required, a signal is raised at run time. For most targets,
22374 the signal is raised before the stack overruns the boundary, so
22375 it is possible to catch the signal without taking special precautions.
22377 For instance, if the stack starts at absolute address @samp{0x80000000}
22378 and grows downwards, you can use the flags
22379 @option{-fstack-limit-symbol=__stack_limit} and
22380 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
22381 of 128KB@. Note that this may only work with the GNU linker.
22383 @item -fsplit-stack
22384 @opindex fsplit-stack
22385 Generate code to automatically split the stack before it overflows.
22386 The resulting program has a discontiguous stack which can only
22387 overflow if the program is unable to allocate any more memory. This
22388 is most useful when running threaded programs, as it is no longer
22389 necessary to calculate a good stack size to use for each thread. This
22390 is currently only implemented for the i386 and x86_64 back ends running
22393 When code compiled with @option{-fsplit-stack} calls code compiled
22394 without @option{-fsplit-stack}, there may not be much stack space
22395 available for the latter code to run. If compiling all code,
22396 including library code, with @option{-fsplit-stack} is not an option,
22397 then the linker can fix up these calls so that the code compiled
22398 without @option{-fsplit-stack} always has a large stack. Support for
22399 this is implemented in the gold linker in GNU binutils release 2.21
22402 @item -fleading-underscore
22403 @opindex fleading-underscore
22404 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
22405 change the way C symbols are represented in the object file. One use
22406 is to help link with legacy assembly code.
22408 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
22409 generate code that is not binary compatible with code generated without that
22410 switch. Use it to conform to a non-default application binary interface.
22411 Not all targets provide complete support for this switch.
22413 @item -ftls-model=@var{model}
22414 @opindex ftls-model
22415 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
22416 The @var{model} argument should be one of @code{global-dynamic},
22417 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
22418 Note that the choice is subject to optimization: the compiler may use
22419 a more efficient model for symbols not visible outside of the translation
22420 unit, or if @option{-fpic} is not given on the command line.
22422 The default without @option{-fpic} is @code{initial-exec}; with
22423 @option{-fpic} the default is @code{global-dynamic}.
22425 @item -fvisibility=@var{default|internal|hidden|protected}
22426 @opindex fvisibility
22427 Set the default ELF image symbol visibility to the specified option---all
22428 symbols are marked with this unless overridden within the code.
22429 Using this feature can very substantially improve linking and
22430 load times of shared object libraries, produce more optimized
22431 code, provide near-perfect API export and prevent symbol clashes.
22432 It is @strong{strongly} recommended that you use this in any shared objects
22435 Despite the nomenclature, @code{default} always means public; i.e.,
22436 available to be linked against from outside the shared object.
22437 @code{protected} and @code{internal} are pretty useless in real-world
22438 usage so the only other commonly used option is @code{hidden}.
22439 The default if @option{-fvisibility} isn't specified is
22440 @code{default}, i.e., make every
22441 symbol public---this causes the same behavior as previous versions of
22444 A good explanation of the benefits offered by ensuring ELF
22445 symbols have the correct visibility is given by ``How To Write
22446 Shared Libraries'' by Ulrich Drepper (which can be found at
22447 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
22448 solution made possible by this option to marking things hidden when
22449 the default is public is to make the default hidden and mark things
22450 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
22451 and @code{__attribute__ ((visibility("default")))} instead of
22452 @code{__declspec(dllexport)} you get almost identical semantics with
22453 identical syntax. This is a great boon to those working with
22454 cross-platform projects.
22456 For those adding visibility support to existing code, you may find
22457 @samp{#pragma GCC visibility} of use. This works by you enclosing
22458 the declarations you wish to set visibility for with (for example)
22459 @samp{#pragma GCC visibility push(hidden)} and
22460 @samp{#pragma GCC visibility pop}.
22461 Bear in mind that symbol visibility should be viewed @strong{as
22462 part of the API interface contract} and thus all new code should
22463 always specify visibility when it is not the default; i.e., declarations
22464 only for use within the local DSO should @strong{always} be marked explicitly
22465 as hidden as so to avoid PLT indirection overheads---making this
22466 abundantly clear also aids readability and self-documentation of the code.
22467 Note that due to ISO C++ specification requirements, @code{operator new} and
22468 @code{operator delete} must always be of default visibility.
22470 Be aware that headers from outside your project, in particular system
22471 headers and headers from any other library you use, may not be
22472 expecting to be compiled with visibility other than the default. You
22473 may need to explicitly say @samp{#pragma GCC visibility push(default)}
22474 before including any such headers.
22476 @samp{extern} declarations are not affected by @option{-fvisibility}, so
22477 a lot of code can be recompiled with @option{-fvisibility=hidden} with
22478 no modifications. However, this means that calls to @code{extern}
22479 functions with no explicit visibility use the PLT, so it is more
22480 effective to use @code{__attribute ((visibility))} and/or
22481 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
22482 declarations should be treated as hidden.
22484 Note that @option{-fvisibility} does affect C++ vague linkage
22485 entities. This means that, for instance, an exception class that is
22486 be thrown between DSOs must be explicitly marked with default
22487 visibility so that the @samp{type_info} nodes are unified between
22490 An overview of these techniques, their benefits and how to use them
22491 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
22493 @item -fstrict-volatile-bitfields
22494 @opindex fstrict-volatile-bitfields
22495 This option should be used if accesses to volatile bit-fields (or other
22496 structure fields, although the compiler usually honors those types
22497 anyway) should use a single access of the width of the
22498 field's type, aligned to a natural alignment if possible. For
22499 example, targets with memory-mapped peripheral registers might require
22500 all such accesses to be 16 bits wide; with this flag you can
22501 declare all peripheral bit-fields as @code{unsigned short} (assuming short
22502 is 16 bits on these targets) to force GCC to use 16-bit accesses
22503 instead of, perhaps, a more efficient 32-bit access.
22505 If this option is disabled, the compiler uses the most efficient
22506 instruction. In the previous example, that might be a 32-bit load
22507 instruction, even though that accesses bytes that do not contain
22508 any portion of the bit-field, or memory-mapped registers unrelated to
22509 the one being updated.
22511 In some cases, such as when the @code{packed} attribute is applied to a
22512 structure field, it may not be possible to access the field with a single
22513 read or write that is correctly aligned for the target machine. In this
22514 case GCC falls back to generating multiple accesses rather than code that
22515 will fault or truncate the result at run time.
22517 Note: Due to restrictions of the C/C++11 memory model, write accesses are
22518 not allowed to touch non bit-field members. It is therefore recommended
22519 to define all bits of the field's type as bit-field members.
22521 The default value of this option is determined by the application binary
22522 interface for the target processor.
22524 @item -fsync-libcalls
22525 @opindex fsync-libcalls
22526 This option controls whether any out-of-line instance of the @code{__sync}
22527 family of functions may be used to implement the C++11 @code{__atomic}
22528 family of functions.
22530 The default value of this option is enabled, thus the only useful form
22531 of the option is @option{-fno-sync-libcalls}. This option is used in
22532 the implementation of the @file{libatomic} runtime library.
22538 @node Environment Variables
22539 @section Environment Variables Affecting GCC
22540 @cindex environment variables
22542 @c man begin ENVIRONMENT
22543 This section describes several environment variables that affect how GCC
22544 operates. Some of them work by specifying directories or prefixes to use
22545 when searching for various kinds of files. Some are used to specify other
22546 aspects of the compilation environment.
22548 Note that you can also specify places to search using options such as
22549 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
22550 take precedence over places specified using environment variables, which
22551 in turn take precedence over those specified by the configuration of GCC@.
22552 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
22553 GNU Compiler Collection (GCC) Internals}.
22558 @c @itemx LC_COLLATE
22560 @c @itemx LC_MONETARY
22561 @c @itemx LC_NUMERIC
22566 @c @findex LC_COLLATE
22567 @findex LC_MESSAGES
22568 @c @findex LC_MONETARY
22569 @c @findex LC_NUMERIC
22573 These environment variables control the way that GCC uses
22574 localization information which allows GCC to work with different
22575 national conventions. GCC inspects the locale categories
22576 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
22577 so. These locale categories can be set to any value supported by your
22578 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
22579 Kingdom encoded in UTF-8.
22581 The @env{LC_CTYPE} environment variable specifies character
22582 classification. GCC uses it to determine the character boundaries in
22583 a string; this is needed for some multibyte encodings that contain quote
22584 and escape characters that are otherwise interpreted as a string
22587 The @env{LC_MESSAGES} environment variable specifies the language to
22588 use in diagnostic messages.
22590 If the @env{LC_ALL} environment variable is set, it overrides the value
22591 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
22592 and @env{LC_MESSAGES} default to the value of the @env{LANG}
22593 environment variable. If none of these variables are set, GCC
22594 defaults to traditional C English behavior.
22598 If @env{TMPDIR} is set, it specifies the directory to use for temporary
22599 files. GCC uses temporary files to hold the output of one stage of
22600 compilation which is to be used as input to the next stage: for example,
22601 the output of the preprocessor, which is the input to the compiler
22604 @item GCC_COMPARE_DEBUG
22605 @findex GCC_COMPARE_DEBUG
22606 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
22607 @option{-fcompare-debug} to the compiler driver. See the documentation
22608 of this option for more details.
22610 @item GCC_EXEC_PREFIX
22611 @findex GCC_EXEC_PREFIX
22612 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
22613 names of the subprograms executed by the compiler. No slash is added
22614 when this prefix is combined with the name of a subprogram, but you can
22615 specify a prefix that ends with a slash if you wish.
22617 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
22618 an appropriate prefix to use based on the pathname it is invoked with.
22620 If GCC cannot find the subprogram using the specified prefix, it
22621 tries looking in the usual places for the subprogram.
22623 The default value of @env{GCC_EXEC_PREFIX} is
22624 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
22625 the installed compiler. In many cases @var{prefix} is the value
22626 of @code{prefix} when you ran the @file{configure} script.
22628 Other prefixes specified with @option{-B} take precedence over this prefix.
22630 This prefix is also used for finding files such as @file{crt0.o} that are
22633 In addition, the prefix is used in an unusual way in finding the
22634 directories to search for header files. For each of the standard
22635 directories whose name normally begins with @samp{/usr/local/lib/gcc}
22636 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
22637 replacing that beginning with the specified prefix to produce an
22638 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
22639 @file{foo/bar} just before it searches the standard directory
22640 @file{/usr/local/lib/bar}.
22641 If a standard directory begins with the configured
22642 @var{prefix} then the value of @var{prefix} is replaced by
22643 @env{GCC_EXEC_PREFIX} when looking for header files.
22645 @item COMPILER_PATH
22646 @findex COMPILER_PATH
22647 The value of @env{COMPILER_PATH} is a colon-separated list of
22648 directories, much like @env{PATH}. GCC tries the directories thus
22649 specified when searching for subprograms, if it can't find the
22650 subprograms using @env{GCC_EXEC_PREFIX}.
22653 @findex LIBRARY_PATH
22654 The value of @env{LIBRARY_PATH} is a colon-separated list of
22655 directories, much like @env{PATH}. When configured as a native compiler,
22656 GCC tries the directories thus specified when searching for special
22657 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
22658 using GCC also uses these directories when searching for ordinary
22659 libraries for the @option{-l} option (but directories specified with
22660 @option{-L} come first).
22664 @cindex locale definition
22665 This variable is used to pass locale information to the compiler. One way in
22666 which this information is used is to determine the character set to be used
22667 when character literals, string literals and comments are parsed in C and C++.
22668 When the compiler is configured to allow multibyte characters,
22669 the following values for @env{LANG} are recognized:
22673 Recognize JIS characters.
22675 Recognize SJIS characters.
22677 Recognize EUCJP characters.
22680 If @env{LANG} is not defined, or if it has some other value, then the
22681 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
22682 recognize and translate multibyte characters.
22686 Some additional environment variables affect the behavior of the
22689 @include cppenv.texi
22693 @node Precompiled Headers
22694 @section Using Precompiled Headers
22695 @cindex precompiled headers
22696 @cindex speed of compilation
22698 Often large projects have many header files that are included in every
22699 source file. The time the compiler takes to process these header files
22700 over and over again can account for nearly all of the time required to
22701 build the project. To make builds faster, GCC allows you to
22702 @dfn{precompile} a header file.
22704 To create a precompiled header file, simply compile it as you would any
22705 other file, if necessary using the @option{-x} option to make the driver
22706 treat it as a C or C++ header file. You may want to use a
22707 tool like @command{make} to keep the precompiled header up-to-date when
22708 the headers it contains change.
22710 A precompiled header file is searched for when @code{#include} is
22711 seen in the compilation. As it searches for the included file
22712 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
22713 compiler looks for a precompiled header in each directory just before it
22714 looks for the include file in that directory. The name searched for is
22715 the name specified in the @code{#include} with @samp{.gch} appended. If
22716 the precompiled header file can't be used, it is ignored.
22718 For instance, if you have @code{#include "all.h"}, and you have
22719 @file{all.h.gch} in the same directory as @file{all.h}, then the
22720 precompiled header file is used if possible, and the original
22721 header is used otherwise.
22723 Alternatively, you might decide to put the precompiled header file in a
22724 directory and use @option{-I} to ensure that directory is searched
22725 before (or instead of) the directory containing the original header.
22726 Then, if you want to check that the precompiled header file is always
22727 used, you can put a file of the same name as the original header in this
22728 directory containing an @code{#error} command.
22730 This also works with @option{-include}. So yet another way to use
22731 precompiled headers, good for projects not designed with precompiled
22732 header files in mind, is to simply take most of the header files used by
22733 a project, include them from another header file, precompile that header
22734 file, and @option{-include} the precompiled header. If the header files
22735 have guards against multiple inclusion, they are skipped because
22736 they've already been included (in the precompiled header).
22738 If you need to precompile the same header file for different
22739 languages, targets, or compiler options, you can instead make a
22740 @emph{directory} named like @file{all.h.gch}, and put each precompiled
22741 header in the directory, perhaps using @option{-o}. It doesn't matter
22742 what you call the files in the directory; every precompiled header in
22743 the directory is considered. The first precompiled header
22744 encountered in the directory that is valid for this compilation is
22745 used; they're searched in no particular order.
22747 There are many other possibilities, limited only by your imagination,
22748 good sense, and the constraints of your build system.
22750 A precompiled header file can be used only when these conditions apply:
22754 Only one precompiled header can be used in a particular compilation.
22757 A precompiled header can't be used once the first C token is seen. You
22758 can have preprocessor directives before a precompiled header; you cannot
22759 include a precompiled header from inside another header.
22762 The precompiled header file must be produced for the same language as
22763 the current compilation. You can't use a C precompiled header for a C++
22767 The precompiled header file must have been produced by the same compiler
22768 binary as the current compilation is using.
22771 Any macros defined before the precompiled header is included must
22772 either be defined in the same way as when the precompiled header was
22773 generated, or must not affect the precompiled header, which usually
22774 means that they don't appear in the precompiled header at all.
22776 The @option{-D} option is one way to define a macro before a
22777 precompiled header is included; using a @code{#define} can also do it.
22778 There are also some options that define macros implicitly, like
22779 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
22782 @item If debugging information is output when using the precompiled
22783 header, using @option{-g} or similar, the same kind of debugging information
22784 must have been output when building the precompiled header. However,
22785 a precompiled header built using @option{-g} can be used in a compilation
22786 when no debugging information is being output.
22788 @item The same @option{-m} options must generally be used when building
22789 and using the precompiled header. @xref{Submodel Options},
22790 for any cases where this rule is relaxed.
22792 @item Each of the following options must be the same when building and using
22793 the precompiled header:
22795 @gccoptlist{-fexceptions}
22798 Some other command-line options starting with @option{-f},
22799 @option{-p}, or @option{-O} must be defined in the same way as when
22800 the precompiled header was generated. At present, it's not clear
22801 which options are safe to change and which are not; the safest choice
22802 is to use exactly the same options when generating and using the
22803 precompiled header. The following are known to be safe:
22805 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
22806 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
22807 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
22812 For all of these except the last, the compiler automatically
22813 ignores the precompiled header if the conditions aren't met. If you
22814 find an option combination that doesn't work and doesn't cause the
22815 precompiled header to be ignored, please consider filing a bug report,
22818 If you do use differing options when generating and using the
22819 precompiled header, the actual behavior is a mixture of the
22820 behavior for the options. For instance, if you use @option{-g} to
22821 generate the precompiled header but not when using it, you may or may
22822 not get debugging information for routines in the precompiled header.