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 @gol
193 -fno-weak -nostdinc++ @gol
194 -fvisibility-inlines-hidden @gol
195 -fvtable-verify=@var{std|preinit|none} @gol
196 -fvtv-counts -fvtv-debug @gol
197 -fvisibility-ms-compat @gol
198 -fext-numeric-literals @gol
199 -Wabi=@var{n} -Wconversion-null -Wctor-dtor-privacy @gol
200 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wnarrowing @gol
201 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
202 -Weffc++ -Wstrict-null-sentinel @gol
203 -Wno-non-template-friend -Wold-style-cast @gol
204 -Woverloaded-virtual -Wno-pmf-conversions @gol
207 @item Objective-C and Objective-C++ Language Options
208 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
209 Objective-C and Objective-C++ Dialects}.
210 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
211 -fgnu-runtime -fnext-runtime @gol
212 -fno-nil-receivers @gol
213 -fobjc-abi-version=@var{n} @gol
214 -fobjc-call-cxx-cdtors @gol
215 -fobjc-direct-dispatch @gol
216 -fobjc-exceptions @gol
219 -fobjc-std=objc1 @gol
220 -fno-local-ivars @gol
221 -fivar-visibility=@var{public|protected|private|package} @gol
222 -freplace-objc-classes @gol
225 -Wassign-intercept @gol
226 -Wno-protocol -Wselector @gol
227 -Wstrict-selector-match @gol
228 -Wundeclared-selector}
230 @item Language Independent Options
231 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
232 @gccoptlist{-fmessage-length=@var{n} @gol
233 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
234 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
235 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
237 @item Warning Options
238 @xref{Warning Options,,Options to Request or Suppress Warnings}.
239 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
240 -pedantic-errors @gol
241 -w -Wextra -Wall -Waddress -Waggregate-return @gol
242 -Waggressive-loop-optimizations -Warray-bounds @gol
243 -Wno-attributes -Wno-builtin-macro-redefined @gol
244 -Wc++-compat -Wc++11-compat -Wcast-align -Wcast-qual @gol
245 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
246 -Wconversion -Wcoverage-mismatch -Wdate-time -Wdelete-incomplete -Wno-cpp @gol
247 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
248 -Wdisabled-optimization -Wno-discarded-qualifiers @gol
249 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
250 -Wno-endif-labels -Werror -Werror=* @gol
251 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
252 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
253 -Wformat-security -Wformat-signedness -Wformat-y2k @gol
254 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
255 -Wignored-qualifiers -Wincompatible-pointer-types @gol
256 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
257 -Winit-self -Winline -Wno-int-conversion @gol
258 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
259 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
260 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
261 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args -Wmissing-braces @gol
262 -Wmissing-field-initializers -Wmissing-include-dirs @gol
263 -Wno-multichar -Wnonnull -Wodr -Wno-overflow -Wopenmp-simd @gol
264 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
265 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
266 -Wpointer-arith -Wno-pointer-to-int-cast @gol
267 -Wredundant-decls -Wno-return-local-addr @gol
268 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
269 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
270 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
271 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
272 -Wstrict-aliasing=n @gol -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
273 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
274 -Wmissing-format-attribute @gol
275 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
276 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
277 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
278 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
279 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
280 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
281 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
282 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
283 -Wvla -Wvolatile-register-var -Wwrite-strings -Wzero-as-null-pointer-constant}
285 @item C and Objective-C-only Warning Options
286 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
287 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
288 -Wold-style-declaration -Wold-style-definition @gol
289 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
290 -Wdeclaration-after-statement -Wpointer-sign}
292 @item Debugging Options
293 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
294 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
295 -fsanitize=@var{style} -fsanitize-recover @gol
296 -fsanitize-undefined-trap-on-error @gol
297 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
298 -fdisable-ipa-@var{pass_name} @gol
299 -fdisable-rtl-@var{pass_name} @gol
300 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
301 -fdisable-tree-@var{pass_name} @gol
302 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
303 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
304 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
305 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
306 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
308 -fdump-statistics @gol
310 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-cfg -fdump-tree-alias @gol
314 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-gimple@r{[}-raw@r{]} @gol
317 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
318 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
319 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
320 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
321 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
322 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
323 -fdump-tree-nrv -fdump-tree-vect @gol
324 -fdump-tree-sink @gol
325 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
326 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
327 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
328 -fdump-tree-vtable-verify @gol
329 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
330 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
331 -fdump-final-insns=@var{file} @gol
332 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
333 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
334 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
335 -fenable-@var{kind}-@var{pass} @gol
336 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
337 -fdebug-types-section -fmem-report-wpa @gol
338 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
340 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
341 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
342 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
343 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
344 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
345 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
346 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
347 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
348 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
349 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
350 -fdebug-prefix-map=@var{old}=@var{new} @gol
351 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
352 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
353 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
354 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
355 -print-prog-name=@var{program} -print-search-dirs -Q @gol
356 -print-sysroot -print-sysroot-headers-suffix @gol
357 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
359 @item Optimization Options
360 @xref{Optimize Options,,Options that Control Optimization}.
361 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
362 -falign-jumps[=@var{n}] @gol
363 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
364 -fassociative-math -fauto-inc-dec -fbranch-probabilities @gol
365 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
366 -fbtr-bb-exclusive -fcaller-saves @gol
367 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
368 -fcompare-elim -fcprop-registers -fcrossjumping @gol
369 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
370 -fcx-limited-range @gol
371 -fdata-sections -fdce -fdelayed-branch @gol
372 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively -fdse @gol
373 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
374 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
375 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
376 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
377 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
378 -fif-conversion2 -findirect-inlining @gol
379 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
380 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
381 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
382 -fira-algorithm=@var{algorithm} @gol
383 -fira-region=@var{region} -fira-hoist-pressure @gol
384 -fira-loop-pressure -fno-ira-share-save-slots @gol
385 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
386 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute
387 -fivopts -fkeep-inline-functions -fkeep-static-consts -flive-range-shrinkage @gol
388 -floop-block -floop-interchange -floop-strip-mine -floop-nest-optimize @gol
389 -floop-parallelize-all -flto -flto-compression-level @gol
390 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
391 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
392 -fmove-loop-invariants -fno-branch-count-reg @gol
393 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
394 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
395 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
396 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
397 -fomit-frame-pointer -foptimize-sibling-calls @gol
398 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
399 -fprefetch-loop-arrays -fprofile-report @gol
400 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
401 -fprofile-generate=@var{path} @gol
402 -fprofile-use -fprofile-use=@var{path} -fprofile-values -fprofile-reorder-functions @gol
403 -freciprocal-math -free -frename-registers -freorder-blocks @gol
404 -freorder-blocks-and-partition -freorder-functions @gol
405 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
406 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
407 -fsched-spec-load -fsched-spec-load-dangerous @gol
408 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
409 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
410 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
411 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
412 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
413 -fselective-scheduling -fselective-scheduling2 @gol
414 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
415 -fsemantic-interposition @gol
416 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
417 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt -fstack-protector @gol
418 -fstack-protector-all -fstack-protector-strong -fstrict-aliasing @gol
419 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
420 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
421 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
422 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
423 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
424 -ftree-loop-if-convert-stores -ftree-loop-im @gol
425 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
426 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
427 -ftree-loop-vectorize @gol
428 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
429 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
430 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
431 -ftree-vectorize -ftree-vrp @gol
432 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
433 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
434 -fuse-caller-save -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
435 -fweb -fwhole-program -fwpa -fuse-ld=@var{linker} -fuse-linker-plugin @gol
436 --param @var{name}=@var{value}
437 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
439 @item Preprocessor Options
440 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
441 @gccoptlist{-A@var{question}=@var{answer} @gol
442 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
443 -C -dD -dI -dM -dN @gol
444 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
445 -idirafter @var{dir} @gol
446 -include @var{file} -imacros @var{file} @gol
447 -iprefix @var{file} -iwithprefix @var{dir} @gol
448 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
449 -imultilib @var{dir} -isysroot @var{dir} @gol
450 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
451 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
452 -remap -trigraphs -undef -U@var{macro} @gol
453 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
455 @item Assembler Option
456 @xref{Assembler Options,,Passing Options to the Assembler}.
457 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
460 @xref{Link Options,,Options for Linking}.
461 @gccoptlist{@var{object-file-name} -l@var{library} @gol
462 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
463 -s -static -static-libgcc -static-libstdc++ @gol
464 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
465 -shared -shared-libgcc -symbolic @gol
466 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
467 -u @var{symbol} -z @var{keyword}}
469 @item Directory Options
470 @xref{Directory Options,,Options for Directory Search}.
471 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
472 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
473 --sysroot=@var{dir} --no-sysroot-suffix}
475 @item Machine Dependent Options
476 @xref{Submodel Options,,Hardware Models and Configurations}.
477 @c This list is ordered alphanumerically by subsection name.
478 @c Try and put the significant identifier (CPU or system) first,
479 @c so users have a clue at guessing where the ones they want will be.
481 @emph{AArch64 Options}
482 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
483 -mgeneral-regs-only @gol
484 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
486 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
487 -mtls-dialect=desc -mtls-dialect=traditional @gol
488 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
490 @emph{Adapteva Epiphany Options}
491 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
492 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
493 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
494 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
495 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
496 -msplit-vecmove-early -m1reg-@var{reg}}
499 @gccoptlist{-mbarrel-shifter @gol
500 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
501 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
502 -mea -mno-mpy -mmul32x16 -mmul64 @gol
503 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
504 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
505 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
506 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
507 -mucb-mcount -mvolatile-cache @gol
508 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
509 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
510 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
511 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
512 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
513 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
516 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
517 -mabi=@var{name} @gol
518 -mapcs-stack-check -mno-apcs-stack-check @gol
519 -mapcs-float -mno-apcs-float @gol
520 -mapcs-reentrant -mno-apcs-reentrant @gol
521 -msched-prolog -mno-sched-prolog @gol
522 -mlittle-endian -mbig-endian @gol
523 -mfloat-abi=@var{name} @gol
524 -mfp16-format=@var{name}
525 -mthumb-interwork -mno-thumb-interwork @gol
526 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
527 -mstructure-size-boundary=@var{n} @gol
528 -mabort-on-noreturn @gol
529 -mlong-calls -mno-long-calls @gol
530 -msingle-pic-base -mno-single-pic-base @gol
531 -mpic-register=@var{reg} @gol
532 -mnop-fun-dllimport @gol
533 -mpoke-function-name @gol
535 -mtpcs-frame -mtpcs-leaf-frame @gol
536 -mcaller-super-interworking -mcallee-super-interworking @gol
537 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
538 -mword-relocations @gol
539 -mfix-cortex-m3-ldrd @gol
540 -munaligned-access @gol
541 -mneon-for-64bits @gol
542 -mslow-flash-data @gol
546 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
547 -mcall-prologues -mint8 -mno-interrupts -mrelax @gol
548 -mstrict-X -mtiny-stack -Waddr-space-convert}
550 @emph{Blackfin Options}
551 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
552 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
553 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
554 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
555 -mno-id-shared-library -mshared-library-id=@var{n} @gol
556 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
557 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
558 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
562 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
563 -msim -msdata=@var{sdata-type}}
566 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
567 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
568 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
569 -mstack-align -mdata-align -mconst-align @gol
570 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
571 -melf -maout -melinux -mlinux -sim -sim2 @gol
572 -mmul-bug-workaround -mno-mul-bug-workaround}
575 @gccoptlist{-mmac @gol
576 -mcr16cplus -mcr16c @gol
577 -msim -mint32 -mbit-ops
578 -mdata-model=@var{model}}
580 @emph{Darwin Options}
581 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
582 -arch_only -bind_at_load -bundle -bundle_loader @gol
583 -client_name -compatibility_version -current_version @gol
585 -dependency-file -dylib_file -dylinker_install_name @gol
586 -dynamic -dynamiclib -exported_symbols_list @gol
587 -filelist -flat_namespace -force_cpusubtype_ALL @gol
588 -force_flat_namespace -headerpad_max_install_names @gol
590 -image_base -init -install_name -keep_private_externs @gol
591 -multi_module -multiply_defined -multiply_defined_unused @gol
592 -noall_load -no_dead_strip_inits_and_terms @gol
593 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
594 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
595 -private_bundle -read_only_relocs -sectalign @gol
596 -sectobjectsymbols -whyload -seg1addr @gol
597 -sectcreate -sectobjectsymbols -sectorder @gol
598 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
599 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
600 -segprot -segs_read_only_addr -segs_read_write_addr @gol
601 -single_module -static -sub_library -sub_umbrella @gol
602 -twolevel_namespace -umbrella -undefined @gol
603 -unexported_symbols_list -weak_reference_mismatches @gol
604 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
605 -mkernel -mone-byte-bool}
607 @emph{DEC Alpha Options}
608 @gccoptlist{-mno-fp-regs -msoft-float @gol
609 -mieee -mieee-with-inexact -mieee-conformant @gol
610 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
611 -mtrap-precision=@var{mode} -mbuild-constants @gol
612 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
613 -mbwx -mmax -mfix -mcix @gol
614 -mfloat-vax -mfloat-ieee @gol
615 -mexplicit-relocs -msmall-data -mlarge-data @gol
616 -msmall-text -mlarge-text @gol
617 -mmemory-latency=@var{time}}
620 @gccoptlist{-msmall-model -mno-lsim}
623 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
624 -mhard-float -msoft-float @gol
625 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
626 -mdouble -mno-double @gol
627 -mmedia -mno-media -mmuladd -mno-muladd @gol
628 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
629 -mlinked-fp -mlong-calls -malign-labels @gol
630 -mlibrary-pic -macc-4 -macc-8 @gol
631 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
632 -moptimize-membar -mno-optimize-membar @gol
633 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
634 -mvliw-branch -mno-vliw-branch @gol
635 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
636 -mno-nested-cond-exec -mtomcat-stats @gol
640 @emph{GNU/Linux Options}
641 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
642 -tno-android-cc -tno-android-ld}
644 @emph{H8/300 Options}
645 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
648 @gccoptlist{-march=@var{architecture-type} @gol
649 -mdisable-fpregs -mdisable-indexing @gol
650 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
651 -mfixed-range=@var{register-range} @gol
652 -mjump-in-delay -mlinker-opt -mlong-calls @gol
653 -mlong-load-store -mno-disable-fpregs @gol
654 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
655 -mno-jump-in-delay -mno-long-load-store @gol
656 -mno-portable-runtime -mno-soft-float @gol
657 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
658 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
659 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
660 -munix=@var{unix-std} -nolibdld -static -threads}
662 @emph{i386 and x86-64 Options}
663 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
664 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
665 -mfpmath=@var{unit} @gol
666 -masm=@var{dialect} -mno-fancy-math-387 @gol
667 -mno-fp-ret-in-387 -msoft-float @gol
668 -mno-wide-multiply -mrtd -malign-double @gol
669 -mpreferred-stack-boundary=@var{num} @gol
670 -mincoming-stack-boundary=@var{num} @gol
671 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
672 -mrecip -mrecip=@var{opt} @gol
673 -mvzeroupper -mprefer-avx128 @gol
674 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
675 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
676 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
677 -mclflushopt -mxsavec -mxsaves @gol
678 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
679 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mthreads @gol
680 -mno-align-stringops -minline-all-stringops @gol
681 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
682 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy}
683 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
684 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
685 -mregparm=@var{num} -msseregparm @gol
686 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
687 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
688 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
689 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
690 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
691 -msse2avx -mfentry -m8bit-idiv @gol
692 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
693 -mstack-protector-guard=@var{guard}}
695 @emph{i386 and x86-64 Windows Options}
696 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
697 -mnop-fun-dllimport -mthread @gol
698 -municode -mwin32 -mwindows -fno-set-stack-executable}
701 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
702 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
703 -mconstant-gp -mauto-pic -mfused-madd @gol
704 -minline-float-divide-min-latency @gol
705 -minline-float-divide-max-throughput @gol
706 -mno-inline-float-divide @gol
707 -minline-int-divide-min-latency @gol
708 -minline-int-divide-max-throughput @gol
709 -mno-inline-int-divide @gol
710 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
711 -mno-inline-sqrt @gol
712 -mdwarf2-asm -mearly-stop-bits @gol
713 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
714 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
715 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
716 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
717 -msched-spec-ldc -msched-spec-control-ldc @gol
718 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
719 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
720 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
721 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
724 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
725 -msign-extend-enabled -muser-enabled}
727 @emph{M32R/D Options}
728 @gccoptlist{-m32r2 -m32rx -m32r @gol
730 -malign-loops -mno-align-loops @gol
731 -missue-rate=@var{number} @gol
732 -mbranch-cost=@var{number} @gol
733 -mmodel=@var{code-size-model-type} @gol
734 -msdata=@var{sdata-type} @gol
735 -mno-flush-func -mflush-func=@var{name} @gol
736 -mno-flush-trap -mflush-trap=@var{number} @gol
740 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
742 @emph{M680x0 Options}
743 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
744 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
745 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
746 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
747 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
748 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
749 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
750 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
754 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
755 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
756 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
757 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
758 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
761 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
762 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
763 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
764 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
767 @emph{MicroBlaze Options}
768 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
769 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
770 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
771 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
772 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
775 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
776 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
777 -mips64 -mips64r2 -mips64r3 -mips64r5 @gol
778 -mips16 -mno-mips16 -mflip-mips16 @gol
779 -minterlink-compressed -mno-interlink-compressed @gol
780 -minterlink-mips16 -mno-interlink-mips16 @gol
781 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
782 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
783 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
784 -mno-float -msingle-float -mdouble-float @gol
785 -mabs=@var{mode} -mnan=@var{encoding} @gol
786 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
789 -mvirt -mno-virt @gol
791 -mmicromips -mno-micromips @gol
792 -mfpu=@var{fpu-type} @gol
793 -msmartmips -mno-smartmips @gol
794 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
795 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
796 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
797 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
798 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
799 -membedded-data -mno-embedded-data @gol
800 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
801 -mcode-readable=@var{setting} @gol
802 -msplit-addresses -mno-split-addresses @gol
803 -mexplicit-relocs -mno-explicit-relocs @gol
804 -mcheck-zero-division -mno-check-zero-division @gol
805 -mdivide-traps -mdivide-breaks @gol
806 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
807 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
808 -mfix-24k -mno-fix-24k @gol
809 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
810 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
811 -mfix-vr4120 -mno-fix-vr4120 @gol
812 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
813 -mflush-func=@var{func} -mno-flush-func @gol
814 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
815 -mfp-exceptions -mno-fp-exceptions @gol
816 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
817 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
820 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
821 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
822 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
823 -mno-base-addresses -msingle-exit -mno-single-exit}
825 @emph{MN10300 Options}
826 @gccoptlist{-mmult-bug -mno-mult-bug @gol
827 -mno-am33 -mam33 -mam33-2 -mam34 @gol
828 -mtune=@var{cpu-type} @gol
829 -mreturn-pointer-on-d0 @gol
830 -mno-crt0 -mrelax -mliw -msetlb}
833 @gccoptlist{-meb -mel -mno-crt0}
835 @emph{MSP430 Options}
836 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
840 @gccoptlist{-mbig-endian -mlittle-endian @gol
841 -mreduced-regs -mfull-regs @gol
842 -mcmov -mno-cmov @gol
843 -mperf-ext -mno-perf-ext @gol
844 -mv3push -mno-v3push @gol
845 -m16bit -mno-16bit @gol
846 -mgp-direct -mno-gp-direct @gol
847 -misr-vector-size=@var{num} @gol
848 -mcache-block-size=@var{num} @gol
849 -march=@var{arch} @gol
850 -mforce-fp-as-gp -mforbid-fp-as-gp @gol
851 -mex9 -mctor-dtor -mrelax}
853 @emph{Nios II Options}
854 @gccoptlist{-G @var{num} -mgpopt -mno-gpopt -mel -meb @gol
855 -mno-bypass-cache -mbypass-cache @gol
856 -mno-cache-volatile -mcache-volatile @gol
857 -mno-fast-sw-div -mfast-sw-div @gol
858 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
859 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
860 -mcustom-fpu-cfg=@var{name} @gol
861 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
863 @emph{PDP-11 Options}
864 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
865 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
866 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
867 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
868 -mbranch-expensive -mbranch-cheap @gol
869 -munix-asm -mdec-asm}
871 @emph{picoChip Options}
872 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
873 -msymbol-as-address -mno-inefficient-warnings}
875 @emph{PowerPC Options}
876 See RS/6000 and PowerPC Options.
879 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78 -m64bit-doubles -m32bit-doubles}
881 @emph{RS/6000 and PowerPC Options}
882 @gccoptlist{-mcpu=@var{cpu-type} @gol
883 -mtune=@var{cpu-type} @gol
884 -mcmodel=@var{code-model} @gol
886 -maltivec -mno-altivec @gol
887 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
888 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
889 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
890 -mfprnd -mno-fprnd @gol
891 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
892 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
893 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
894 -malign-power -malign-natural @gol
895 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
896 -msingle-float -mdouble-float -msimple-fpu @gol
897 -mstring -mno-string -mupdate -mno-update @gol
898 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
899 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
900 -mstrict-align -mno-strict-align -mrelocatable @gol
901 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
902 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
903 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
904 -mprioritize-restricted-insns=@var{priority} @gol
905 -msched-costly-dep=@var{dependence_type} @gol
906 -minsert-sched-nops=@var{scheme} @gol
907 -mcall-sysv -mcall-netbsd @gol
908 -maix-struct-return -msvr4-struct-return @gol
909 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
910 -mblock-move-inline-limit=@var{num} @gol
911 -misel -mno-isel @gol
912 -misel=yes -misel=no @gol
914 -mspe=yes -mspe=no @gol
916 -mgen-cell-microcode -mwarn-cell-microcode @gol
917 -mvrsave -mno-vrsave @gol
918 -mmulhw -mno-mulhw @gol
919 -mdlmzb -mno-dlmzb @gol
920 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
921 -mprototype -mno-prototype @gol
922 -msim -mmvme -mads -myellowknife -memb -msdata @gol
923 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
924 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
925 -mno-recip-precision @gol
926 -mveclibabi=@var{type} -mfriz -mno-friz @gol
927 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
928 -msave-toc-indirect -mno-save-toc-indirect @gol
929 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
930 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
931 -mquad-memory -mno-quad-memory @gol
932 -mquad-memory-atomic -mno-quad-memory-atomic @gol
933 -mcompat-align-parm -mno-compat-align-parm}
936 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
938 -mbig-endian-data -mlittle-endian-data @gol
941 -mas100-syntax -mno-as100-syntax@gol
943 -mmax-constant-size=@gol
946 -mno-warn-multiple-fast-interrupts@gol
947 -msave-acc-in-interrupts}
949 @emph{S/390 and zSeries Options}
950 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
951 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
952 -mlong-double-64 -mlong-double-128 @gol
953 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
954 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
955 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
956 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
957 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
958 -mhotpatch[=@var{halfwords}] -mno-hotpatch}
961 @gccoptlist{-meb -mel @gol
965 -mscore5 -mscore5u -mscore7 -mscore7d}
968 @gccoptlist{-m1 -m2 -m2e @gol
969 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
971 -m4-nofpu -m4-single-only -m4-single -m4 @gol
972 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
973 -m5-64media -m5-64media-nofpu @gol
974 -m5-32media -m5-32media-nofpu @gol
975 -m5-compact -m5-compact-nofpu @gol
976 -mb -ml -mdalign -mrelax @gol
977 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
978 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
979 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
980 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
981 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
982 -maccumulate-outgoing-args -minvalid-symbols @gol
983 -matomic-model=@var{atomic-model} @gol
984 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
985 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
986 -mpretend-cmove -mtas}
988 @emph{Solaris 2 Options}
989 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
993 @gccoptlist{-mcpu=@var{cpu-type} @gol
994 -mtune=@var{cpu-type} @gol
995 -mcmodel=@var{code-model} @gol
996 -mmemory-model=@var{mem-model} @gol
997 -m32 -m64 -mapp-regs -mno-app-regs @gol
998 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
999 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1000 -mhard-quad-float -msoft-quad-float @gol
1001 -mstack-bias -mno-stack-bias @gol
1002 -munaligned-doubles -mno-unaligned-doubles @gol
1003 -muser-mode -mno-user-mode @gol
1004 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1005 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1006 -mcbcond -mno-cbcond @gol
1007 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1008 -mfix-at697f -mfix-ut699}
1011 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1012 -msafe-dma -munsafe-dma @gol
1014 -msmall-mem -mlarge-mem -mstdmain @gol
1015 -mfixed-range=@var{register-range} @gol
1017 -maddress-space-conversion -mno-address-space-conversion @gol
1018 -mcache-size=@var{cache-size} @gol
1019 -matomic-updates -mno-atomic-updates}
1021 @emph{System V Options}
1022 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1024 @emph{TILE-Gx Options}
1025 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1026 -mcmodel=@var{code-model}}
1028 @emph{TILEPro Options}
1029 @gccoptlist{-mcpu=@var{cpu} -m32}
1032 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1033 -mprolog-function -mno-prolog-function -mspace @gol
1034 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1035 -mapp-regs -mno-app-regs @gol
1036 -mdisable-callt -mno-disable-callt @gol
1037 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1038 -mv850e -mv850 -mv850e3v5 @gol
1049 @gccoptlist{-mg -mgnu -munix}
1052 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1053 -mpointer-size=@var{size}}
1055 @emph{VxWorks Options}
1056 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1057 -Xbind-lazy -Xbind-now}
1059 @emph{x86-64 Options}
1060 See i386 and x86-64 Options.
1062 @emph{Xstormy16 Options}
1065 @emph{Xtensa Options}
1066 @gccoptlist{-mconst16 -mno-const16 @gol
1067 -mfused-madd -mno-fused-madd @gol
1069 -mserialize-volatile -mno-serialize-volatile @gol
1070 -mtext-section-literals -mno-text-section-literals @gol
1071 -mtarget-align -mno-target-align @gol
1072 -mlongcalls -mno-longcalls}
1074 @emph{zSeries Options}
1075 See S/390 and zSeries Options.
1077 @item Code Generation Options
1078 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1079 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1080 -ffixed-@var{reg} -fexceptions @gol
1081 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1082 -fasynchronous-unwind-tables @gol
1083 -fno-gnu-unique @gol
1084 -finhibit-size-directive -finstrument-functions @gol
1085 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1086 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1087 -fno-common -fno-ident @gol
1088 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1089 -fno-jump-tables @gol
1090 -frecord-gcc-switches @gol
1091 -freg-struct-return -fshort-enums @gol
1092 -fshort-double -fshort-wchar @gol
1093 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1094 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1095 -fno-stack-limit -fsplit-stack @gol
1096 -fleading-underscore -ftls-model=@var{model} @gol
1097 -fstack-reuse=@var{reuse_level} @gol
1098 -ftrapv -fwrapv -fbounds-check @gol
1099 -fvisibility -fstrict-volatile-bitfields -fsync-libcalls}
1103 @node Overall Options
1104 @section Options Controlling the Kind of Output
1106 Compilation can involve up to four stages: preprocessing, compilation
1107 proper, assembly and linking, always in that order. GCC is capable of
1108 preprocessing and compiling several files either into several
1109 assembler input files, or into one assembler input file; then each
1110 assembler input file produces an object file, and linking combines all
1111 the object files (those newly compiled, and those specified as input)
1112 into an executable file.
1114 @cindex file name suffix
1115 For any given input file, the file name suffix determines what kind of
1116 compilation is done:
1120 C source code that must be preprocessed.
1123 C source code that should not be preprocessed.
1126 C++ source code that should not be preprocessed.
1129 Objective-C source code. Note that you must link with the @file{libobjc}
1130 library to make an Objective-C program work.
1133 Objective-C source code that should not be preprocessed.
1137 Objective-C++ source code. Note that you must link with the @file{libobjc}
1138 library to make an Objective-C++ program work. Note that @samp{.M} refers
1139 to a literal capital M@.
1141 @item @var{file}.mii
1142 Objective-C++ source code that should not be preprocessed.
1145 C, C++, Objective-C or Objective-C++ header file to be turned into a
1146 precompiled header (default), or C, C++ header file to be turned into an
1147 Ada spec (via the @option{-fdump-ada-spec} switch).
1150 @itemx @var{file}.cp
1151 @itemx @var{file}.cxx
1152 @itemx @var{file}.cpp
1153 @itemx @var{file}.CPP
1154 @itemx @var{file}.c++
1156 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1157 the last two letters must both be literally @samp{x}. Likewise,
1158 @samp{.C} refers to a literal capital C@.
1162 Objective-C++ source code that must be preprocessed.
1164 @item @var{file}.mii
1165 Objective-C++ source code that should not be preprocessed.
1169 @itemx @var{file}.hp
1170 @itemx @var{file}.hxx
1171 @itemx @var{file}.hpp
1172 @itemx @var{file}.HPP
1173 @itemx @var{file}.h++
1174 @itemx @var{file}.tcc
1175 C++ header file to be turned into a precompiled header or Ada spec.
1178 @itemx @var{file}.for
1179 @itemx @var{file}.ftn
1180 Fixed form Fortran source code that should not be preprocessed.
1183 @itemx @var{file}.FOR
1184 @itemx @var{file}.fpp
1185 @itemx @var{file}.FPP
1186 @itemx @var{file}.FTN
1187 Fixed form Fortran source code that must be preprocessed (with the traditional
1190 @item @var{file}.f90
1191 @itemx @var{file}.f95
1192 @itemx @var{file}.f03
1193 @itemx @var{file}.f08
1194 Free form Fortran source code that should not be preprocessed.
1196 @item @var{file}.F90
1197 @itemx @var{file}.F95
1198 @itemx @var{file}.F03
1199 @itemx @var{file}.F08
1200 Free form Fortran source code that must be preprocessed (with the
1201 traditional preprocessor).
1206 @c FIXME: Descriptions of Java file types.
1212 @item @var{file}.ads
1213 Ada source code file that contains a library unit declaration (a
1214 declaration of a package, subprogram, or generic, or a generic
1215 instantiation), or a library unit renaming declaration (a package,
1216 generic, or subprogram renaming declaration). Such files are also
1219 @item @var{file}.adb
1220 Ada source code file containing a library unit body (a subprogram or
1221 package body). Such files are also called @dfn{bodies}.
1223 @c GCC also knows about some suffixes for languages not yet included:
1234 @itemx @var{file}.sx
1235 Assembler code that must be preprocessed.
1238 An object file to be fed straight into linking.
1239 Any file name with no recognized suffix is treated this way.
1243 You can specify the input language explicitly with the @option{-x} option:
1246 @item -x @var{language}
1247 Specify explicitly the @var{language} for the following input files
1248 (rather than letting the compiler choose a default based on the file
1249 name suffix). This option applies to all following input files until
1250 the next @option{-x} option. Possible values for @var{language} are:
1252 c c-header cpp-output
1253 c++ c++-header c++-cpp-output
1254 objective-c objective-c-header objective-c-cpp-output
1255 objective-c++ objective-c++-header objective-c++-cpp-output
1256 assembler assembler-with-cpp
1258 f77 f77-cpp-input f95 f95-cpp-input
1264 Turn off any specification of a language, so that subsequent files are
1265 handled according to their file name suffixes (as they are if @option{-x}
1266 has not been used at all).
1268 @item -pass-exit-codes
1269 @opindex pass-exit-codes
1270 Normally the @command{gcc} program exits with the code of 1 if any
1271 phase of the compiler returns a non-success return code. If you specify
1272 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1273 the numerically highest error produced by any phase returning an error
1274 indication. The C, C++, and Fortran front ends return 4 if an internal
1275 compiler error is encountered.
1278 If you only want some of the stages of compilation, you can use
1279 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1280 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1281 @command{gcc} is to stop. Note that some combinations (for example,
1282 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1287 Compile or assemble the source files, but do not link. The linking
1288 stage simply is not done. The ultimate output is in the form of an
1289 object file for each source file.
1291 By default, the object file name for a source file is made by replacing
1292 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1294 Unrecognized input files, not requiring compilation or assembly, are
1299 Stop after the stage of compilation proper; do not assemble. The output
1300 is in the form of an assembler code file for each non-assembler input
1303 By default, the assembler file name for a source file is made by
1304 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1306 Input files that don't require compilation are ignored.
1310 Stop after the preprocessing stage; do not run the compiler proper. The
1311 output is in the form of preprocessed source code, which is sent to the
1314 Input files that don't require preprocessing are ignored.
1316 @cindex output file option
1319 Place output in file @var{file}. This applies to whatever
1320 sort of output is being produced, whether it be an executable file,
1321 an object file, an assembler file or preprocessed C code.
1323 If @option{-o} is not specified, the default is to put an executable
1324 file in @file{a.out}, the object file for
1325 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1326 assembler file in @file{@var{source}.s}, a precompiled header file in
1327 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1332 Print (on standard error output) the commands executed to run the stages
1333 of compilation. Also print the version number of the compiler driver
1334 program and of the preprocessor and the compiler proper.
1338 Like @option{-v} except the commands are not executed and arguments
1339 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1340 This is useful for shell scripts to capture the driver-generated command lines.
1344 Use pipes rather than temporary files for communication between the
1345 various stages of compilation. This fails to work on some systems where
1346 the assembler is unable to read from a pipe; but the GNU assembler has
1351 Print (on the standard output) a description of the command-line options
1352 understood by @command{gcc}. If the @option{-v} option is also specified
1353 then @option{--help} is also passed on to the various processes
1354 invoked by @command{gcc}, so that they can display the command-line options
1355 they accept. If the @option{-Wextra} option has also been specified
1356 (prior to the @option{--help} option), then command-line options that
1357 have no documentation associated with them are also displayed.
1360 @opindex target-help
1361 Print (on the standard output) a description of target-specific command-line
1362 options for each tool. For some targets extra target-specific
1363 information may also be printed.
1365 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1366 Print (on the standard output) a description of the command-line
1367 options understood by the compiler that fit into all specified classes
1368 and qualifiers. These are the supported classes:
1371 @item @samp{optimizers}
1372 Display all of the optimization options supported by the
1375 @item @samp{warnings}
1376 Display all of the options controlling warning messages
1377 produced by the compiler.
1380 Display target-specific options. Unlike the
1381 @option{--target-help} option however, target-specific options of the
1382 linker and assembler are not displayed. This is because those
1383 tools do not currently support the extended @option{--help=} syntax.
1386 Display the values recognized by the @option{--param}
1389 @item @var{language}
1390 Display the options supported for @var{language}, where
1391 @var{language} is the name of one of the languages supported in this
1395 Display the options that are common to all languages.
1398 These are the supported qualifiers:
1401 @item @samp{undocumented}
1402 Display only those options that are undocumented.
1405 Display options taking an argument that appears after an equal
1406 sign in the same continuous piece of text, such as:
1407 @samp{--help=target}.
1409 @item @samp{separate}
1410 Display options taking an argument that appears as a separate word
1411 following the original option, such as: @samp{-o output-file}.
1414 Thus for example to display all the undocumented target-specific
1415 switches supported by the compiler, use:
1418 --help=target,undocumented
1421 The sense of a qualifier can be inverted by prefixing it with the
1422 @samp{^} character, so for example to display all binary warning
1423 options (i.e., ones that are either on or off and that do not take an
1424 argument) that have a description, use:
1427 --help=warnings,^joined,^undocumented
1430 The argument to @option{--help=} should not consist solely of inverted
1433 Combining several classes is possible, although this usually
1434 restricts the output so much that there is nothing to display. One
1435 case where it does work, however, is when one of the classes is
1436 @var{target}. For example, to display all the target-specific
1437 optimization options, use:
1440 --help=target,optimizers
1443 The @option{--help=} option can be repeated on the command line. Each
1444 successive use displays its requested class of options, skipping
1445 those that have already been displayed.
1447 If the @option{-Q} option appears on the command line before the
1448 @option{--help=} option, then the descriptive text displayed by
1449 @option{--help=} is changed. Instead of describing the displayed
1450 options, an indication is given as to whether the option is enabled,
1451 disabled or set to a specific value (assuming that the compiler
1452 knows this at the point where the @option{--help=} option is used).
1454 Here is a truncated example from the ARM port of @command{gcc}:
1457 % gcc -Q -mabi=2 --help=target -c
1458 The following options are target specific:
1460 -mabort-on-noreturn [disabled]
1464 The output is sensitive to the effects of previous command-line
1465 options, so for example it is possible to find out which optimizations
1466 are enabled at @option{-O2} by using:
1469 -Q -O2 --help=optimizers
1472 Alternatively you can discover which binary optimizations are enabled
1473 by @option{-O3} by using:
1476 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1477 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1478 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1481 @item -no-canonical-prefixes
1482 @opindex no-canonical-prefixes
1483 Do not expand any symbolic links, resolve references to @samp{/../}
1484 or @samp{/./}, or make the path absolute when generating a relative
1489 Display the version number and copyrights of the invoked GCC@.
1493 Invoke all subcommands under a wrapper program. The name of the
1494 wrapper program and its parameters are passed as a comma separated
1498 gcc -c t.c -wrapper gdb,--args
1502 This invokes all subprograms of @command{gcc} under
1503 @samp{gdb --args}, thus the invocation of @command{cc1} is
1504 @samp{gdb --args cc1 @dots{}}.
1506 @item -fplugin=@var{name}.so
1508 Load the plugin code in file @var{name}.so, assumed to be a
1509 shared object to be dlopen'd by the compiler. The base name of
1510 the shared object file is used to identify the plugin for the
1511 purposes of argument parsing (See
1512 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1513 Each plugin should define the callback functions specified in the
1516 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1517 @opindex fplugin-arg
1518 Define an argument called @var{key} with a value of @var{value}
1519 for the plugin called @var{name}.
1521 @item -fdump-ada-spec@r{[}-slim@r{]}
1522 @opindex fdump-ada-spec
1523 For C and C++ source and include files, generate corresponding Ada specs.
1524 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1525 GNAT User's Guide}, which provides detailed documentation on this feature.
1527 @item -fada-spec-parent=@var{unit}
1528 @opindex fada-spec-parent
1529 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1530 Ada specs as child units of parent @var{unit}.
1532 @item -fdump-go-spec=@var{file}
1533 @opindex fdump-go-spec
1534 For input files in any language, generate corresponding Go
1535 declarations in @var{file}. This generates Go @code{const},
1536 @code{type}, @code{var}, and @code{func} declarations which may be a
1537 useful way to start writing a Go interface to code written in some
1540 @include @value{srcdir}/../libiberty/at-file.texi
1544 @section Compiling C++ Programs
1546 @cindex suffixes for C++ source
1547 @cindex C++ source file suffixes
1548 C++ source files conventionally use one of the suffixes @samp{.C},
1549 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1550 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1551 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1552 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1553 files with these names and compiles them as C++ programs even if you
1554 call the compiler the same way as for compiling C programs (usually
1555 with the name @command{gcc}).
1559 However, the use of @command{gcc} does not add the C++ library.
1560 @command{g++} is a program that calls GCC and automatically specifies linking
1561 against the C++ library. It treats @samp{.c},
1562 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1563 files unless @option{-x} is used. This program is also useful when
1564 precompiling a C header file with a @samp{.h} extension for use in C++
1565 compilations. On many systems, @command{g++} is also installed with
1566 the name @command{c++}.
1568 @cindex invoking @command{g++}
1569 When you compile C++ programs, you may specify many of the same
1570 command-line options that you use for compiling programs in any
1571 language; or command-line options meaningful for C and related
1572 languages; or options that are meaningful only for C++ programs.
1573 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1574 explanations of options for languages related to C@.
1575 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1576 explanations of options that are meaningful only for C++ programs.
1578 @node C Dialect Options
1579 @section Options Controlling C Dialect
1580 @cindex dialect options
1581 @cindex language dialect options
1582 @cindex options, dialect
1584 The following options control the dialect of C (or languages derived
1585 from C, such as C++, Objective-C and Objective-C++) that the compiler
1589 @cindex ANSI support
1593 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1594 equivalent to @option{-std=c++98}.
1596 This turns off certain features of GCC that are incompatible with ISO
1597 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1598 such as the @code{asm} and @code{typeof} keywords, and
1599 predefined macros such as @code{unix} and @code{vax} that identify the
1600 type of system you are using. It also enables the undesirable and
1601 rarely used ISO trigraph feature. For the C compiler,
1602 it disables recognition of C++ style @samp{//} comments as well as
1603 the @code{inline} keyword.
1605 The alternate keywords @code{__asm__}, @code{__extension__},
1606 @code{__inline__} and @code{__typeof__} continue to work despite
1607 @option{-ansi}. You would not want to use them in an ISO C program, of
1608 course, but it is useful to put them in header files that might be included
1609 in compilations done with @option{-ansi}. Alternate predefined macros
1610 such as @code{__unix__} and @code{__vax__} are also available, with or
1611 without @option{-ansi}.
1613 The @option{-ansi} option does not cause non-ISO programs to be
1614 rejected gratuitously. For that, @option{-Wpedantic} is required in
1615 addition to @option{-ansi}. @xref{Warning Options}.
1617 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1618 option is used. Some header files may notice this macro and refrain
1619 from declaring certain functions or defining certain macros that the
1620 ISO standard doesn't call for; this is to avoid interfering with any
1621 programs that might use these names for other things.
1623 Functions that are normally built in but do not have semantics
1624 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1625 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1626 built-in functions provided by GCC}, for details of the functions
1631 Determine the language standard. @xref{Standards,,Language Standards
1632 Supported by GCC}, for details of these standard versions. This option
1633 is currently only supported when compiling C or C++.
1635 The compiler can accept several base standards, such as @samp{c90} or
1636 @samp{c++98}, and GNU dialects of those standards, such as
1637 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1638 compiler accepts all programs following that standard plus those
1639 using GNU extensions that do not contradict it. For example,
1640 @option{-std=c90} turns off certain features of GCC that are
1641 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1642 keywords, but not other GNU extensions that do not have a meaning in
1643 ISO C90, such as omitting the middle term of a @code{?:}
1644 expression. On the other hand, when a GNU dialect of a standard is
1645 specified, all features supported by the compiler are enabled, even when
1646 those features change the meaning of the base standard. As a result, some
1647 strict-conforming programs may be rejected. The particular standard
1648 is used by @option{-Wpedantic} to identify which features are GNU
1649 extensions given that version of the standard. For example
1650 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1651 comments, while @option{-std=gnu99 -Wpedantic} does not.
1653 A value for this option must be provided; possible values are
1659 Support all ISO C90 programs (certain GNU extensions that conflict
1660 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1662 @item iso9899:199409
1663 ISO C90 as modified in amendment 1.
1669 ISO C99. This standard is substantially completely supported, modulo
1670 bugs, extended identifiers (supported except for corner cases when
1671 @option{-fextended-identifiers} is used) and floating-point issues
1672 (mainly but not entirely relating to optional C99 features from
1673 Annexes F and G). See
1674 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1675 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1680 ISO C11, the 2011 revision of the ISO C standard. This standard is
1681 substantially completely supported, modulo bugs, extended identifiers
1682 (supported except for corner cases when
1683 @option{-fextended-identifiers} is used), floating-point issues
1684 (mainly but not entirely relating to optional C11 features from
1685 Annexes F and G) and the optional Annexes K (Bounds-checking
1686 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1690 GNU dialect of ISO C90 (including some C99 features). This
1691 is the default for C code.
1695 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1699 GNU dialect of ISO C11. This is intended to become the default in a
1700 future release of GCC. The name @samp{gnu1x} is deprecated.
1704 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1705 additional defect reports. Same as @option{-ansi} for C++ code.
1709 GNU dialect of @option{-std=c++98}. This is the default for
1714 The 2011 ISO C++ standard plus amendments.
1715 The name @samp{c++0x} is deprecated.
1719 GNU dialect of @option{-std=c++11}.
1720 The name @samp{gnu++0x} is deprecated.
1723 The next revision of the ISO C++ standard, tentatively planned for
1724 2014. Support is highly experimental, and will almost certainly
1725 change in incompatible ways in future releases.
1728 GNU dialect of @option{-std=c++1y}. Support is highly experimental,
1729 and will almost certainly change in incompatible ways in future
1733 @item -fgnu89-inline
1734 @opindex fgnu89-inline
1735 The option @option{-fgnu89-inline} tells GCC to use the traditional
1736 GNU semantics for @code{inline} functions when in C99 mode.
1737 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1738 is accepted and ignored by GCC versions 4.1.3 up to but not including
1739 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1740 C99 mode. Using this option is roughly equivalent to adding the
1741 @code{gnu_inline} function attribute to all inline functions
1742 (@pxref{Function Attributes}).
1744 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1745 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1746 specifies the default behavior). This option was first supported in
1747 GCC 4.3. This option is not supported in @option{-std=c90} or
1748 @option{-std=gnu90} mode.
1750 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1751 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1752 in effect for @code{inline} functions. @xref{Common Predefined
1753 Macros,,,cpp,The C Preprocessor}.
1755 @item -aux-info @var{filename}
1757 Output to the given filename prototyped declarations for all functions
1758 declared and/or defined in a translation unit, including those in header
1759 files. This option is silently ignored in any language other than C@.
1761 Besides declarations, the file indicates, in comments, the origin of
1762 each declaration (source file and line), whether the declaration was
1763 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1764 @samp{O} for old, respectively, in the first character after the line
1765 number and the colon), and whether it came from a declaration or a
1766 definition (@samp{C} or @samp{F}, respectively, in the following
1767 character). In the case of function definitions, a K&R-style list of
1768 arguments followed by their declarations is also provided, inside
1769 comments, after the declaration.
1771 @item -fallow-parameterless-variadic-functions
1772 @opindex fallow-parameterless-variadic-functions
1773 Accept variadic functions without named parameters.
1775 Although it is possible to define such a function, this is not very
1776 useful as it is not possible to read the arguments. This is only
1777 supported for C as this construct is allowed by C++.
1781 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1782 keyword, so that code can use these words as identifiers. You can use
1783 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1784 instead. @option{-ansi} implies @option{-fno-asm}.
1786 In C++, this switch only affects the @code{typeof} keyword, since
1787 @code{asm} and @code{inline} are standard keywords. You may want to
1788 use the @option{-fno-gnu-keywords} flag instead, which has the same
1789 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1790 switch only affects the @code{asm} and @code{typeof} keywords, since
1791 @code{inline} is a standard keyword in ISO C99.
1794 @itemx -fno-builtin-@var{function}
1795 @opindex fno-builtin
1796 @cindex built-in functions
1797 Don't recognize built-in functions that do not begin with
1798 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1799 functions provided by GCC}, for details of the functions affected,
1800 including those which are not built-in functions when @option{-ansi} or
1801 @option{-std} options for strict ISO C conformance are used because they
1802 do not have an ISO standard meaning.
1804 GCC normally generates special code to handle certain built-in functions
1805 more efficiently; for instance, calls to @code{alloca} may become single
1806 instructions which adjust the stack directly, and calls to @code{memcpy}
1807 may become inline copy loops. The resulting code is often both smaller
1808 and faster, but since the function calls no longer appear as such, you
1809 cannot set a breakpoint on those calls, nor can you change the behavior
1810 of the functions by linking with a different library. In addition,
1811 when a function is recognized as a built-in function, GCC may use
1812 information about that function to warn about problems with calls to
1813 that function, or to generate more efficient code, even if the
1814 resulting code still contains calls to that function. For example,
1815 warnings are given with @option{-Wformat} for bad calls to
1816 @code{printf} when @code{printf} is built in and @code{strlen} is
1817 known not to modify global memory.
1819 With the @option{-fno-builtin-@var{function}} option
1820 only the built-in function @var{function} is
1821 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1822 function is named that is not built-in in this version of GCC, this
1823 option is ignored. There is no corresponding
1824 @option{-fbuiltin-@var{function}} option; if you wish to enable
1825 built-in functions selectively when using @option{-fno-builtin} or
1826 @option{-ffreestanding}, you may define macros such as:
1829 #define abs(n) __builtin_abs ((n))
1830 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1835 @cindex hosted environment
1837 Assert that compilation targets a hosted environment. This implies
1838 @option{-fbuiltin}. A hosted environment is one in which the
1839 entire standard library is available, and in which @code{main} has a return
1840 type of @code{int}. Examples are nearly everything except a kernel.
1841 This is equivalent to @option{-fno-freestanding}.
1843 @item -ffreestanding
1844 @opindex ffreestanding
1845 @cindex hosted environment
1847 Assert that compilation targets a freestanding environment. This
1848 implies @option{-fno-builtin}. A freestanding environment
1849 is one in which the standard library may not exist, and program startup may
1850 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1851 This is equivalent to @option{-fno-hosted}.
1853 @xref{Standards,,Language Standards Supported by GCC}, for details of
1854 freestanding and hosted environments.
1858 @cindex OpenMP parallel
1859 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1860 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1861 compiler generates parallel code according to the OpenMP Application
1862 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1863 implies @option{-pthread}, and thus is only supported on targets that
1864 have support for @option{-pthread}. @option{-fopenmp} implies
1865 @option{-fopenmp-simd}.
1868 @opindex fopenmp-simd
1871 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1872 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1877 @cindex Enable Cilk Plus
1878 Enable the usage of Cilk Plus language extension features for C/C++.
1879 When the option @option{-fcilkplus} is specified, enable the usage of
1880 the Cilk Plus Language extension features for C/C++. The present
1881 implementation follows ABI version 1.2. This is an experimental
1882 feature that is only partially complete, and whose interface may
1883 change in future versions of GCC as the official specification
1884 changes. Currently, all features but @code{_Cilk_for} have been
1889 When the option @option{-fgnu-tm} is specified, the compiler
1890 generates code for the Linux variant of Intel's current Transactional
1891 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1892 an experimental feature whose interface may change in future versions
1893 of GCC, as the official specification changes. Please note that not
1894 all architectures are supported for this feature.
1896 For more information on GCC's support for transactional memory,
1897 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1898 Transactional Memory Library}.
1900 Note that the transactional memory feature is not supported with
1901 non-call exceptions (@option{-fnon-call-exceptions}).
1903 @item -fms-extensions
1904 @opindex fms-extensions
1905 Accept some non-standard constructs used in Microsoft header files.
1907 In C++ code, this allows member names in structures to be similar
1908 to previous types declarations.
1917 Some cases of unnamed fields in structures and unions are only
1918 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1919 fields within structs/unions}, for details.
1921 Note that this option is off for all targets but i?86 and x86_64
1922 targets using ms-abi.
1923 @item -fplan9-extensions
1924 Accept some non-standard constructs used in Plan 9 code.
1926 This enables @option{-fms-extensions}, permits passing pointers to
1927 structures with anonymous fields to functions that expect pointers to
1928 elements of the type of the field, and permits referring to anonymous
1929 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1930 struct/union fields within structs/unions}, for details. This is only
1931 supported for C, not C++.
1935 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1936 options for strict ISO C conformance) implies @option{-trigraphs}.
1938 @cindex traditional C language
1939 @cindex C language, traditional
1941 @itemx -traditional-cpp
1942 @opindex traditional-cpp
1943 @opindex traditional
1944 Formerly, these options caused GCC to attempt to emulate a pre-standard
1945 C compiler. They are now only supported with the @option{-E} switch.
1946 The preprocessor continues to support a pre-standard mode. See the GNU
1947 CPP manual for details.
1949 @item -fcond-mismatch
1950 @opindex fcond-mismatch
1951 Allow conditional expressions with mismatched types in the second and
1952 third arguments. The value of such an expression is void. This option
1953 is not supported for C++.
1955 @item -flax-vector-conversions
1956 @opindex flax-vector-conversions
1957 Allow implicit conversions between vectors with differing numbers of
1958 elements and/or incompatible element types. This option should not be
1961 @item -funsigned-char
1962 @opindex funsigned-char
1963 Let the type @code{char} be unsigned, like @code{unsigned char}.
1965 Each kind of machine has a default for what @code{char} should
1966 be. It is either like @code{unsigned char} by default or like
1967 @code{signed char} by default.
1969 Ideally, a portable program should always use @code{signed char} or
1970 @code{unsigned char} when it depends on the signedness of an object.
1971 But many programs have been written to use plain @code{char} and
1972 expect it to be signed, or expect it to be unsigned, depending on the
1973 machines they were written for. This option, and its inverse, let you
1974 make such a program work with the opposite default.
1976 The type @code{char} is always a distinct type from each of
1977 @code{signed char} or @code{unsigned char}, even though its behavior
1978 is always just like one of those two.
1981 @opindex fsigned-char
1982 Let the type @code{char} be signed, like @code{signed char}.
1984 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1985 the negative form of @option{-funsigned-char}. Likewise, the option
1986 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1988 @item -fsigned-bitfields
1989 @itemx -funsigned-bitfields
1990 @itemx -fno-signed-bitfields
1991 @itemx -fno-unsigned-bitfields
1992 @opindex fsigned-bitfields
1993 @opindex funsigned-bitfields
1994 @opindex fno-signed-bitfields
1995 @opindex fno-unsigned-bitfields
1996 These options control whether a bit-field is signed or unsigned, when the
1997 declaration does not use either @code{signed} or @code{unsigned}. By
1998 default, such a bit-field is signed, because this is consistent: the
1999 basic integer types such as @code{int} are signed types.
2002 @node C++ Dialect Options
2003 @section Options Controlling C++ Dialect
2005 @cindex compiler options, C++
2006 @cindex C++ options, command-line
2007 @cindex options, C++
2008 This section describes the command-line options that are only meaningful
2009 for C++ programs. You can also use most of the GNU compiler options
2010 regardless of what language your program is in. For example, you
2011 might compile a file @code{firstClass.C} like this:
2014 g++ -g -frepo -O -c firstClass.C
2018 In this example, only @option{-frepo} is an option meant
2019 only for C++ programs; you can use the other options with any
2020 language supported by GCC@.
2022 Here is a list of options that are @emph{only} for compiling C++ programs:
2026 @item -fabi-version=@var{n}
2027 @opindex fabi-version
2028 Use version @var{n} of the C++ ABI@. The default is version 0.
2030 Version 0 refers to the version conforming most closely to
2031 the C++ ABI specification. Therefore, the ABI obtained using version 0
2032 will change in different versions of G++ as ABI bugs are fixed.
2034 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2036 Version 2 is the version of the C++ ABI that first appeared in G++
2037 3.4, and was the default through G++ 4.9.
2039 Version 3 corrects an error in mangling a constant address as a
2042 Version 4, which first appeared in G++ 4.5, implements a standard
2043 mangling for vector types.
2045 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2046 attribute const/volatile on function pointer types, decltype of a
2047 plain decl, and use of a function parameter in the declaration of
2050 Version 6, which first appeared in G++ 4.7, corrects the promotion
2051 behavior of C++11 scoped enums and the mangling of template argument
2052 packs, const/static_cast, prefix ++ and --, and a class scope function
2053 used as a template argument.
2055 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2056 builtin type and corrects the mangling of lambdas in default argument
2059 Version 8, which first appeared in G++ 4.9, corrects the substitution
2060 behavior of function types with function-cv-qualifiers.
2062 See also @option{-Wabi}.
2064 @item -fabi-compat-version=@var{n}
2065 @opindex fabi-compat-version
2066 Starting with GCC 4.5, on targets that support strong aliases, G++
2067 works around mangling changes by creating an alias with the correct
2068 mangled name when defining a symbol with an incorrect mangled name.
2069 This switch specifies which ABI version to use for the alias.
2071 With @option{-fabi-version=0} (the default), this defaults to 2. If
2072 another ABI version is explicitly selected, this defaults to 0.
2074 The compatibility version is also set by @option{-Wabi=@var{n}}.
2076 @item -fno-access-control
2077 @opindex fno-access-control
2078 Turn off all access checking. This switch is mainly useful for working
2079 around bugs in the access control code.
2083 Check that the pointer returned by @code{operator new} is non-null
2084 before attempting to modify the storage allocated. This check is
2085 normally unnecessary because the C++ standard specifies that
2086 @code{operator new} only returns @code{0} if it is declared
2087 @samp{throw()}, in which case the compiler always checks the
2088 return value even without this option. In all other cases, when
2089 @code{operator new} has a non-empty exception specification, memory
2090 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2091 @samp{new (nothrow)}.
2093 @item -fconstexpr-depth=@var{n}
2094 @opindex fconstexpr-depth
2095 Set the maximum nested evaluation depth for C++11 constexpr functions
2096 to @var{n}. A limit is needed to detect endless recursion during
2097 constant expression evaluation. The minimum specified by the standard
2100 @item -fdeduce-init-list
2101 @opindex fdeduce-init-list
2102 Enable deduction of a template type parameter as
2103 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2106 template <class T> auto forward(T t) -> decltype (realfn (t))
2113 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2117 This deduction was implemented as a possible extension to the
2118 originally proposed semantics for the C++11 standard, but was not part
2119 of the final standard, so it is disabled by default. This option is
2120 deprecated, and may be removed in a future version of G++.
2122 @item -ffriend-injection
2123 @opindex ffriend-injection
2124 Inject friend functions into the enclosing namespace, so that they are
2125 visible outside the scope of the class in which they are declared.
2126 Friend functions were documented to work this way in the old Annotated
2127 C++ Reference Manual, and versions of G++ before 4.1 always worked
2128 that way. However, in ISO C++ a friend function that is not declared
2129 in an enclosing scope can only be found using argument dependent
2130 lookup. This option causes friends to be injected as they were in
2133 This option is for compatibility, and may be removed in a future
2136 @item -fno-elide-constructors
2137 @opindex fno-elide-constructors
2138 The C++ standard allows an implementation to omit creating a temporary
2139 that is only used to initialize another object of the same type.
2140 Specifying this option disables that optimization, and forces G++ to
2141 call the copy constructor in all cases.
2143 @item -fno-enforce-eh-specs
2144 @opindex fno-enforce-eh-specs
2145 Don't generate code to check for violation of exception specifications
2146 at run time. This option violates the C++ standard, but may be useful
2147 for reducing code size in production builds, much like defining
2148 @samp{NDEBUG}. This does not give user code permission to throw
2149 exceptions in violation of the exception specifications; the compiler
2150 still optimizes based on the specifications, so throwing an
2151 unexpected exception results in undefined behavior at run time.
2153 @item -fextern-tls-init
2154 @itemx -fno-extern-tls-init
2155 @opindex fextern-tls-init
2156 @opindex fno-extern-tls-init
2157 The C++11 and OpenMP standards allow @samp{thread_local} and
2158 @samp{threadprivate} variables to have dynamic (runtime)
2159 initialization. To support this, any use of such a variable goes
2160 through a wrapper function that performs any necessary initialization.
2161 When the use and definition of the variable are in the same
2162 translation unit, this overhead can be optimized away, but when the
2163 use is in a different translation unit there is significant overhead
2164 even if the variable doesn't actually need dynamic initialization. If
2165 the programmer can be sure that no use of the variable in a
2166 non-defining TU needs to trigger dynamic initialization (either
2167 because the variable is statically initialized, or a use of the
2168 variable in the defining TU will be executed before any uses in
2169 another TU), they can avoid this overhead with the
2170 @option{-fno-extern-tls-init} option.
2172 On targets that support symbol aliases, the default is
2173 @option{-fextern-tls-init}. On targets that do not support symbol
2174 aliases, the default is @option{-fno-extern-tls-init}.
2177 @itemx -fno-for-scope
2179 @opindex fno-for-scope
2180 If @option{-ffor-scope} is specified, the scope of variables declared in
2181 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2182 as specified by the C++ standard.
2183 If @option{-fno-for-scope} is specified, the scope of variables declared in
2184 a @i{for-init-statement} extends to the end of the enclosing scope,
2185 as was the case in old versions of G++, and other (traditional)
2186 implementations of C++.
2188 If neither flag is given, the default is to follow the standard,
2189 but to allow and give a warning for old-style code that would
2190 otherwise be invalid, or have different behavior.
2192 @item -fno-gnu-keywords
2193 @opindex fno-gnu-keywords
2194 Do not recognize @code{typeof} as a keyword, so that code can use this
2195 word as an identifier. You can use the keyword @code{__typeof__} instead.
2196 @option{-ansi} implies @option{-fno-gnu-keywords}.
2198 @item -fno-implicit-templates
2199 @opindex fno-implicit-templates
2200 Never emit code for non-inline templates that are instantiated
2201 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2202 @xref{Template Instantiation}, for more information.
2204 @item -fno-implicit-inline-templates
2205 @opindex fno-implicit-inline-templates
2206 Don't emit code for implicit instantiations of inline templates, either.
2207 The default is to handle inlines differently so that compiles with and
2208 without optimization need the same set of explicit instantiations.
2210 @item -fno-implement-inlines
2211 @opindex fno-implement-inlines
2212 To save space, do not emit out-of-line copies of inline functions
2213 controlled by @samp{#pragma implementation}. This causes linker
2214 errors if these functions are not inlined everywhere they are called.
2216 @item -fms-extensions
2217 @opindex fms-extensions
2218 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2219 int and getting a pointer to member function via non-standard syntax.
2221 @item -fno-nonansi-builtins
2222 @opindex fno-nonansi-builtins
2223 Disable built-in declarations of functions that are not mandated by
2224 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2225 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2228 @opindex fnothrow-opt
2229 Treat a @code{throw()} exception specification as if it were a
2230 @code{noexcept} specification to reduce or eliminate the text size
2231 overhead relative to a function with no exception specification. If
2232 the function has local variables of types with non-trivial
2233 destructors, the exception specification actually makes the
2234 function smaller because the EH cleanups for those variables can be
2235 optimized away. The semantic effect is that an exception thrown out of
2236 a function with such an exception specification results in a call
2237 to @code{terminate} rather than @code{unexpected}.
2239 @item -fno-operator-names
2240 @opindex fno-operator-names
2241 Do not treat the operator name keywords @code{and}, @code{bitand},
2242 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2243 synonyms as keywords.
2245 @item -fno-optional-diags
2246 @opindex fno-optional-diags
2247 Disable diagnostics that the standard says a compiler does not need to
2248 issue. Currently, the only such diagnostic issued by G++ is the one for
2249 a name having multiple meanings within a class.
2252 @opindex fpermissive
2253 Downgrade some diagnostics about nonconformant code from errors to
2254 warnings. Thus, using @option{-fpermissive} allows some
2255 nonconforming code to compile.
2257 @item -fno-pretty-templates
2258 @opindex fno-pretty-templates
2259 When an error message refers to a specialization of a function
2260 template, the compiler normally prints the signature of the
2261 template followed by the template arguments and any typedefs or
2262 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2263 rather than @code{void f(int)}) so that it's clear which template is
2264 involved. When an error message refers to a specialization of a class
2265 template, the compiler omits any template arguments that match
2266 the default template arguments for that template. If either of these
2267 behaviors make it harder to understand the error message rather than
2268 easier, you can use @option{-fno-pretty-templates} to disable them.
2272 Enable automatic template instantiation at link time. This option also
2273 implies @option{-fno-implicit-templates}. @xref{Template
2274 Instantiation}, for more information.
2278 Disable generation of information about every class with virtual
2279 functions for use by the C++ run-time type identification features
2280 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2281 of the language, you can save some space by using this flag. Note that
2282 exception handling uses the same information, but G++ generates it as
2283 needed. The @samp{dynamic_cast} operator can still be used for casts that
2284 do not require run-time type information, i.e.@: casts to @code{void *} or to
2285 unambiguous base classes.
2289 Emit statistics about front-end processing at the end of the compilation.
2290 This information is generally only useful to the G++ development team.
2292 @item -fstrict-enums
2293 @opindex fstrict-enums
2294 Allow the compiler to optimize using the assumption that a value of
2295 enumerated type can only be one of the values of the enumeration (as
2296 defined in the C++ standard; basically, a value that can be
2297 represented in the minimum number of bits needed to represent all the
2298 enumerators). This assumption may not be valid if the program uses a
2299 cast to convert an arbitrary integer value to the enumerated type.
2301 @item -ftemplate-backtrace-limit=@var{n}
2302 @opindex ftemplate-backtrace-limit
2303 Set the maximum number of template instantiation notes for a single
2304 warning or error to @var{n}. The default value is 10.
2306 @item -ftemplate-depth=@var{n}
2307 @opindex ftemplate-depth
2308 Set the maximum instantiation depth for template classes to @var{n}.
2309 A limit on the template instantiation depth is needed to detect
2310 endless recursions during template class instantiation. ANSI/ISO C++
2311 conforming programs must not rely on a maximum depth greater than 17
2312 (changed to 1024 in C++11). The default value is 900, as the compiler
2313 can run out of stack space before hitting 1024 in some situations.
2315 @item -fno-threadsafe-statics
2316 @opindex fno-threadsafe-statics
2317 Do not emit the extra code to use the routines specified in the C++
2318 ABI for thread-safe initialization of local statics. You can use this
2319 option to reduce code size slightly in code that doesn't need to be
2322 @item -fuse-cxa-atexit
2323 @opindex fuse-cxa-atexit
2324 Register destructors for objects with static storage duration with the
2325 @code{__cxa_atexit} function rather than the @code{atexit} function.
2326 This option is required for fully standards-compliant handling of static
2327 destructors, but only works if your C library supports
2328 @code{__cxa_atexit}.
2330 @item -fno-use-cxa-get-exception-ptr
2331 @opindex fno-use-cxa-get-exception-ptr
2332 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2333 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2334 if the runtime routine is not available.
2336 @item -fvisibility-inlines-hidden
2337 @opindex fvisibility-inlines-hidden
2338 This switch declares that the user does not attempt to compare
2339 pointers to inline functions or methods where the addresses of the two functions
2340 are taken in different shared objects.
2342 The effect of this is that GCC may, effectively, mark inline methods with
2343 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2344 appear in the export table of a DSO and do not require a PLT indirection
2345 when used within the DSO@. Enabling this option can have a dramatic effect
2346 on load and link times of a DSO as it massively reduces the size of the
2347 dynamic export table when the library makes heavy use of templates.
2349 The behavior of this switch is not quite the same as marking the
2350 methods as hidden directly, because it does not affect static variables
2351 local to the function or cause the compiler to deduce that
2352 the function is defined in only one shared object.
2354 You may mark a method as having a visibility explicitly to negate the
2355 effect of the switch for that method. For example, if you do want to
2356 compare pointers to a particular inline method, you might mark it as
2357 having default visibility. Marking the enclosing class with explicit
2358 visibility has no effect.
2360 Explicitly instantiated inline methods are unaffected by this option
2361 as their linkage might otherwise cross a shared library boundary.
2362 @xref{Template Instantiation}.
2364 @item -fvisibility-ms-compat
2365 @opindex fvisibility-ms-compat
2366 This flag attempts to use visibility settings to make GCC's C++
2367 linkage model compatible with that of Microsoft Visual Studio.
2369 The flag makes these changes to GCC's linkage model:
2373 It sets the default visibility to @code{hidden}, like
2374 @option{-fvisibility=hidden}.
2377 Types, but not their members, are not hidden by default.
2380 The One Definition Rule is relaxed for types without explicit
2381 visibility specifications that are defined in more than one
2382 shared object: those declarations are permitted if they are
2383 permitted when this option is not used.
2386 In new code it is better to use @option{-fvisibility=hidden} and
2387 export those classes that are intended to be externally visible.
2388 Unfortunately it is possible for code to rely, perhaps accidentally,
2389 on the Visual Studio behavior.
2391 Among the consequences of these changes are that static data members
2392 of the same type with the same name but defined in different shared
2393 objects are different, so changing one does not change the other;
2394 and that pointers to function members defined in different shared
2395 objects may not compare equal. When this flag is given, it is a
2396 violation of the ODR to define types with the same name differently.
2398 @item -fvtable-verify=@var{std|preinit|none}
2399 @opindex fvtable-verify
2400 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2401 feature that verifies at runtime, for every virtual call that is made, that
2402 the vtable pointer through which the call is made is valid for the type of
2403 the object, and has not been corrupted or overwritten. If an invalid vtable
2404 pointer is detected (at runtime), an error is reported and execution of the
2405 program is immediately halted.
2407 This option causes runtime data structures to be built, at program start up,
2408 for verifying the vtable pointers. The options @code{std} and @code{preinit}
2409 control the timing of when these data structures are built. In both cases the
2410 data structures are built before execution reaches 'main'. The
2411 @option{-fvtable-verify=std} causes these data structure to be built after the
2412 shared libraries have been loaded and initialized.
2413 @option{-fvtable-verify=preinit} causes them to be built before the shared
2414 libraries have been loaded and initialized.
2416 If this option appears multiple times in the compiler line, with different
2417 values specified, 'none' will take highest priority over both 'std' and
2418 'preinit'; 'preinit' will take priority over 'std'.
2421 @opindex (fvtv-debug)
2422 Causes debug versions of the runtime functions for the vtable verification
2423 feature to be called. This assumes the @option{-fvtable-verify=std} or
2424 @option{-fvtable-verify=preinit} has been used. This flag will also cause the
2425 compiler to keep track of which vtable pointers it found for each class, and
2426 record that information in the file ``vtv_set_ptr_data.log'', in the dump
2427 file directory on the user's machine.
2429 Note: This feature APPENDS data to the log file. If you want a fresh log
2430 file, be sure to delete any existing one.
2433 @opindex fvtv-counts
2434 This is a debugging flag. When used in conjunction with
2435 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2436 causes the compiler to keep track of the total number of virtual calls
2437 it encountered and the number of verifications it inserted. It also
2438 counts the number of calls to certain runtime library functions
2439 that it inserts. This information, for each compilation unit, is written
2440 to a file named ``vtv_count_data.log'', in the dump_file directory on
2441 the user's machine. It also counts the size of the vtable pointer sets
2442 for each class, and writes this information to ``vtv_class_set_sizes.log''
2443 in the same directory.
2445 Note: This feature APPENDS data to the log files. To get a fresh log
2446 files, be sure to delete any existing ones.
2450 Do not use weak symbol support, even if it is provided by the linker.
2451 By default, G++ uses weak symbols if they are available. This
2452 option exists only for testing, and should not be used by end-users;
2453 it results in inferior code and has no benefits. This option may
2454 be removed in a future release of G++.
2458 Do not search for header files in the standard directories specific to
2459 C++, but do still search the other standard directories. (This option
2460 is used when building the C++ library.)
2463 In addition, these optimization, warning, and code generation options
2464 have meanings only for C++ programs:
2467 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2470 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2471 G++ to warn when it generates code that is probably not compatible with the
2472 vendor-neutral C++ ABI@. Since G++ now defaults to
2473 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2474 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2475 or an older compatibility version is selected (with
2476 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2478 Although an effort has been made to warn about
2479 all such cases, there are probably some cases that are not warned about,
2480 even though G++ is generating incompatible code. There may also be
2481 cases where warnings are emitted even though the code that is generated
2484 You should rewrite your code to avoid these warnings if you are
2485 concerned about the fact that code generated by G++ may not be binary
2486 compatible with code generated by other compilers.
2488 @option{-Wabi} can also be used with an explicit version number to
2489 warn about compatibility with a particular @option{-fabi-version}
2490 level, e.g. @option{-Wabi=2} to warn about changes relative to
2491 @option{-fabi-version=2}. Specifying a version number also sets
2492 @option{-fabi-compat-version=@var{n}}.
2494 The known incompatibilities in @option{-fabi-version=2} (which was the
2495 default from GCC 3.4 to 4.9) include:
2500 A template with a non-type template parameter of reference type was
2501 mangled incorrectly:
2504 template <int &> struct S @{@};
2508 This was fixed in @option{-fabi-version=3}.
2511 SIMD vector types declared using @code{__attribute ((vector_size))} were
2512 mangled in a non-standard way that does not allow for overloading of
2513 functions taking vectors of different sizes.
2515 The mangling was changed in @option{-fabi-version=4}.
2518 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2519 qualifiers, and @code{decltype} of a plain declaration was folded away.
2521 These mangling issues were fixed in @option{-fabi-version=5}.
2524 Scoped enumerators passed as arguments to a variadic function are
2525 promoted like unscoped enumerators, causing @samp{va_arg} to complain.
2526 On most targets this does not actually affect the parameter passing
2527 ABI, as there is no way to pass an argument smaller than @samp{int}.
2529 Also, the ABI changed the mangling of template argument packs,
2530 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2531 a class scope function used as a template argument.
2533 These issues were corrected in @option{-fabi-version=6}.
2536 Lambdas in default argument scope were mangled incorrectly, and the
2537 ABI changed the mangling of nullptr_t.
2539 These issues were corrected in @option{-fabi-version=7}.
2542 When mangling a function type with function-cv-qualifiers, the
2543 un-qualified function type was incorrectly treated as a substitution
2546 This was fixed in @option{-fabi-version=8}.
2549 It also warns about psABI-related changes. The known psABI changes at this
2555 For SysV/x86-64, unions with @code{long double} members are
2556 passed in memory as specified in psABI. For example:
2566 @code{union U} is always passed in memory.
2570 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2571 @opindex Wctor-dtor-privacy
2572 @opindex Wno-ctor-dtor-privacy
2573 Warn when a class seems unusable because all the constructors or
2574 destructors in that class are private, and it has neither friends nor
2575 public static member functions. Also warn if there are no non-private
2576 methods, and there's at least one private member function that isn't
2577 a constructor or destructor.
2579 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2580 @opindex Wdelete-non-virtual-dtor
2581 @opindex Wno-delete-non-virtual-dtor
2582 Warn when @samp{delete} is used to destroy an instance of a class that
2583 has virtual functions and non-virtual destructor. It is unsafe to delete
2584 an instance of a derived class through a pointer to a base class if the
2585 base class does not have a virtual destructor. This warning is enabled
2588 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2589 @opindex Wliteral-suffix
2590 @opindex Wno-literal-suffix
2591 Warn when a string or character literal is followed by a ud-suffix which does
2592 not begin with an underscore. As a conforming extension, GCC treats such
2593 suffixes as separate preprocessing tokens in order to maintain backwards
2594 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2598 #define __STDC_FORMAT_MACROS
2599 #include <inttypes.h>
2604 printf("My int64: %"PRId64"\n", i64);
2608 In this case, @code{PRId64} is treated as a separate preprocessing token.
2610 This warning is enabled by default.
2612 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2614 @opindex Wno-narrowing
2615 Warn when a narrowing conversion prohibited by C++11 occurs within
2619 int i = @{ 2.2 @}; // error: narrowing from double to int
2622 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2624 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2625 required by the standard. Note that this does not affect the meaning
2626 of well-formed code; narrowing conversions are still considered
2627 ill-formed in SFINAE context.
2629 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2631 @opindex Wno-noexcept
2632 Warn when a noexcept-expression evaluates to false because of a call
2633 to a function that does not have a non-throwing exception
2634 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2635 the compiler to never throw an exception.
2637 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2638 @opindex Wnon-virtual-dtor
2639 @opindex Wno-non-virtual-dtor
2640 Warn when a class has virtual functions and an accessible non-virtual
2641 destructor itself or in an accessible polymorphic base class, in which
2642 case it is possible but unsafe to delete an instance of a derived
2643 class through a pointer to the class itself or base class. This
2644 warning is automatically enabled if @option{-Weffc++} is specified.
2646 @item -Wreorder @r{(C++ and Objective-C++ only)}
2648 @opindex Wno-reorder
2649 @cindex reordering, warning
2650 @cindex warning for reordering of member initializers
2651 Warn when the order of member initializers given in the code does not
2652 match the order in which they must be executed. For instance:
2658 A(): j (0), i (1) @{ @}
2663 The compiler rearranges the member initializers for @samp{i}
2664 and @samp{j} to match the declaration order of the members, emitting
2665 a warning to that effect. This warning is enabled by @option{-Wall}.
2667 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2668 @opindex fext-numeric-literals
2669 @opindex fno-ext-numeric-literals
2670 Accept imaginary, fixed-point, or machine-defined
2671 literal number suffixes as GNU extensions.
2672 When this option is turned off these suffixes are treated
2673 as C++11 user-defined literal numeric suffixes.
2674 This is on by default for all pre-C++11 dialects and all GNU dialects:
2675 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2676 @option{-std=gnu++1y}.
2677 This option is off by default
2678 for ISO C++11 onwards (@option{-std=c++11}, ...).
2681 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2684 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2687 Warn about violations of the following style guidelines from Scott Meyers'
2688 @cite{Effective C++} series of books:
2692 Define a copy constructor and an assignment operator for classes
2693 with dynamically-allocated memory.
2696 Prefer initialization to assignment in constructors.
2699 Have @code{operator=} return a reference to @code{*this}.
2702 Don't try to return a reference when you must return an object.
2705 Distinguish between prefix and postfix forms of increment and
2706 decrement operators.
2709 Never overload @code{&&}, @code{||}, or @code{,}.
2713 This option also enables @option{-Wnon-virtual-dtor}, which is also
2714 one of the effective C++ recommendations. However, the check is
2715 extended to warn about the lack of virtual destructor in accessible
2716 non-polymorphic bases classes too.
2718 When selecting this option, be aware that the standard library
2719 headers do not obey all of these guidelines; use @samp{grep -v}
2720 to filter out those warnings.
2722 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2723 @opindex Wstrict-null-sentinel
2724 @opindex Wno-strict-null-sentinel
2725 Warn about the use of an uncasted @code{NULL} as sentinel. When
2726 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2727 to @code{__null}. Although it is a null pointer constant rather than a
2728 null pointer, it is guaranteed to be of the same size as a pointer.
2729 But this use is not portable across different compilers.
2731 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2732 @opindex Wno-non-template-friend
2733 @opindex Wnon-template-friend
2734 Disable warnings when non-templatized friend functions are declared
2735 within a template. Since the advent of explicit template specification
2736 support in G++, if the name of the friend is an unqualified-id (i.e.,
2737 @samp{friend foo(int)}), the C++ language specification demands that the
2738 friend declare or define an ordinary, nontemplate function. (Section
2739 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2740 could be interpreted as a particular specialization of a templatized
2741 function. Because this non-conforming behavior is no longer the default
2742 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2743 check existing code for potential trouble spots and is on by default.
2744 This new compiler behavior can be turned off with
2745 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2746 but disables the helpful warning.
2748 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2749 @opindex Wold-style-cast
2750 @opindex Wno-old-style-cast
2751 Warn if an old-style (C-style) cast to a non-void type is used within
2752 a C++ program. The new-style casts (@samp{dynamic_cast},
2753 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2754 less vulnerable to unintended effects and much easier to search for.
2756 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2757 @opindex Woverloaded-virtual
2758 @opindex Wno-overloaded-virtual
2759 @cindex overloaded virtual function, warning
2760 @cindex warning for overloaded virtual function
2761 Warn when a function declaration hides virtual functions from a
2762 base class. For example, in:
2769 struct B: public A @{
2774 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2785 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2786 @opindex Wno-pmf-conversions
2787 @opindex Wpmf-conversions
2788 Disable the diagnostic for converting a bound pointer to member function
2791 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2792 @opindex Wsign-promo
2793 @opindex Wno-sign-promo
2794 Warn when overload resolution chooses a promotion from unsigned or
2795 enumerated type to a signed type, over a conversion to an unsigned type of
2796 the same size. Previous versions of G++ tried to preserve
2797 unsignedness, but the standard mandates the current behavior.
2800 @node Objective-C and Objective-C++ Dialect Options
2801 @section Options Controlling Objective-C and Objective-C++ Dialects
2803 @cindex compiler options, Objective-C and Objective-C++
2804 @cindex Objective-C and Objective-C++ options, command-line
2805 @cindex options, Objective-C and Objective-C++
2806 (NOTE: This manual does not describe the Objective-C and Objective-C++
2807 languages themselves. @xref{Standards,,Language Standards
2808 Supported by GCC}, for references.)
2810 This section describes the command-line options that are only meaningful
2811 for Objective-C and Objective-C++ programs. You can also use most of
2812 the language-independent GNU compiler options.
2813 For example, you might compile a file @code{some_class.m} like this:
2816 gcc -g -fgnu-runtime -O -c some_class.m
2820 In this example, @option{-fgnu-runtime} is an option meant only for
2821 Objective-C and Objective-C++ programs; you can use the other options with
2822 any language supported by GCC@.
2824 Note that since Objective-C is an extension of the C language, Objective-C
2825 compilations may also use options specific to the C front-end (e.g.,
2826 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2827 C++-specific options (e.g., @option{-Wabi}).
2829 Here is a list of options that are @emph{only} for compiling Objective-C
2830 and Objective-C++ programs:
2833 @item -fconstant-string-class=@var{class-name}
2834 @opindex fconstant-string-class
2835 Use @var{class-name} as the name of the class to instantiate for each
2836 literal string specified with the syntax @code{@@"@dots{}"}. The default
2837 class name is @code{NXConstantString} if the GNU runtime is being used, and
2838 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2839 @option{-fconstant-cfstrings} option, if also present, overrides the
2840 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2841 to be laid out as constant CoreFoundation strings.
2844 @opindex fgnu-runtime
2845 Generate object code compatible with the standard GNU Objective-C
2846 runtime. This is the default for most types of systems.
2848 @item -fnext-runtime
2849 @opindex fnext-runtime
2850 Generate output compatible with the NeXT runtime. This is the default
2851 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2852 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2855 @item -fno-nil-receivers
2856 @opindex fno-nil-receivers
2857 Assume that all Objective-C message dispatches (@code{[receiver
2858 message:arg]}) in this translation unit ensure that the receiver is
2859 not @code{nil}. This allows for more efficient entry points in the
2860 runtime to be used. This option is only available in conjunction with
2861 the NeXT runtime and ABI version 0 or 1.
2863 @item -fobjc-abi-version=@var{n}
2864 @opindex fobjc-abi-version
2865 Use version @var{n} of the Objective-C ABI for the selected runtime.
2866 This option is currently supported only for the NeXT runtime. In that
2867 case, Version 0 is the traditional (32-bit) ABI without support for
2868 properties and other Objective-C 2.0 additions. Version 1 is the
2869 traditional (32-bit) ABI with support for properties and other
2870 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2871 nothing is specified, the default is Version 0 on 32-bit target
2872 machines, and Version 2 on 64-bit target machines.
2874 @item -fobjc-call-cxx-cdtors
2875 @opindex fobjc-call-cxx-cdtors
2876 For each Objective-C class, check if any of its instance variables is a
2877 C++ object with a non-trivial default constructor. If so, synthesize a
2878 special @code{- (id) .cxx_construct} instance method which runs
2879 non-trivial default constructors on any such instance variables, in order,
2880 and then return @code{self}. Similarly, check if any instance variable
2881 is a C++ object with a non-trivial destructor, and if so, synthesize a
2882 special @code{- (void) .cxx_destruct} method which runs
2883 all such default destructors, in reverse order.
2885 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2886 methods thusly generated only operate on instance variables
2887 declared in the current Objective-C class, and not those inherited
2888 from superclasses. It is the responsibility of the Objective-C
2889 runtime to invoke all such methods in an object's inheritance
2890 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2891 by the runtime immediately after a new object instance is allocated;
2892 the @code{- (void) .cxx_destruct} methods are invoked immediately
2893 before the runtime deallocates an object instance.
2895 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2896 support for invoking the @code{- (id) .cxx_construct} and
2897 @code{- (void) .cxx_destruct} methods.
2899 @item -fobjc-direct-dispatch
2900 @opindex fobjc-direct-dispatch
2901 Allow fast jumps to the message dispatcher. On Darwin this is
2902 accomplished via the comm page.
2904 @item -fobjc-exceptions
2905 @opindex fobjc-exceptions
2906 Enable syntactic support for structured exception handling in
2907 Objective-C, similar to what is offered by C++ and Java. This option
2908 is required to use the Objective-C keywords @code{@@try},
2909 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2910 @code{@@synchronized}. This option is available with both the GNU
2911 runtime and the NeXT runtime (but not available in conjunction with
2912 the NeXT runtime on Mac OS X 10.2 and earlier).
2916 Enable garbage collection (GC) in Objective-C and Objective-C++
2917 programs. This option is only available with the NeXT runtime; the
2918 GNU runtime has a different garbage collection implementation that
2919 does not require special compiler flags.
2921 @item -fobjc-nilcheck
2922 @opindex fobjc-nilcheck
2923 For the NeXT runtime with version 2 of the ABI, check for a nil
2924 receiver in method invocations before doing the actual method call.
2925 This is the default and can be disabled using
2926 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2927 checked for nil in this way no matter what this flag is set to.
2928 Currently this flag does nothing when the GNU runtime, or an older
2929 version of the NeXT runtime ABI, is used.
2931 @item -fobjc-std=objc1
2933 Conform to the language syntax of Objective-C 1.0, the language
2934 recognized by GCC 4.0. This only affects the Objective-C additions to
2935 the C/C++ language; it does not affect conformance to C/C++ standards,
2936 which is controlled by the separate C/C++ dialect option flags. When
2937 this option is used with the Objective-C or Objective-C++ compiler,
2938 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2939 This is useful if you need to make sure that your Objective-C code can
2940 be compiled with older versions of GCC@.
2942 @item -freplace-objc-classes
2943 @opindex freplace-objc-classes
2944 Emit a special marker instructing @command{ld(1)} not to statically link in
2945 the resulting object file, and allow @command{dyld(1)} to load it in at
2946 run time instead. This is used in conjunction with the Fix-and-Continue
2947 debugging mode, where the object file in question may be recompiled and
2948 dynamically reloaded in the course of program execution, without the need
2949 to restart the program itself. Currently, Fix-and-Continue functionality
2950 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2955 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2956 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2957 compile time) with static class references that get initialized at load time,
2958 which improves run-time performance. Specifying the @option{-fzero-link} flag
2959 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2960 to be retained. This is useful in Zero-Link debugging mode, since it allows
2961 for individual class implementations to be modified during program execution.
2962 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2963 regardless of command-line options.
2965 @item -fno-local-ivars
2966 @opindex fno-local-ivars
2967 @opindex flocal-ivars
2968 By default instance variables in Objective-C can be accessed as if
2969 they were local variables from within the methods of the class they're
2970 declared in. This can lead to shadowing between instance variables
2971 and other variables declared either locally inside a class method or
2972 globally with the same name. Specifying the @option{-fno-local-ivars}
2973 flag disables this behavior thus avoiding variable shadowing issues.
2975 @item -fivar-visibility=@var{public|protected|private|package}
2976 @opindex fivar-visibility
2977 Set the default instance variable visibility to the specified option
2978 so that instance variables declared outside the scope of any access
2979 modifier directives default to the specified visibility.
2983 Dump interface declarations for all classes seen in the source file to a
2984 file named @file{@var{sourcename}.decl}.
2986 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2987 @opindex Wassign-intercept
2988 @opindex Wno-assign-intercept
2989 Warn whenever an Objective-C assignment is being intercepted by the
2992 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2993 @opindex Wno-protocol
2995 If a class is declared to implement a protocol, a warning is issued for
2996 every method in the protocol that is not implemented by the class. The
2997 default behavior is to issue a warning for every method not explicitly
2998 implemented in the class, even if a method implementation is inherited
2999 from the superclass. If you use the @option{-Wno-protocol} option, then
3000 methods inherited from the superclass are considered to be implemented,
3001 and no warning is issued for them.
3003 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3005 @opindex Wno-selector
3006 Warn if multiple methods of different types for the same selector are
3007 found during compilation. The check is performed on the list of methods
3008 in the final stage of compilation. Additionally, a check is performed
3009 for each selector appearing in a @code{@@selector(@dots{})}
3010 expression, and a corresponding method for that selector has been found
3011 during compilation. Because these checks scan the method table only at
3012 the end of compilation, these warnings are not produced if the final
3013 stage of compilation is not reached, for example because an error is
3014 found during compilation, or because the @option{-fsyntax-only} option is
3017 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3018 @opindex Wstrict-selector-match
3019 @opindex Wno-strict-selector-match
3020 Warn if multiple methods with differing argument and/or return types are
3021 found for a given selector when attempting to send a message using this
3022 selector to a receiver of type @code{id} or @code{Class}. When this flag
3023 is off (which is the default behavior), the compiler omits such warnings
3024 if any differences found are confined to types that share the same size
3027 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3028 @opindex Wundeclared-selector
3029 @opindex Wno-undeclared-selector
3030 Warn if a @code{@@selector(@dots{})} expression referring to an
3031 undeclared selector is found. A selector is considered undeclared if no
3032 method with that name has been declared before the
3033 @code{@@selector(@dots{})} expression, either explicitly in an
3034 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3035 an @code{@@implementation} section. This option always performs its
3036 checks as soon as a @code{@@selector(@dots{})} expression is found,
3037 while @option{-Wselector} only performs its checks in the final stage of
3038 compilation. This also enforces the coding style convention
3039 that methods and selectors must be declared before being used.
3041 @item -print-objc-runtime-info
3042 @opindex print-objc-runtime-info
3043 Generate C header describing the largest structure that is passed by
3048 @node Language Independent Options
3049 @section Options to Control Diagnostic Messages Formatting
3050 @cindex options to control diagnostics formatting
3051 @cindex diagnostic messages
3052 @cindex message formatting
3054 Traditionally, diagnostic messages have been formatted irrespective of
3055 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3056 options described below
3057 to control the formatting algorithm for diagnostic messages,
3058 e.g.@: how many characters per line, how often source location
3059 information should be reported. Note that some language front ends may not
3060 honor these options.
3063 @item -fmessage-length=@var{n}
3064 @opindex fmessage-length
3065 Try to format error messages so that they fit on lines of about @var{n}
3066 characters. The default is 72 characters for @command{g++} and 0 for the rest of
3067 the front ends supported by GCC@. If @var{n} is zero, then no
3068 line-wrapping is done; each error message appears on a single
3071 @item -fdiagnostics-show-location=once
3072 @opindex fdiagnostics-show-location
3073 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3074 reporter to emit source location information @emph{once}; that is, in
3075 case the message is too long to fit on a single physical line and has to
3076 be wrapped, the source location won't be emitted (as prefix) again,
3077 over and over, in subsequent continuation lines. This is the default
3080 @item -fdiagnostics-show-location=every-line
3081 Only meaningful in line-wrapping mode. Instructs the diagnostic
3082 messages reporter to emit the same source location information (as
3083 prefix) for physical lines that result from the process of breaking
3084 a message which is too long to fit on a single line.
3086 @item -fdiagnostics-color[=@var{WHEN}]
3087 @itemx -fno-diagnostics-color
3088 @opindex fdiagnostics-color
3089 @cindex highlight, color, colour
3090 @vindex GCC_COLORS @r{environment variable}
3091 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3092 or @samp{auto}. The default is @samp{never} if @env{GCC_COLORS} environment
3093 variable isn't present in the environment, and @samp{auto} otherwise.
3094 @samp{auto} means to use color only when the standard error is a terminal.
3095 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3096 aliases for @option{-fdiagnostics-color=always} and
3097 @option{-fdiagnostics-color=never}, respectively.
3099 The colors are defined by the environment variable @env{GCC_COLORS}.
3100 Its value is a colon-separated list of capabilities and Select Graphic
3101 Rendition (SGR) substrings. SGR commands are interpreted by the
3102 terminal or terminal emulator. (See the section in the documentation
3103 of your text terminal for permitted values and their meanings as
3104 character attributes.) These substring values are integers in decimal
3105 representation and can be concatenated with semicolons.
3106 Common values to concatenate include
3108 @samp{4} for underline,
3110 @samp{7} for inverse,
3111 @samp{39} for default foreground color,
3112 @samp{30} to @samp{37} for foreground colors,
3113 @samp{90} to @samp{97} for 16-color mode foreground colors,
3114 @samp{38;5;0} to @samp{38;5;255}
3115 for 88-color and 256-color modes foreground colors,
3116 @samp{49} for default background color,
3117 @samp{40} to @samp{47} for background colors,
3118 @samp{100} to @samp{107} for 16-color mode background colors,
3119 and @samp{48;5;0} to @samp{48;5;255}
3120 for 88-color and 256-color modes background colors.
3122 The default @env{GCC_COLORS} is
3123 @samp{error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01}
3124 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3125 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3126 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3127 string disables colors.
3128 Supported capabilities are as follows.
3132 @vindex error GCC_COLORS @r{capability}
3133 SGR substring for error: markers.
3136 @vindex warning GCC_COLORS @r{capability}
3137 SGR substring for warning: markers.
3140 @vindex note GCC_COLORS @r{capability}
3141 SGR substring for note: markers.
3144 @vindex caret GCC_COLORS @r{capability}
3145 SGR substring for caret line.
3148 @vindex locus GCC_COLORS @r{capability}
3149 SGR substring for location information, @samp{file:line} or
3150 @samp{file:line:column} etc.
3153 @vindex quote GCC_COLORS @r{capability}
3154 SGR substring for information printed within quotes.
3157 @item -fno-diagnostics-show-option
3158 @opindex fno-diagnostics-show-option
3159 @opindex fdiagnostics-show-option
3160 By default, each diagnostic emitted includes text indicating the
3161 command-line option that directly controls the diagnostic (if such an
3162 option is known to the diagnostic machinery). Specifying the
3163 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3165 @item -fno-diagnostics-show-caret
3166 @opindex fno-diagnostics-show-caret
3167 @opindex fdiagnostics-show-caret
3168 By default, each diagnostic emitted includes the original source line
3169 and a caret '^' indicating the column. This option suppresses this
3174 @node Warning Options
3175 @section Options to Request or Suppress Warnings
3176 @cindex options to control warnings
3177 @cindex warning messages
3178 @cindex messages, warning
3179 @cindex suppressing warnings
3181 Warnings are diagnostic messages that report constructions that
3182 are not inherently erroneous but that are risky or suggest there
3183 may have been an error.
3185 The following language-independent options do not enable specific
3186 warnings but control the kinds of diagnostics produced by GCC@.
3189 @cindex syntax checking
3191 @opindex fsyntax-only
3192 Check the code for syntax errors, but don't do anything beyond that.
3194 @item -fmax-errors=@var{n}
3195 @opindex fmax-errors
3196 Limits the maximum number of error messages to @var{n}, at which point
3197 GCC bails out rather than attempting to continue processing the source
3198 code. If @var{n} is 0 (the default), there is no limit on the number
3199 of error messages produced. If @option{-Wfatal-errors} is also
3200 specified, then @option{-Wfatal-errors} takes precedence over this
3205 Inhibit all warning messages.
3210 Make all warnings into errors.
3215 Make the specified warning into an error. The specifier for a warning
3216 is appended; for example @option{-Werror=switch} turns the warnings
3217 controlled by @option{-Wswitch} into errors. This switch takes a
3218 negative form, to be used to negate @option{-Werror} for specific
3219 warnings; for example @option{-Wno-error=switch} makes
3220 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3223 The warning message for each controllable warning includes the
3224 option that controls the warning. That option can then be used with
3225 @option{-Werror=} and @option{-Wno-error=} as described above.
3226 (Printing of the option in the warning message can be disabled using the
3227 @option{-fno-diagnostics-show-option} flag.)
3229 Note that specifying @option{-Werror=}@var{foo} automatically implies
3230 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3233 @item -Wfatal-errors
3234 @opindex Wfatal-errors
3235 @opindex Wno-fatal-errors
3236 This option causes the compiler to abort compilation on the first error
3237 occurred rather than trying to keep going and printing further error
3242 You can request many specific warnings with options beginning with
3243 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3244 implicit declarations. Each of these specific warning options also
3245 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3246 example, @option{-Wno-implicit}. This manual lists only one of the
3247 two forms, whichever is not the default. For further
3248 language-specific options also refer to @ref{C++ Dialect Options} and
3249 @ref{Objective-C and Objective-C++ Dialect Options}.
3251 When an unrecognized warning option is requested (e.g.,
3252 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3253 that the option is not recognized. However, if the @option{-Wno-} form
3254 is used, the behavior is slightly different: no diagnostic is
3255 produced for @option{-Wno-unknown-warning} unless other diagnostics
3256 are being produced. This allows the use of new @option{-Wno-} options
3257 with old compilers, but if something goes wrong, the compiler
3258 warns that an unrecognized option is present.
3265 Issue all the warnings demanded by strict ISO C and ISO C++;
3266 reject all programs that use forbidden extensions, and some other
3267 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3268 version of the ISO C standard specified by any @option{-std} option used.
3270 Valid ISO C and ISO C++ programs should compile properly with or without
3271 this option (though a rare few require @option{-ansi} or a
3272 @option{-std} option specifying the required version of ISO C)@. However,
3273 without this option, certain GNU extensions and traditional C and C++
3274 features are supported as well. With this option, they are rejected.
3276 @option{-Wpedantic} does not cause warning messages for use of the
3277 alternate keywords whose names begin and end with @samp{__}. Pedantic
3278 warnings are also disabled in the expression that follows
3279 @code{__extension__}. However, only system header files should use
3280 these escape routes; application programs should avoid them.
3281 @xref{Alternate Keywords}.
3283 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3284 C conformance. They soon find that it does not do quite what they want:
3285 it finds some non-ISO practices, but not all---only those for which
3286 ISO C @emph{requires} a diagnostic, and some others for which
3287 diagnostics have been added.
3289 A feature to report any failure to conform to ISO C might be useful in
3290 some instances, but would require considerable additional work and would
3291 be quite different from @option{-Wpedantic}. We don't have plans to
3292 support such a feature in the near future.
3294 Where the standard specified with @option{-std} represents a GNU
3295 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3296 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3297 extended dialect is based. Warnings from @option{-Wpedantic} are given
3298 where they are required by the base standard. (It does not make sense
3299 for such warnings to be given only for features not in the specified GNU
3300 C dialect, since by definition the GNU dialects of C include all
3301 features the compiler supports with the given option, and there would be
3302 nothing to warn about.)
3304 @item -pedantic-errors
3305 @opindex pedantic-errors
3306 Like @option{-Wpedantic}, except that errors are produced rather than
3312 This enables all the warnings about constructions that some users
3313 consider questionable, and that are easy to avoid (or modify to
3314 prevent the warning), even in conjunction with macros. This also
3315 enables some language-specific warnings described in @ref{C++ Dialect
3316 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3318 @option{-Wall} turns on the following warning flags:
3320 @gccoptlist{-Waddress @gol
3321 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3323 -Wchar-subscripts @gol
3324 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3325 -Wimplicit-int @r{(C and Objective-C only)} @gol
3326 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3329 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3330 -Wmaybe-uninitialized @gol
3331 -Wmissing-braces @r{(only for C/ObjC)} @gol
3338 -Wsequence-point @gol
3339 -Wsign-compare @r{(only in C++)} @gol
3340 -Wstrict-aliasing @gol
3341 -Wstrict-overflow=1 @gol
3344 -Wuninitialized @gol
3345 -Wunknown-pragmas @gol
3346 -Wunused-function @gol
3349 -Wunused-variable @gol
3350 -Wvolatile-register-var @gol
3353 Note that some warning flags are not implied by @option{-Wall}. Some of
3354 them warn about constructions that users generally do not consider
3355 questionable, but which occasionally you might wish to check for;
3356 others warn about constructions that are necessary or hard to avoid in
3357 some cases, and there is no simple way to modify the code to suppress
3358 the warning. Some of them are enabled by @option{-Wextra} but many of
3359 them must be enabled individually.
3365 This enables some extra warning flags that are not enabled by
3366 @option{-Wall}. (This option used to be called @option{-W}. The older
3367 name is still supported, but the newer name is more descriptive.)
3369 @gccoptlist{-Wclobbered @gol
3371 -Wignored-qualifiers @gol
3372 -Wmissing-field-initializers @gol
3373 -Wmissing-parameter-type @r{(C only)} @gol
3374 -Wold-style-declaration @r{(C only)} @gol
3375 -Woverride-init @gol
3378 -Wuninitialized @gol
3379 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3380 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3383 The option @option{-Wextra} also prints warning messages for the
3389 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3390 @samp{>}, or @samp{>=}.
3393 (C++ only) An enumerator and a non-enumerator both appear in a
3394 conditional expression.
3397 (C++ only) Ambiguous virtual bases.
3400 (C++ only) Subscripting an array that has been declared @samp{register}.
3403 (C++ only) Taking the address of a variable that has been declared
3407 (C++ only) A base class is not initialized in a derived class's copy
3412 @item -Wchar-subscripts
3413 @opindex Wchar-subscripts
3414 @opindex Wno-char-subscripts
3415 Warn if an array subscript has type @code{char}. This is a common cause
3416 of error, as programmers often forget that this type is signed on some
3418 This warning is enabled by @option{-Wall}.
3422 @opindex Wno-comment
3423 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3424 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3425 This warning is enabled by @option{-Wall}.
3427 @item -Wno-coverage-mismatch
3428 @opindex Wno-coverage-mismatch
3429 Warn if feedback profiles do not match when using the
3430 @option{-fprofile-use} option.
3431 If a source file is changed between compiling with @option{-fprofile-gen} and
3432 with @option{-fprofile-use}, the files with the profile feedback can fail
3433 to match the source file and GCC cannot use the profile feedback
3434 information. By default, this warning is enabled and is treated as an
3435 error. @option{-Wno-coverage-mismatch} can be used to disable the
3436 warning or @option{-Wno-error=coverage-mismatch} can be used to
3437 disable the error. Disabling the error for this warning can result in
3438 poorly optimized code and is useful only in the
3439 case of very minor changes such as bug fixes to an existing code-base.
3440 Completely disabling the warning is not recommended.
3443 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3445 Suppress warning messages emitted by @code{#warning} directives.
3447 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3448 @opindex Wdouble-promotion
3449 @opindex Wno-double-promotion
3450 Give a warning when a value of type @code{float} is implicitly
3451 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3452 floating-point unit implement @code{float} in hardware, but emulate
3453 @code{double} in software. On such a machine, doing computations
3454 using @code{double} values is much more expensive because of the
3455 overhead required for software emulation.
3457 It is easy to accidentally do computations with @code{double} because
3458 floating-point literals are implicitly of type @code{double}. For
3462 float area(float radius)
3464 return 3.14159 * radius * radius;
3468 the compiler performs the entire computation with @code{double}
3469 because the floating-point literal is a @code{double}.
3472 @itemx -Wformat=@var{n}
3475 @opindex ffreestanding
3476 @opindex fno-builtin
3478 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3479 the arguments supplied have types appropriate to the format string
3480 specified, and that the conversions specified in the format string make
3481 sense. This includes standard functions, and others specified by format
3482 attributes (@pxref{Function Attributes}), in the @code{printf},
3483 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3484 not in the C standard) families (or other target-specific families).
3485 Which functions are checked without format attributes having been
3486 specified depends on the standard version selected, and such checks of
3487 functions without the attribute specified are disabled by
3488 @option{-ffreestanding} or @option{-fno-builtin}.
3490 The formats are checked against the format features supported by GNU
3491 libc version 2.2. These include all ISO C90 and C99 features, as well
3492 as features from the Single Unix Specification and some BSD and GNU
3493 extensions. Other library implementations may not support all these
3494 features; GCC does not support warning about features that go beyond a
3495 particular library's limitations. However, if @option{-Wpedantic} is used
3496 with @option{-Wformat}, warnings are given about format features not
3497 in the selected standard version (but not for @code{strfmon} formats,
3498 since those are not in any version of the C standard). @xref{C Dialect
3499 Options,,Options Controlling C Dialect}.
3506 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3507 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3508 @option{-Wformat} also checks for null format arguments for several
3509 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3510 aspects of this level of format checking can be disabled by the
3511 options: @option{-Wno-format-contains-nul},
3512 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3513 @option{-Wformat} is enabled by @option{-Wall}.
3515 @item -Wno-format-contains-nul
3516 @opindex Wno-format-contains-nul
3517 @opindex Wformat-contains-nul
3518 If @option{-Wformat} is specified, do not warn about format strings that
3521 @item -Wno-format-extra-args
3522 @opindex Wno-format-extra-args
3523 @opindex Wformat-extra-args
3524 If @option{-Wformat} is specified, do not warn about excess arguments to a
3525 @code{printf} or @code{scanf} format function. The C standard specifies
3526 that such arguments are ignored.
3528 Where the unused arguments lie between used arguments that are
3529 specified with @samp{$} operand number specifications, normally
3530 warnings are still given, since the implementation could not know what
3531 type to pass to @code{va_arg} to skip the unused arguments. However,
3532 in the case of @code{scanf} formats, this option suppresses the
3533 warning if the unused arguments are all pointers, since the Single
3534 Unix Specification says that such unused arguments are allowed.
3536 @item -Wno-format-zero-length
3537 @opindex Wno-format-zero-length
3538 @opindex Wformat-zero-length
3539 If @option{-Wformat} is specified, do not warn about zero-length formats.
3540 The C standard specifies that zero-length formats are allowed.
3545 Enable @option{-Wformat} plus additional format checks. Currently
3546 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3547 -Wformat-signedness -Wformat-y2k}.
3549 @item -Wformat-nonliteral
3550 @opindex Wformat-nonliteral
3551 @opindex Wno-format-nonliteral
3552 If @option{-Wformat} is specified, also warn if the format string is not a
3553 string literal and so cannot be checked, unless the format function
3554 takes its format arguments as a @code{va_list}.
3556 @item -Wformat-security
3557 @opindex Wformat-security
3558 @opindex Wno-format-security
3559 If @option{-Wformat} is specified, also warn about uses of format
3560 functions that represent possible security problems. At present, this
3561 warns about calls to @code{printf} and @code{scanf} functions where the
3562 format string is not a string literal and there are no format arguments,
3563 as in @code{printf (foo);}. This may be a security hole if the format
3564 string came from untrusted input and contains @samp{%n}. (This is
3565 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3566 in future warnings may be added to @option{-Wformat-security} that are not
3567 included in @option{-Wformat-nonliteral}.)
3569 @item -Wformat-signedness
3570 @opindex Wformat-signedness
3571 @opindex Wno-format-signedness
3572 If @option{-Wformat} is specified, also warn if the format string
3573 requires an unsigned argument and the argument is signed and vice versa.
3576 @opindex Wformat-y2k
3577 @opindex Wno-format-y2k
3578 If @option{-Wformat} is specified, also warn about @code{strftime}
3579 formats that may yield only a two-digit year.
3584 @opindex Wno-nonnull
3585 Warn about passing a null pointer for arguments marked as
3586 requiring a non-null value by the @code{nonnull} function attribute.
3588 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3589 can be disabled with the @option{-Wno-nonnull} option.
3591 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3593 @opindex Wno-init-self
3594 Warn about uninitialized variables that are initialized with themselves.
3595 Note this option can only be used with the @option{-Wuninitialized} option.
3597 For example, GCC warns about @code{i} being uninitialized in the
3598 following snippet only when @option{-Winit-self} has been specified:
3609 This warning is enabled by @option{-Wall} in C++.
3611 @item -Wimplicit-int @r{(C and Objective-C only)}
3612 @opindex Wimplicit-int
3613 @opindex Wno-implicit-int
3614 Warn when a declaration does not specify a type.
3615 This warning is enabled by @option{-Wall}.
3617 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3618 @opindex Wimplicit-function-declaration
3619 @opindex Wno-implicit-function-declaration
3620 Give a warning whenever a function is used before being declared. In
3621 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3622 enabled by default and it is made into an error by
3623 @option{-pedantic-errors}. This warning is also enabled by
3626 @item -Wimplicit @r{(C and Objective-C only)}
3628 @opindex Wno-implicit
3629 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3630 This warning is enabled by @option{-Wall}.
3632 @item -Wignored-qualifiers @r{(C and C++ only)}
3633 @opindex Wignored-qualifiers
3634 @opindex Wno-ignored-qualifiers
3635 Warn if the return type of a function has a type qualifier
3636 such as @code{const}. For ISO C such a type qualifier has no effect,
3637 since the value returned by a function is not an lvalue.
3638 For C++, the warning is only emitted for scalar types or @code{void}.
3639 ISO C prohibits qualified @code{void} return types on function
3640 definitions, so such return types always receive a warning
3641 even without this option.
3643 This warning is also enabled by @option{-Wextra}.
3648 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3649 a function with external linkage, returning int, taking either zero
3650 arguments, two, or three arguments of appropriate types. This warning
3651 is enabled by default in C++ and is enabled by either @option{-Wall}
3652 or @option{-Wpedantic}.
3654 @item -Wmissing-braces
3655 @opindex Wmissing-braces
3656 @opindex Wno-missing-braces
3657 Warn if an aggregate or union initializer is not fully bracketed. In
3658 the following example, the initializer for @samp{a} is not fully
3659 bracketed, but that for @samp{b} is fully bracketed. This warning is
3660 enabled by @option{-Wall} in C.
3663 int a[2][2] = @{ 0, 1, 2, 3 @};
3664 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3667 This warning is enabled by @option{-Wall}.
3669 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3670 @opindex Wmissing-include-dirs
3671 @opindex Wno-missing-include-dirs
3672 Warn if a user-supplied include directory does not exist.
3675 @opindex Wparentheses
3676 @opindex Wno-parentheses
3677 Warn if parentheses are omitted in certain contexts, such
3678 as when there is an assignment in a context where a truth value
3679 is expected, or when operators are nested whose precedence people
3680 often get confused about.
3682 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3683 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3684 interpretation from that of ordinary mathematical notation.
3686 Also warn about constructions where there may be confusion to which
3687 @code{if} statement an @code{else} branch belongs. Here is an example of
3702 In C/C++, every @code{else} branch belongs to the innermost possible
3703 @code{if} statement, which in this example is @code{if (b)}. This is
3704 often not what the programmer expected, as illustrated in the above
3705 example by indentation the programmer chose. When there is the
3706 potential for this confusion, GCC issues a warning when this flag
3707 is specified. To eliminate the warning, add explicit braces around
3708 the innermost @code{if} statement so there is no way the @code{else}
3709 can belong to the enclosing @code{if}. The resulting code
3726 Also warn for dangerous uses of the GNU extension to
3727 @code{?:} with omitted middle operand. When the condition
3728 in the @code{?}: operator is a boolean expression, the omitted value is
3729 always 1. Often programmers expect it to be a value computed
3730 inside the conditional expression instead.
3732 This warning is enabled by @option{-Wall}.
3734 @item -Wsequence-point
3735 @opindex Wsequence-point
3736 @opindex Wno-sequence-point
3737 Warn about code that may have undefined semantics because of violations
3738 of sequence point rules in the C and C++ standards.
3740 The C and C++ standards define the order in which expressions in a C/C++
3741 program are evaluated in terms of @dfn{sequence points}, which represent
3742 a partial ordering between the execution of parts of the program: those
3743 executed before the sequence point, and those executed after it. These
3744 occur after the evaluation of a full expression (one which is not part
3745 of a larger expression), after the evaluation of the first operand of a
3746 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3747 function is called (but after the evaluation of its arguments and the
3748 expression denoting the called function), and in certain other places.
3749 Other than as expressed by the sequence point rules, the order of
3750 evaluation of subexpressions of an expression is not specified. All
3751 these rules describe only a partial order rather than a total order,
3752 since, for example, if two functions are called within one expression
3753 with no sequence point between them, the order in which the functions
3754 are called is not specified. However, the standards committee have
3755 ruled that function calls do not overlap.
3757 It is not specified when between sequence points modifications to the
3758 values of objects take effect. Programs whose behavior depends on this
3759 have undefined behavior; the C and C++ standards specify that ``Between
3760 the previous and next sequence point an object shall have its stored
3761 value modified at most once by the evaluation of an expression.
3762 Furthermore, the prior value shall be read only to determine the value
3763 to be stored.''. If a program breaks these rules, the results on any
3764 particular implementation are entirely unpredictable.
3766 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3767 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3768 diagnosed by this option, and it may give an occasional false positive
3769 result, but in general it has been found fairly effective at detecting
3770 this sort of problem in programs.
3772 The standard is worded confusingly, therefore there is some debate
3773 over the precise meaning of the sequence point rules in subtle cases.
3774 Links to discussions of the problem, including proposed formal
3775 definitions, may be found on the GCC readings page, at
3776 @uref{http://gcc.gnu.org/@/readings.html}.
3778 This warning is enabled by @option{-Wall} for C and C++.
3780 @item -Wno-return-local-addr
3781 @opindex Wno-return-local-addr
3782 @opindex Wreturn-local-addr
3783 Do not warn about returning a pointer (or in C++, a reference) to a
3784 variable that goes out of scope after the function returns.
3787 @opindex Wreturn-type
3788 @opindex Wno-return-type
3789 Warn whenever a function is defined with a return type that defaults
3790 to @code{int}. Also warn about any @code{return} statement with no
3791 return value in a function whose return type is not @code{void}
3792 (falling off the end of the function body is considered returning
3793 without a value), and about a @code{return} statement with an
3794 expression in a function whose return type is @code{void}.
3796 For C++, a function without return type always produces a diagnostic
3797 message, even when @option{-Wno-return-type} is specified. The only
3798 exceptions are @samp{main} and functions defined in system headers.
3800 This warning is enabled by @option{-Wall}.
3805 Warn whenever a @code{switch} statement has an index of enumerated type
3806 and lacks a @code{case} for one or more of the named codes of that
3807 enumeration. (The presence of a @code{default} label prevents this
3808 warning.) @code{case} labels outside the enumeration range also
3809 provoke warnings when this option is used (even if there is a
3810 @code{default} label).
3811 This warning is enabled by @option{-Wall}.
3813 @item -Wswitch-default
3814 @opindex Wswitch-default
3815 @opindex Wno-switch-default
3816 Warn whenever a @code{switch} statement does not have a @code{default}
3820 @opindex Wswitch-enum
3821 @opindex Wno-switch-enum
3822 Warn whenever a @code{switch} statement has an index of enumerated type
3823 and lacks a @code{case} for one or more of the named codes of that
3824 enumeration. @code{case} labels outside the enumeration range also
3825 provoke warnings when this option is used. The only difference
3826 between @option{-Wswitch} and this option is that this option gives a
3827 warning about an omitted enumeration code even if there is a
3828 @code{default} label.
3831 @opindex Wswitch-bool
3832 @opindex Wno-switch-bool
3833 Warn whenever a @code{switch} statement has an index of boolean type.
3834 It is possible to suppress this warning by casting the controlling
3835 expression to a type other than @code{bool}. For example:
3838 switch ((int) (a == 4))
3844 This warning is enabled by default for C and C++ programs.
3846 @item -Wsync-nand @r{(C and C++ only)}
3848 @opindex Wno-sync-nand
3849 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3850 built-in functions are used. These functions changed semantics in GCC 4.4.
3854 @opindex Wno-trigraphs
3855 Warn if any trigraphs are encountered that might change the meaning of
3856 the program (trigraphs within comments are not warned about).
3857 This warning is enabled by @option{-Wall}.
3859 @item -Wunused-but-set-parameter
3860 @opindex Wunused-but-set-parameter
3861 @opindex Wno-unused-but-set-parameter
3862 Warn whenever a function parameter is assigned to, but otherwise unused
3863 (aside from its declaration).
3865 To suppress this warning use the @samp{unused} attribute
3866 (@pxref{Variable Attributes}).
3868 This warning is also enabled by @option{-Wunused} together with
3871 @item -Wunused-but-set-variable
3872 @opindex Wunused-but-set-variable
3873 @opindex Wno-unused-but-set-variable
3874 Warn whenever a local variable is assigned to, but otherwise unused
3875 (aside from its declaration).
3876 This warning is enabled by @option{-Wall}.
3878 To suppress this warning use the @samp{unused} attribute
3879 (@pxref{Variable Attributes}).
3881 This warning is also enabled by @option{-Wunused}, which is enabled
3884 @item -Wunused-function
3885 @opindex Wunused-function
3886 @opindex Wno-unused-function
3887 Warn whenever a static function is declared but not defined or a
3888 non-inline static function is unused.
3889 This warning is enabled by @option{-Wall}.
3891 @item -Wunused-label
3892 @opindex Wunused-label
3893 @opindex Wno-unused-label
3894 Warn whenever a label is declared but not used.
3895 This warning is enabled by @option{-Wall}.
3897 To suppress this warning use the @samp{unused} attribute
3898 (@pxref{Variable Attributes}).
3900 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3901 @opindex Wunused-local-typedefs
3902 Warn when a typedef locally defined in a function is not used.
3903 This warning is enabled by @option{-Wall}.
3905 @item -Wunused-parameter
3906 @opindex Wunused-parameter
3907 @opindex Wno-unused-parameter
3908 Warn whenever a function parameter is unused aside from its declaration.
3910 To suppress this warning use the @samp{unused} attribute
3911 (@pxref{Variable Attributes}).
3913 @item -Wno-unused-result
3914 @opindex Wunused-result
3915 @opindex Wno-unused-result
3916 Do not warn if a caller of a function marked with attribute
3917 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3918 its return value. The default is @option{-Wunused-result}.
3920 @item -Wunused-variable
3921 @opindex Wunused-variable
3922 @opindex Wno-unused-variable
3923 Warn whenever a local variable or non-constant static variable is unused
3924 aside from its declaration.
3925 This warning is enabled by @option{-Wall}.
3927 To suppress this warning use the @samp{unused} attribute
3928 (@pxref{Variable Attributes}).
3930 @item -Wunused-value
3931 @opindex Wunused-value
3932 @opindex Wno-unused-value
3933 Warn whenever a statement computes a result that is explicitly not
3934 used. To suppress this warning cast the unused expression to
3935 @samp{void}. This includes an expression-statement or the left-hand
3936 side of a comma expression that contains no side effects. For example,
3937 an expression such as @samp{x[i,j]} causes a warning, while
3938 @samp{x[(void)i,j]} does not.
3940 This warning is enabled by @option{-Wall}.
3945 All the above @option{-Wunused} options combined.
3947 In order to get a warning about an unused function parameter, you must
3948 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
3949 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
3951 @item -Wuninitialized
3952 @opindex Wuninitialized
3953 @opindex Wno-uninitialized
3954 Warn if an automatic variable is used without first being initialized
3955 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3956 warn if a non-static reference or non-static @samp{const} member
3957 appears in a class without constructors.
3959 If you want to warn about code that uses the uninitialized value of the
3960 variable in its own initializer, use the @option{-Winit-self} option.
3962 These warnings occur for individual uninitialized or clobbered
3963 elements of structure, union or array variables as well as for
3964 variables that are uninitialized or clobbered as a whole. They do
3965 not occur for variables or elements declared @code{volatile}. Because
3966 these warnings depend on optimization, the exact variables or elements
3967 for which there are warnings depends on the precise optimization
3968 options and version of GCC used.
3970 Note that there may be no warning about a variable that is used only
3971 to compute a value that itself is never used, because such
3972 computations may be deleted by data flow analysis before the warnings
3975 @item -Wmaybe-uninitialized
3976 @opindex Wmaybe-uninitialized
3977 @opindex Wno-maybe-uninitialized
3978 For an automatic variable, if there exists a path from the function
3979 entry to a use of the variable that is initialized, but there exist
3980 some other paths for which the variable is not initialized, the compiler
3981 emits a warning if it cannot prove the uninitialized paths are not
3982 executed at run time. These warnings are made optional because GCC is
3983 not smart enough to see all the reasons why the code might be correct
3984 in spite of appearing to have an error. Here is one example of how
4005 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4006 always initialized, but GCC doesn't know this. To suppress the
4007 warning, you need to provide a default case with assert(0) or
4010 @cindex @code{longjmp} warnings
4011 This option also warns when a non-volatile automatic variable might be
4012 changed by a call to @code{longjmp}. These warnings as well are possible
4013 only in optimizing compilation.
4015 The compiler sees only the calls to @code{setjmp}. It cannot know
4016 where @code{longjmp} will be called; in fact, a signal handler could
4017 call it at any point in the code. As a result, you may get a warning
4018 even when there is in fact no problem because @code{longjmp} cannot
4019 in fact be called at the place that would cause a problem.
4021 Some spurious warnings can be avoided if you declare all the functions
4022 you use that never return as @code{noreturn}. @xref{Function
4025 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4027 @item -Wunknown-pragmas
4028 @opindex Wunknown-pragmas
4029 @opindex Wno-unknown-pragmas
4030 @cindex warning for unknown pragmas
4031 @cindex unknown pragmas, warning
4032 @cindex pragmas, warning of unknown
4033 Warn when a @code{#pragma} directive is encountered that is not understood by
4034 GCC@. If this command-line option is used, warnings are even issued
4035 for unknown pragmas in system header files. This is not the case if
4036 the warnings are only enabled by the @option{-Wall} command-line option.
4039 @opindex Wno-pragmas
4041 Do not warn about misuses of pragmas, such as incorrect parameters,
4042 invalid syntax, or conflicts between pragmas. See also
4043 @option{-Wunknown-pragmas}.
4045 @item -Wstrict-aliasing
4046 @opindex Wstrict-aliasing
4047 @opindex Wno-strict-aliasing
4048 This option is only active when @option{-fstrict-aliasing} is active.
4049 It warns about code that might break the strict aliasing rules that the
4050 compiler is using for optimization. The warning does not catch all
4051 cases, but does attempt to catch the more common pitfalls. It is
4052 included in @option{-Wall}.
4053 It is equivalent to @option{-Wstrict-aliasing=3}
4055 @item -Wstrict-aliasing=n
4056 @opindex Wstrict-aliasing=n
4057 This option is only active when @option{-fstrict-aliasing} is active.
4058 It warns about code that might break the strict aliasing rules that the
4059 compiler is using for optimization.
4060 Higher levels correspond to higher accuracy (fewer false positives).
4061 Higher levels also correspond to more effort, similar to the way @option{-O}
4063 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4065 Level 1: Most aggressive, quick, least accurate.
4066 Possibly useful when higher levels
4067 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4068 false negatives. However, it has many false positives.
4069 Warns for all pointer conversions between possibly incompatible types,
4070 even if never dereferenced. Runs in the front end only.
4072 Level 2: Aggressive, quick, not too precise.
4073 May still have many false positives (not as many as level 1 though),
4074 and few false negatives (but possibly more than level 1).
4075 Unlike level 1, it only warns when an address is taken. Warns about
4076 incomplete types. Runs in the front end only.
4078 Level 3 (default for @option{-Wstrict-aliasing}):
4079 Should have very few false positives and few false
4080 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4081 Takes care of the common pun+dereference pattern in the front end:
4082 @code{*(int*)&some_float}.
4083 If optimization is enabled, it also runs in the back end, where it deals
4084 with multiple statement cases using flow-sensitive points-to information.
4085 Only warns when the converted pointer is dereferenced.
4086 Does not warn about incomplete types.
4088 @item -Wstrict-overflow
4089 @itemx -Wstrict-overflow=@var{n}
4090 @opindex Wstrict-overflow
4091 @opindex Wno-strict-overflow
4092 This option is only active when @option{-fstrict-overflow} is active.
4093 It warns about cases where the compiler optimizes based on the
4094 assumption that signed overflow does not occur. Note that it does not
4095 warn about all cases where the code might overflow: it only warns
4096 about cases where the compiler implements some optimization. Thus
4097 this warning depends on the optimization level.
4099 An optimization that assumes that signed overflow does not occur is
4100 perfectly safe if the values of the variables involved are such that
4101 overflow never does, in fact, occur. Therefore this warning can
4102 easily give a false positive: a warning about code that is not
4103 actually a problem. To help focus on important issues, several
4104 warning levels are defined. No warnings are issued for the use of
4105 undefined signed overflow when estimating how many iterations a loop
4106 requires, in particular when determining whether a loop will be
4110 @item -Wstrict-overflow=1
4111 Warn about cases that are both questionable and easy to avoid. For
4112 example, with @option{-fstrict-overflow}, the compiler simplifies
4113 @code{x + 1 > x} to @code{1}. This level of
4114 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4115 are not, and must be explicitly requested.
4117 @item -Wstrict-overflow=2
4118 Also warn about other cases where a comparison is simplified to a
4119 constant. For example: @code{abs (x) >= 0}. This can only be
4120 simplified when @option{-fstrict-overflow} is in effect, because
4121 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4122 zero. @option{-Wstrict-overflow} (with no level) is the same as
4123 @option{-Wstrict-overflow=2}.
4125 @item -Wstrict-overflow=3
4126 Also warn about other cases where a comparison is simplified. For
4127 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4129 @item -Wstrict-overflow=4
4130 Also warn about other simplifications not covered by the above cases.
4131 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4133 @item -Wstrict-overflow=5
4134 Also warn about cases where the compiler reduces the magnitude of a
4135 constant involved in a comparison. For example: @code{x + 2 > y} is
4136 simplified to @code{x + 1 >= y}. This is reported only at the
4137 highest warning level because this simplification applies to many
4138 comparisons, so this warning level gives a very large number of
4142 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4143 @opindex Wsuggest-attribute=
4144 @opindex Wno-suggest-attribute=
4145 Warn for cases where adding an attribute may be beneficial. The
4146 attributes currently supported are listed below.
4149 @item -Wsuggest-attribute=pure
4150 @itemx -Wsuggest-attribute=const
4151 @itemx -Wsuggest-attribute=noreturn
4152 @opindex Wsuggest-attribute=pure
4153 @opindex Wno-suggest-attribute=pure
4154 @opindex Wsuggest-attribute=const
4155 @opindex Wno-suggest-attribute=const
4156 @opindex Wsuggest-attribute=noreturn
4157 @opindex Wno-suggest-attribute=noreturn
4159 Warn about functions that might be candidates for attributes
4160 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4161 functions visible in other compilation units or (in the case of @code{pure} and
4162 @code{const}) if it cannot prove that the function returns normally. A function
4163 returns normally if it doesn't contain an infinite loop or return abnormally
4164 by throwing, calling @code{abort()} or trapping. This analysis requires option
4165 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4166 higher. Higher optimization levels improve the accuracy of the analysis.
4168 @item -Wsuggest-attribute=format
4169 @itemx -Wmissing-format-attribute
4170 @opindex Wsuggest-attribute=format
4171 @opindex Wmissing-format-attribute
4172 @opindex Wno-suggest-attribute=format
4173 @opindex Wno-missing-format-attribute
4177 Warn about function pointers that might be candidates for @code{format}
4178 attributes. Note these are only possible candidates, not absolute ones.
4179 GCC guesses that function pointers with @code{format} attributes that
4180 are used in assignment, initialization, parameter passing or return
4181 statements should have a corresponding @code{format} attribute in the
4182 resulting type. I.e.@: the left-hand side of the assignment or
4183 initialization, the type of the parameter variable, or the return type
4184 of the containing function respectively should also have a @code{format}
4185 attribute to avoid the warning.
4187 GCC also warns about function definitions that might be
4188 candidates for @code{format} attributes. Again, these are only
4189 possible candidates. GCC guesses that @code{format} attributes
4190 might be appropriate for any function that calls a function like
4191 @code{vprintf} or @code{vscanf}, but this might not always be the
4192 case, and some functions for which @code{format} attributes are
4193 appropriate may not be detected.
4196 @item -Warray-bounds
4197 @opindex Wno-array-bounds
4198 @opindex Warray-bounds
4199 This option is only active when @option{-ftree-vrp} is active
4200 (default for @option{-O2} and above). It warns about subscripts to arrays
4201 that are always out of bounds. This warning is enabled by @option{-Wall}.
4203 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4204 @opindex Wno-discarded-qualifiers
4205 @opindex Wdiscarded-qualifiers
4206 Do not warn if type qualifiers on pointers are being discarded.
4207 Typically, the compiler will warn if a @code{const char *} variable is
4208 passed to a function that takes @code{char *} parameter. This option
4209 can be used to suppress such a warning.
4211 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4212 @opindex Wno-incompatible-pointer-types
4213 @opindex Wincompatible-pointer-types
4214 Do not warn when there is a conversion between pointers that have incompatible
4215 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4216 which warns for pointer argument passing or assignment with different signedness
4218 @item -Wno-int-conversion @r{(C and Objective-C only)}
4219 @opindex Wno-int-conversion
4220 @opindex Wint-conversion
4221 Do not warn about incompatible integer to pointer and pointer to integer
4222 conversions. This warning is about implicit conversions; for explicit
4223 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4224 @option{-Wno-pointer-to-int-cast} may be used.
4226 @item -Wno-div-by-zero
4227 @opindex Wno-div-by-zero
4228 @opindex Wdiv-by-zero
4229 Do not warn about compile-time integer division by zero. Floating-point
4230 division by zero is not warned about, as it can be a legitimate way of
4231 obtaining infinities and NaNs.
4233 @item -Wsystem-headers
4234 @opindex Wsystem-headers
4235 @opindex Wno-system-headers
4236 @cindex warnings from system headers
4237 @cindex system headers, warnings from
4238 Print warning messages for constructs found in system header files.
4239 Warnings from system headers are normally suppressed, on the assumption
4240 that they usually do not indicate real problems and would only make the
4241 compiler output harder to read. Using this command-line option tells
4242 GCC to emit warnings from system headers as if they occurred in user
4243 code. However, note that using @option{-Wall} in conjunction with this
4244 option does @emph{not} warn about unknown pragmas in system
4245 headers---for that, @option{-Wunknown-pragmas} must also be used.
4248 @opindex Wtrampolines
4249 @opindex Wno-trampolines
4250 Warn about trampolines generated for pointers to nested functions.
4252 A trampoline is a small piece of data or code that is created at run
4253 time on the stack when the address of a nested function is taken, and
4254 is used to call the nested function indirectly. For some targets, it
4255 is made up of data only and thus requires no special treatment. But,
4256 for most targets, it is made up of code and thus requires the stack
4257 to be made executable in order for the program to work properly.
4260 @opindex Wfloat-equal
4261 @opindex Wno-float-equal
4262 Warn if floating-point values are used in equality comparisons.
4264 The idea behind this is that sometimes it is convenient (for the
4265 programmer) to consider floating-point values as approximations to
4266 infinitely precise real numbers. If you are doing this, then you need
4267 to compute (by analyzing the code, or in some other way) the maximum or
4268 likely maximum error that the computation introduces, and allow for it
4269 when performing comparisons (and when producing output, but that's a
4270 different problem). In particular, instead of testing for equality, you
4271 should check to see whether the two values have ranges that overlap; and
4272 this is done with the relational operators, so equality comparisons are
4275 @item -Wtraditional @r{(C and Objective-C only)}
4276 @opindex Wtraditional
4277 @opindex Wno-traditional
4278 Warn about certain constructs that behave differently in traditional and
4279 ISO C@. Also warn about ISO C constructs that have no traditional C
4280 equivalent, and/or problematic constructs that should be avoided.
4284 Macro parameters that appear within string literals in the macro body.
4285 In traditional C macro replacement takes place within string literals,
4286 but in ISO C it does not.
4289 In traditional C, some preprocessor directives did not exist.
4290 Traditional preprocessors only considered a line to be a directive
4291 if the @samp{#} appeared in column 1 on the line. Therefore
4292 @option{-Wtraditional} warns about directives that traditional C
4293 understands but ignores because the @samp{#} does not appear as the
4294 first character on the line. It also suggests you hide directives like
4295 @samp{#pragma} not understood by traditional C by indenting them. Some
4296 traditional implementations do not recognize @samp{#elif}, so this option
4297 suggests avoiding it altogether.
4300 A function-like macro that appears without arguments.
4303 The unary plus operator.
4306 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4307 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4308 constants.) Note, these suffixes appear in macros defined in the system
4309 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4310 Use of these macros in user code might normally lead to spurious
4311 warnings, however GCC's integrated preprocessor has enough context to
4312 avoid warning in these cases.
4315 A function declared external in one block and then used after the end of
4319 A @code{switch} statement has an operand of type @code{long}.
4322 A non-@code{static} function declaration follows a @code{static} one.
4323 This construct is not accepted by some traditional C compilers.
4326 The ISO type of an integer constant has a different width or
4327 signedness from its traditional type. This warning is only issued if
4328 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4329 typically represent bit patterns, are not warned about.
4332 Usage of ISO string concatenation is detected.
4335 Initialization of automatic aggregates.
4338 Identifier conflicts with labels. Traditional C lacks a separate
4339 namespace for labels.
4342 Initialization of unions. If the initializer is zero, the warning is
4343 omitted. This is done under the assumption that the zero initializer in
4344 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4345 initializer warnings and relies on default initialization to zero in the
4349 Conversions by prototypes between fixed/floating-point values and vice
4350 versa. The absence of these prototypes when compiling with traditional
4351 C causes serious problems. This is a subset of the possible
4352 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4355 Use of ISO C style function definitions. This warning intentionally is
4356 @emph{not} issued for prototype declarations or variadic functions
4357 because these ISO C features appear in your code when using
4358 libiberty's traditional C compatibility macros, @code{PARAMS} and
4359 @code{VPARAMS}. This warning is also bypassed for nested functions
4360 because that feature is already a GCC extension and thus not relevant to
4361 traditional C compatibility.
4364 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4365 @opindex Wtraditional-conversion
4366 @opindex Wno-traditional-conversion
4367 Warn if a prototype causes a type conversion that is different from what
4368 would happen to the same argument in the absence of a prototype. This
4369 includes conversions of fixed point to floating and vice versa, and
4370 conversions changing the width or signedness of a fixed-point argument
4371 except when the same as the default promotion.
4373 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4374 @opindex Wdeclaration-after-statement
4375 @opindex Wno-declaration-after-statement
4376 Warn when a declaration is found after a statement in a block. This
4377 construct, known from C++, was introduced with ISO C99 and is by default
4378 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4379 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4384 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4386 @item -Wno-endif-labels
4387 @opindex Wno-endif-labels
4388 @opindex Wendif-labels
4389 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4394 Warn whenever a local variable or type declaration shadows another
4395 variable, parameter, type, class member (in C++), or instance variable
4396 (in Objective-C) or whenever a built-in function is shadowed. Note
4397 that in C++, the compiler warns if a local variable shadows an
4398 explicit typedef, but not if it shadows a struct/class/enum.
4400 @item -Wno-shadow-ivar @r{(Objective-C only)}
4401 @opindex Wno-shadow-ivar
4402 @opindex Wshadow-ivar
4403 Do not warn whenever a local variable shadows an instance variable in an
4406 @item -Wlarger-than=@var{len}
4407 @opindex Wlarger-than=@var{len}
4408 @opindex Wlarger-than-@var{len}
4409 Warn whenever an object of larger than @var{len} bytes is defined.
4411 @item -Wframe-larger-than=@var{len}
4412 @opindex Wframe-larger-than
4413 Warn if the size of a function frame is larger than @var{len} bytes.
4414 The computation done to determine the stack frame size is approximate
4415 and not conservative.
4416 The actual requirements may be somewhat greater than @var{len}
4417 even if you do not get a warning. In addition, any space allocated
4418 via @code{alloca}, variable-length arrays, or related constructs
4419 is not included by the compiler when determining
4420 whether or not to issue a warning.
4422 @item -Wno-free-nonheap-object
4423 @opindex Wno-free-nonheap-object
4424 @opindex Wfree-nonheap-object
4425 Do not warn when attempting to free an object that was not allocated
4428 @item -Wstack-usage=@var{len}
4429 @opindex Wstack-usage
4430 Warn if the stack usage of a function might be larger than @var{len} bytes.
4431 The computation done to determine the stack usage is conservative.
4432 Any space allocated via @code{alloca}, variable-length arrays, or related
4433 constructs is included by the compiler when determining whether or not to
4436 The message is in keeping with the output of @option{-fstack-usage}.
4440 If the stack usage is fully static but exceeds the specified amount, it's:
4443 warning: stack usage is 1120 bytes
4446 If the stack usage is (partly) dynamic but bounded, it's:
4449 warning: stack usage might be 1648 bytes
4452 If the stack usage is (partly) dynamic and not bounded, it's:
4455 warning: stack usage might be unbounded
4459 @item -Wunsafe-loop-optimizations
4460 @opindex Wunsafe-loop-optimizations
4461 @opindex Wno-unsafe-loop-optimizations
4462 Warn if the loop cannot be optimized because the compiler cannot
4463 assume anything on the bounds of the loop indices. With
4464 @option{-funsafe-loop-optimizations} warn if the compiler makes
4467 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4468 @opindex Wno-pedantic-ms-format
4469 @opindex Wpedantic-ms-format
4470 When used in combination with @option{-Wformat}
4471 and @option{-pedantic} without GNU extensions, this option
4472 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4473 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4474 which depend on the MS runtime.
4476 @item -Wpointer-arith
4477 @opindex Wpointer-arith
4478 @opindex Wno-pointer-arith
4479 Warn about anything that depends on the ``size of'' a function type or
4480 of @code{void}. GNU C assigns these types a size of 1, for
4481 convenience in calculations with @code{void *} pointers and pointers
4482 to functions. In C++, warn also when an arithmetic operation involves
4483 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4486 @opindex Wtype-limits
4487 @opindex Wno-type-limits
4488 Warn if a comparison is always true or always false due to the limited
4489 range of the data type, but do not warn for constant expressions. For
4490 example, warn if an unsigned variable is compared against zero with
4491 @samp{<} or @samp{>=}. This warning is also enabled by
4494 @item -Wbad-function-cast @r{(C and Objective-C only)}
4495 @opindex Wbad-function-cast
4496 @opindex Wno-bad-function-cast
4497 Warn whenever a function call is cast to a non-matching type.
4498 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4500 @item -Wc++-compat @r{(C and Objective-C only)}
4501 Warn about ISO C constructs that are outside of the common subset of
4502 ISO C and ISO C++, e.g.@: request for implicit conversion from
4503 @code{void *} to a pointer to non-@code{void} type.
4505 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4506 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4507 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4508 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4509 enabled by @option{-Wall}.
4513 @opindex Wno-cast-qual
4514 Warn whenever a pointer is cast so as to remove a type qualifier from
4515 the target type. For example, warn if a @code{const char *} is cast
4516 to an ordinary @code{char *}.
4518 Also warn when making a cast that introduces a type qualifier in an
4519 unsafe way. For example, casting @code{char **} to @code{const char **}
4520 is unsafe, as in this example:
4523 /* p is char ** value. */
4524 const char **q = (const char **) p;
4525 /* Assignment of readonly string to const char * is OK. */
4527 /* Now char** pointer points to read-only memory. */
4532 @opindex Wcast-align
4533 @opindex Wno-cast-align
4534 Warn whenever a pointer is cast such that the required alignment of the
4535 target is increased. For example, warn if a @code{char *} is cast to
4536 an @code{int *} on machines where integers can only be accessed at
4537 two- or four-byte boundaries.
4539 @item -Wwrite-strings
4540 @opindex Wwrite-strings
4541 @opindex Wno-write-strings
4542 When compiling C, give string constants the type @code{const
4543 char[@var{length}]} so that copying the address of one into a
4544 non-@code{const} @code{char *} pointer produces a warning. These
4545 warnings help you find at compile time code that can try to write
4546 into a string constant, but only if you have been very careful about
4547 using @code{const} in declarations and prototypes. Otherwise, it is
4548 just a nuisance. This is why we did not make @option{-Wall} request
4551 When compiling C++, warn about the deprecated conversion from string
4552 literals to @code{char *}. This warning is enabled by default for C++
4557 @opindex Wno-clobbered
4558 Warn for variables that might be changed by @samp{longjmp} or
4559 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4561 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4562 @opindex Wconditionally-supported
4563 @opindex Wno-conditionally-supported
4564 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4567 @opindex Wconversion
4568 @opindex Wno-conversion
4569 Warn for implicit conversions that may alter a value. This includes
4570 conversions between real and integer, like @code{abs (x)} when
4571 @code{x} is @code{double}; conversions between signed and unsigned,
4572 like @code{unsigned ui = -1}; and conversions to smaller types, like
4573 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4574 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4575 changed by the conversion like in @code{abs (2.0)}. Warnings about
4576 conversions between signed and unsigned integers can be disabled by
4577 using @option{-Wno-sign-conversion}.
4579 For C++, also warn for confusing overload resolution for user-defined
4580 conversions; and conversions that never use a type conversion
4581 operator: conversions to @code{void}, the same type, a base class or a
4582 reference to them. Warnings about conversions between signed and
4583 unsigned integers are disabled by default in C++ unless
4584 @option{-Wsign-conversion} is explicitly enabled.
4586 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4587 @opindex Wconversion-null
4588 @opindex Wno-conversion-null
4589 Do not warn for conversions between @code{NULL} and non-pointer
4590 types. @option{-Wconversion-null} is enabled by default.
4592 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4593 @opindex Wzero-as-null-pointer-constant
4594 @opindex Wno-zero-as-null-pointer-constant
4595 Warn when a literal '0' is used as null pointer constant. This can
4596 be useful to facilitate the conversion to @code{nullptr} in C++11.
4600 @opindex Wno-date-time
4601 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4602 are encountered as they might prevent bit-wise-identical reproducible
4605 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4606 @opindex Wdelete-incomplete
4607 @opindex Wno-delete-incomplete
4608 Warn when deleting a pointer to incomplete type, which may cause
4609 undefined behavior at runtime. This warning is enabled by default.
4611 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4612 @opindex Wuseless-cast
4613 @opindex Wno-useless-cast
4614 Warn when an expression is casted to its own type.
4617 @opindex Wempty-body
4618 @opindex Wno-empty-body
4619 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4620 while} statement. This warning is also enabled by @option{-Wextra}.
4622 @item -Wenum-compare
4623 @opindex Wenum-compare
4624 @opindex Wno-enum-compare
4625 Warn about a comparison between values of different enumerated types.
4626 In C++ enumeral mismatches in conditional expressions are also
4627 diagnosed and the warning is enabled by default. In C this warning is
4628 enabled by @option{-Wall}.
4630 @item -Wjump-misses-init @r{(C, Objective-C only)}
4631 @opindex Wjump-misses-init
4632 @opindex Wno-jump-misses-init
4633 Warn if a @code{goto} statement or a @code{switch} statement jumps
4634 forward across the initialization of a variable, or jumps backward to a
4635 label after the variable has been initialized. This only warns about
4636 variables that are initialized when they are declared. This warning is
4637 only supported for C and Objective-C; in C++ this sort of branch is an
4640 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4641 can be disabled with the @option{-Wno-jump-misses-init} option.
4643 @item -Wsign-compare
4644 @opindex Wsign-compare
4645 @opindex Wno-sign-compare
4646 @cindex warning for comparison of signed and unsigned values
4647 @cindex comparison of signed and unsigned values, warning
4648 @cindex signed and unsigned values, comparison warning
4649 Warn when a comparison between signed and unsigned values could produce
4650 an incorrect result when the signed value is converted to unsigned.
4651 This warning is also enabled by @option{-Wextra}; to get the other warnings
4652 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4654 @item -Wsign-conversion
4655 @opindex Wsign-conversion
4656 @opindex Wno-sign-conversion
4657 Warn for implicit conversions that may change the sign of an integer
4658 value, like assigning a signed integer expression to an unsigned
4659 integer variable. An explicit cast silences the warning. In C, this
4660 option is enabled also by @option{-Wconversion}.
4662 @item -Wfloat-conversion
4663 @opindex Wfloat-conversion
4664 @opindex Wno-float-conversion
4665 Warn for implicit conversions that reduce the precision of a real value.
4666 This includes conversions from real to integer, and from higher precision
4667 real to lower precision real values. This option is also enabled by
4668 @option{-Wconversion}.
4670 @item -Wsizeof-pointer-memaccess
4671 @opindex Wsizeof-pointer-memaccess
4672 @opindex Wno-sizeof-pointer-memaccess
4673 Warn for suspicious length parameters to certain string and memory built-in
4674 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4675 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4676 but a pointer, and suggests a possible fix, or about
4677 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4680 @item -Wsizeof-array-argument
4681 @opindex Wsizeof-array-argument
4682 @opindex Wno-sizeof-array-argument
4683 Warn when the @code{sizeof} operator is applied to a parameter that is
4684 declared as an array in a function definition. This warning is enabled by
4685 default for C and C++ programs.
4687 @item -Wmemset-transposed-args
4688 @opindex Wmemset-transposed-args
4689 @opindex Wno-memset-transposed-args
4690 Warn for suspicious calls to the @code{memset} built-in function, if the
4691 second argument is not zero and the third argument is zero. This warns e.g.@
4692 about @code{memset (buf, sizeof buf, 0)} where most probably
4693 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4694 is only emitted if the third argument is literal zero, if it is some expression
4695 that is folded to zero, or e.g. a cast of zero to some type etc., it
4696 is far less likely that user has mistakenly exchanged the arguments and
4697 no warning is emitted. This warning is enabled by @option{-Wall}.
4701 @opindex Wno-address
4702 Warn about suspicious uses of memory addresses. These include using
4703 the address of a function in a conditional expression, such as
4704 @code{void func(void); if (func)}, and comparisons against the memory
4705 address of a string literal, such as @code{if (x == "abc")}. Such
4706 uses typically indicate a programmer error: the address of a function
4707 always evaluates to true, so their use in a conditional usually
4708 indicate that the programmer forgot the parentheses in a function
4709 call; and comparisons against string literals result in unspecified
4710 behavior and are not portable in C, so they usually indicate that the
4711 programmer intended to use @code{strcmp}. This warning is enabled by
4715 @opindex Wlogical-op
4716 @opindex Wno-logical-op
4717 Warn about suspicious uses of logical operators in expressions.
4718 This includes using logical operators in contexts where a
4719 bit-wise operator is likely to be expected.
4721 @item -Wlogical-not-parentheses
4722 @opindex Wlogical-not-parentheses
4723 @opindex Wno-logical-not-parentheses
4724 Warn about logical not used on the left hand side operand of a comparison.
4725 This option does not warn if the LHS or RHS operand is of a boolean or
4726 a vector type. Its purpose is to detect suspicious code like the following:
4730 if (!a > 1) @{ @dots{} @}
4733 It is possible to suppress the warning by wrapping the LHS into
4736 if ((!a) > 1) @{ @dots{} @}
4739 @item -Waggregate-return
4740 @opindex Waggregate-return
4741 @opindex Wno-aggregate-return
4742 Warn if any functions that return structures or unions are defined or
4743 called. (In languages where you can return an array, this also elicits
4746 @item -Wno-aggressive-loop-optimizations
4747 @opindex Wno-aggressive-loop-optimizations
4748 @opindex Waggressive-loop-optimizations
4749 Warn if in a loop with constant number of iterations the compiler detects
4750 undefined behavior in some statement during one or more of the iterations.
4752 @item -Wno-attributes
4753 @opindex Wno-attributes
4754 @opindex Wattributes
4755 Do not warn if an unexpected @code{__attribute__} is used, such as
4756 unrecognized attributes, function attributes applied to variables,
4757 etc. This does not stop errors for incorrect use of supported
4760 @item -Wno-builtin-macro-redefined
4761 @opindex Wno-builtin-macro-redefined
4762 @opindex Wbuiltin-macro-redefined
4763 Do not warn if certain built-in macros are redefined. This suppresses
4764 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4765 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4767 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4768 @opindex Wstrict-prototypes
4769 @opindex Wno-strict-prototypes
4770 Warn if a function is declared or defined without specifying the
4771 argument types. (An old-style function definition is permitted without
4772 a warning if preceded by a declaration that specifies the argument
4775 @item -Wold-style-declaration @r{(C and Objective-C only)}
4776 @opindex Wold-style-declaration
4777 @opindex Wno-old-style-declaration
4778 Warn for obsolescent usages, according to the C Standard, in a
4779 declaration. For example, warn if storage-class specifiers like
4780 @code{static} are not the first things in a declaration. This warning
4781 is also enabled by @option{-Wextra}.
4783 @item -Wold-style-definition @r{(C and Objective-C only)}
4784 @opindex Wold-style-definition
4785 @opindex Wno-old-style-definition
4786 Warn if an old-style function definition is used. A warning is given
4787 even if there is a previous prototype.
4789 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4790 @opindex Wmissing-parameter-type
4791 @opindex Wno-missing-parameter-type
4792 A function parameter is declared without a type specifier in K&R-style
4799 This warning is also enabled by @option{-Wextra}.
4801 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4802 @opindex Wmissing-prototypes
4803 @opindex Wno-missing-prototypes
4804 Warn if a global function is defined without a previous prototype
4805 declaration. This warning is issued even if the definition itself
4806 provides a prototype. Use this option to detect global functions
4807 that do not have a matching prototype declaration in a header file.
4808 This option is not valid for C++ because all function declarations
4809 provide prototypes and a non-matching declaration will declare an
4810 overload rather than conflict with an earlier declaration.
4811 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
4813 @item -Wmissing-declarations
4814 @opindex Wmissing-declarations
4815 @opindex Wno-missing-declarations
4816 Warn if a global function is defined without a previous declaration.
4817 Do so even if the definition itself provides a prototype.
4818 Use this option to detect global functions that are not declared in
4819 header files. In C, no warnings are issued for functions with previous
4820 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
4821 missing prototypes. In C++, no warnings are issued for function templates,
4822 or for inline functions, or for functions in anonymous namespaces.
4824 @item -Wmissing-field-initializers
4825 @opindex Wmissing-field-initializers
4826 @opindex Wno-missing-field-initializers
4830 Warn if a structure's initializer has some fields missing. For
4831 example, the following code causes such a warning, because
4832 @code{x.h} is implicitly zero:
4835 struct s @{ int f, g, h; @};
4836 struct s x = @{ 3, 4 @};
4839 This option does not warn about designated initializers, so the following
4840 modification does not trigger a warning:
4843 struct s @{ int f, g, h; @};
4844 struct s x = @{ .f = 3, .g = 4 @};
4847 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4848 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
4850 @item -Wno-multichar
4851 @opindex Wno-multichar
4853 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4854 Usually they indicate a typo in the user's code, as they have
4855 implementation-defined values, and should not be used in portable code.
4857 @item -Wnormalized=<none|id|nfc|nfkc>
4858 @opindex Wnormalized=
4861 @cindex character set, input normalization
4862 In ISO C and ISO C++, two identifiers are different if they are
4863 different sequences of characters. However, sometimes when characters
4864 outside the basic ASCII character set are used, you can have two
4865 different character sequences that look the same. To avoid confusion,
4866 the ISO 10646 standard sets out some @dfn{normalization rules} which
4867 when applied ensure that two sequences that look the same are turned into
4868 the same sequence. GCC can warn you if you are using identifiers that
4869 have not been normalized; this option controls that warning.
4871 There are four levels of warning supported by GCC@. The default is
4872 @option{-Wnormalized=nfc}, which warns about any identifier that is
4873 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4874 recommended form for most uses.
4876 Unfortunately, there are some characters allowed in identifiers by
4877 ISO C and ISO C++ that, when turned into NFC, are not allowed in
4878 identifiers. That is, there's no way to use these symbols in portable
4879 ISO C or C++ and have all your identifiers in NFC@.
4880 @option{-Wnormalized=id} suppresses the warning for these characters.
4881 It is hoped that future versions of the standards involved will correct
4882 this, which is why this option is not the default.
4884 You can switch the warning off for all characters by writing
4885 @option{-Wnormalized=none}. You should only do this if you
4886 are using some other normalization scheme (like ``D''), because
4887 otherwise you can easily create bugs that are literally impossible to see.
4889 Some characters in ISO 10646 have distinct meanings but look identical
4890 in some fonts or display methodologies, especially once formatting has
4891 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4892 LETTER N'', displays just like a regular @code{n} that has been
4893 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4894 normalization scheme to convert all these into a standard form as
4895 well, and GCC warns if your code is not in NFKC if you use
4896 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4897 about every identifier that contains the letter O because it might be
4898 confused with the digit 0, and so is not the default, but may be
4899 useful as a local coding convention if the programming environment
4900 cannot be fixed to display these characters distinctly.
4902 @item -Wno-deprecated
4903 @opindex Wno-deprecated
4904 @opindex Wdeprecated
4905 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4907 @item -Wno-deprecated-declarations
4908 @opindex Wno-deprecated-declarations
4909 @opindex Wdeprecated-declarations
4910 Do not warn about uses of functions (@pxref{Function Attributes}),
4911 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4912 Attributes}) marked as deprecated by using the @code{deprecated}
4916 @opindex Wno-overflow
4918 Do not warn about compile-time overflow in constant expressions.
4923 Warn about One Definition Rule violations during link-time optimization.
4924 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
4927 @opindex Wopenm-simd
4928 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
4929 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
4930 be used to relax the cost model.
4932 @item -Woverride-init @r{(C and Objective-C only)}
4933 @opindex Woverride-init
4934 @opindex Wno-override-init
4938 Warn if an initialized field without side effects is overridden when
4939 using designated initializers (@pxref{Designated Inits, , Designated
4942 This warning is included in @option{-Wextra}. To get other
4943 @option{-Wextra} warnings without this one, use @option{-Wextra
4944 -Wno-override-init}.
4949 Warn if a structure is given the packed attribute, but the packed
4950 attribute has no effect on the layout or size of the structure.
4951 Such structures may be mis-aligned for little benefit. For
4952 instance, in this code, the variable @code{f.x} in @code{struct bar}
4953 is misaligned even though @code{struct bar} does not itself
4954 have the packed attribute:
4961 @} __attribute__((packed));
4969 @item -Wpacked-bitfield-compat
4970 @opindex Wpacked-bitfield-compat
4971 @opindex Wno-packed-bitfield-compat
4972 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4973 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4974 the change can lead to differences in the structure layout. GCC
4975 informs you when the offset of such a field has changed in GCC 4.4.
4976 For example there is no longer a 4-bit padding between field @code{a}
4977 and @code{b} in this structure:
4984 @} __attribute__ ((packed));
4987 This warning is enabled by default. Use
4988 @option{-Wno-packed-bitfield-compat} to disable this warning.
4993 Warn if padding is included in a structure, either to align an element
4994 of the structure or to align the whole structure. Sometimes when this
4995 happens it is possible to rearrange the fields of the structure to
4996 reduce the padding and so make the structure smaller.
4998 @item -Wredundant-decls
4999 @opindex Wredundant-decls
5000 @opindex Wno-redundant-decls
5001 Warn if anything is declared more than once in the same scope, even in
5002 cases where multiple declaration is valid and changes nothing.
5004 @item -Wnested-externs @r{(C and Objective-C only)}
5005 @opindex Wnested-externs
5006 @opindex Wno-nested-externs
5007 Warn if an @code{extern} declaration is encountered within a function.
5009 @item -Wno-inherited-variadic-ctor
5010 @opindex Winherited-variadic-ctor
5011 @opindex Wno-inherited-variadic-ctor
5012 Suppress warnings about use of C++11 inheriting constructors when the
5013 base class inherited from has a C variadic constructor; the warning is
5014 on by default because the ellipsis is not inherited.
5019 Warn if a function that is declared as inline cannot be inlined.
5020 Even with this option, the compiler does not warn about failures to
5021 inline functions declared in system headers.
5023 The compiler uses a variety of heuristics to determine whether or not
5024 to inline a function. For example, the compiler takes into account
5025 the size of the function being inlined and the amount of inlining
5026 that has already been done in the current function. Therefore,
5027 seemingly insignificant changes in the source program can cause the
5028 warnings produced by @option{-Winline} to appear or disappear.
5030 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5031 @opindex Wno-invalid-offsetof
5032 @opindex Winvalid-offsetof
5033 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
5034 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
5035 to a non-POD type is undefined. In existing C++ implementations,
5036 however, @samp{offsetof} typically gives meaningful results even when
5037 applied to certain kinds of non-POD types (such as a simple
5038 @samp{struct} that fails to be a POD type only by virtue of having a
5039 constructor). This flag is for users who are aware that they are
5040 writing nonportable code and who have deliberately chosen to ignore the
5043 The restrictions on @samp{offsetof} may be relaxed in a future version
5044 of the C++ standard.
5046 @item -Wno-int-to-pointer-cast
5047 @opindex Wno-int-to-pointer-cast
5048 @opindex Wint-to-pointer-cast
5049 Suppress warnings from casts to pointer type of an integer of a
5050 different size. In C++, casting to a pointer type of smaller size is
5051 an error. @option{Wint-to-pointer-cast} is enabled by default.
5054 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5055 @opindex Wno-pointer-to-int-cast
5056 @opindex Wpointer-to-int-cast
5057 Suppress warnings from casts from a pointer to an integer type of a
5061 @opindex Winvalid-pch
5062 @opindex Wno-invalid-pch
5063 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5064 the search path but can't be used.
5068 @opindex Wno-long-long
5069 Warn if @samp{long long} type is used. This is enabled by either
5070 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5071 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5073 @item -Wvariadic-macros
5074 @opindex Wvariadic-macros
5075 @opindex Wno-variadic-macros
5076 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
5077 alternate syntax when in pedantic ISO C99 mode. This is default.
5078 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
5082 @opindex Wno-varargs
5083 Warn upon questionable usage of the macros used to handle variable
5084 arguments like @samp{va_start}. This is default. To inhibit the
5085 warning messages, use @option{-Wno-varargs}.
5087 @item -Wvector-operation-performance
5088 @opindex Wvector-operation-performance
5089 @opindex Wno-vector-operation-performance
5090 Warn if vector operation is not implemented via SIMD capabilities of the
5091 architecture. Mainly useful for the performance tuning.
5092 Vector operation can be implemented @code{piecewise}, which means that the
5093 scalar operation is performed on every vector element;
5094 @code{in parallel}, which means that the vector operation is implemented
5095 using scalars of wider type, which normally is more performance efficient;
5096 and @code{as a single scalar}, which means that vector fits into a
5099 @item -Wno-virtual-move-assign
5100 @opindex Wvirtual-move-assign
5101 @opindex Wno-virtual-move-assign
5102 Suppress warnings about inheriting from a virtual base with a
5103 non-trivial C++11 move assignment operator. This is dangerous because
5104 if the virtual base is reachable along more than one path, it will be
5105 moved multiple times, which can mean both objects end up in the
5106 moved-from state. If the move assignment operator is written to avoid
5107 moving from a moved-from object, this warning can be disabled.
5112 Warn if variable length array is used in the code.
5113 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5114 the variable length array.
5116 @item -Wvolatile-register-var
5117 @opindex Wvolatile-register-var
5118 @opindex Wno-volatile-register-var
5119 Warn if a register variable is declared volatile. The volatile
5120 modifier does not inhibit all optimizations that may eliminate reads
5121 and/or writes to register variables. This warning is enabled by
5124 @item -Wdisabled-optimization
5125 @opindex Wdisabled-optimization
5126 @opindex Wno-disabled-optimization
5127 Warn if a requested optimization pass is disabled. This warning does
5128 not generally indicate that there is anything wrong with your code; it
5129 merely indicates that GCC's optimizers are unable to handle the code
5130 effectively. Often, the problem is that your code is too big or too
5131 complex; GCC refuses to optimize programs when the optimization
5132 itself is likely to take inordinate amounts of time.
5134 @item -Wpointer-sign @r{(C and Objective-C only)}
5135 @opindex Wpointer-sign
5136 @opindex Wno-pointer-sign
5137 Warn for pointer argument passing or assignment with different signedness.
5138 This option is only supported for C and Objective-C@. It is implied by
5139 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5140 @option{-Wno-pointer-sign}.
5142 @item -Wstack-protector
5143 @opindex Wstack-protector
5144 @opindex Wno-stack-protector
5145 This option is only active when @option{-fstack-protector} is active. It
5146 warns about functions that are not protected against stack smashing.
5148 @item -Woverlength-strings
5149 @opindex Woverlength-strings
5150 @opindex Wno-overlength-strings
5151 Warn about string constants that are longer than the ``minimum
5152 maximum'' length specified in the C standard. Modern compilers
5153 generally allow string constants that are much longer than the
5154 standard's minimum limit, but very portable programs should avoid
5155 using longer strings.
5157 The limit applies @emph{after} string constant concatenation, and does
5158 not count the trailing NUL@. In C90, the limit was 509 characters; in
5159 C99, it was raised to 4095. C++98 does not specify a normative
5160 minimum maximum, so we do not diagnose overlength strings in C++@.
5162 This option is implied by @option{-Wpedantic}, and can be disabled with
5163 @option{-Wno-overlength-strings}.
5165 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5166 @opindex Wunsuffixed-float-constants
5168 Issue a warning for any floating constant that does not have
5169 a suffix. When used together with @option{-Wsystem-headers} it
5170 warns about such constants in system header files. This can be useful
5171 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5172 from the decimal floating-point extension to C99.
5174 @item -Wno-designated-init @r{(C and Objective-C only)}
5175 Suppress warnings when a positional initializer is used to initialize
5176 a structure that has been marked with the @code{designated_init}
5181 @node Debugging Options
5182 @section Options for Debugging Your Program or GCC
5183 @cindex options, debugging
5184 @cindex debugging information options
5186 GCC has various special options that are used for debugging
5187 either your program or GCC:
5192 Produce debugging information in the operating system's native format
5193 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5196 On most systems that use stabs format, @option{-g} enables use of extra
5197 debugging information that only GDB can use; this extra information
5198 makes debugging work better in GDB but probably makes other debuggers
5200 refuse to read the program. If you want to control for certain whether
5201 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5202 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5204 GCC allows you to use @option{-g} with
5205 @option{-O}. The shortcuts taken by optimized code may occasionally
5206 produce surprising results: some variables you declared may not exist
5207 at all; flow of control may briefly move where you did not expect it;
5208 some statements may not be executed because they compute constant
5209 results or their values are already at hand; some statements may
5210 execute in different places because they have been moved out of loops.
5212 Nevertheless it proves possible to debug optimized output. This makes
5213 it reasonable to use the optimizer for programs that might have bugs.
5215 The following options are useful when GCC is generated with the
5216 capability for more than one debugging format.
5219 @opindex gsplit-dwarf
5220 Separate as much dwarf debugging information as possible into a
5221 separate output file with the extension .dwo. This option allows
5222 the build system to avoid linking files with debug information. To
5223 be useful, this option requires a debugger capable of reading .dwo
5228 Produce debugging information for use by GDB@. This means to use the
5229 most expressive format available (DWARF 2, stabs, or the native format
5230 if neither of those are supported), including GDB extensions if at all
5235 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5237 @item -ggnu-pubnames
5238 @opindex ggnu-pubnames
5239 Generate .debug_pubnames and .debug_pubtypes sections in a format
5240 suitable for conversion into a GDB@ index. This option is only useful
5241 with a linker that can produce GDB@ index version 7.
5245 Produce debugging information in stabs format (if that is supported),
5246 without GDB extensions. This is the format used by DBX on most BSD
5247 systems. On MIPS, Alpha and System V Release 4 systems this option
5248 produces stabs debugging output that is not understood by DBX or SDB@.
5249 On System V Release 4 systems this option requires the GNU assembler.
5251 @item -feliminate-unused-debug-symbols
5252 @opindex feliminate-unused-debug-symbols
5253 Produce debugging information in stabs format (if that is supported),
5254 for only symbols that are actually used.
5256 @item -femit-class-debug-always
5257 Instead of emitting debugging information for a C++ class in only one
5258 object file, emit it in all object files using the class. This option
5259 should be used only with debuggers that are unable to handle the way GCC
5260 normally emits debugging information for classes because using this
5261 option increases the size of debugging information by as much as a
5264 @item -fdebug-types-section
5265 @opindex fdebug-types-section
5266 @opindex fno-debug-types-section
5267 When using DWARF Version 4 or higher, type DIEs can be put into
5268 their own @code{.debug_types} section instead of making them part of the
5269 @code{.debug_info} section. It is more efficient to put them in a separate
5270 comdat sections since the linker can then remove duplicates.
5271 But not all DWARF consumers support @code{.debug_types} sections yet
5272 and on some objects @code{.debug_types} produces larger instead of smaller
5273 debugging information.
5277 Produce debugging information in stabs format (if that is supported),
5278 using GNU extensions understood only by the GNU debugger (GDB)@. The
5279 use of these extensions is likely to make other debuggers crash or
5280 refuse to read the program.
5284 Produce debugging information in COFF format (if that is supported).
5285 This is the format used by SDB on most System V systems prior to
5290 Produce debugging information in XCOFF format (if that is supported).
5291 This is the format used by the DBX debugger on IBM RS/6000 systems.
5295 Produce debugging information in XCOFF format (if that is supported),
5296 using GNU extensions understood only by the GNU debugger (GDB)@. The
5297 use of these extensions is likely to make other debuggers crash or
5298 refuse to read the program, and may cause assemblers other than the GNU
5299 assembler (GAS) to fail with an error.
5301 @item -gdwarf-@var{version}
5302 @opindex gdwarf-@var{version}
5303 Produce debugging information in DWARF format (if that is supported).
5304 The value of @var{version} may be either 2, 3 or 4; the default version
5305 for most targets is 4.
5307 Note that with DWARF Version 2, some ports require and always
5308 use some non-conflicting DWARF 3 extensions in the unwind tables.
5310 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5311 for maximum benefit.
5313 @item -grecord-gcc-switches
5314 @opindex grecord-gcc-switches
5315 This switch causes the command-line options used to invoke the
5316 compiler that may affect code generation to be appended to the
5317 DW_AT_producer attribute in DWARF debugging information. The options
5318 are concatenated with spaces separating them from each other and from
5319 the compiler version. See also @option{-frecord-gcc-switches} for another
5320 way of storing compiler options into the object file. This is the default.
5322 @item -gno-record-gcc-switches
5323 @opindex gno-record-gcc-switches
5324 Disallow appending command-line options to the DW_AT_producer attribute
5325 in DWARF debugging information.
5327 @item -gstrict-dwarf
5328 @opindex gstrict-dwarf
5329 Disallow using extensions of later DWARF standard version than selected
5330 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5331 DWARF extensions from later standard versions is allowed.
5333 @item -gno-strict-dwarf
5334 @opindex gno-strict-dwarf
5335 Allow using extensions of later DWARF standard version than selected with
5336 @option{-gdwarf-@var{version}}.
5338 @item -gz@r{[}=@var{type}@r{]}
5340 Produce compressed debug sections in DWARF format, if that is supported.
5341 If @var{type} is not given, the default type depends on the capabilities
5342 of the assembler and linker used. @var{type} may be one of
5343 @option{none} (don't compress debug sections), @option{zlib} (use zlib
5344 compression in ELF gABI format), or @option{zlib-gnu} (use zlib
5345 compression in traditional GNU format). If the linker doesn't support
5346 writing compressed debug sections, the option is rejected. Otherwise,
5347 if the assembler does not support them, @option{-gz} is silently ignored
5348 when producing object files.
5352 Produce debugging information in Alpha/VMS debug format (if that is
5353 supported). This is the format used by DEBUG on Alpha/VMS systems.
5356 @itemx -ggdb@var{level}
5357 @itemx -gstabs@var{level}
5358 @itemx -gcoff@var{level}
5359 @itemx -gxcoff@var{level}
5360 @itemx -gvms@var{level}
5361 Request debugging information and also use @var{level} to specify how
5362 much information. The default level is 2.
5364 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5367 Level 1 produces minimal information, enough for making backtraces in
5368 parts of the program that you don't plan to debug. This includes
5369 descriptions of functions and external variables, and line number
5370 tables, but no information about local variables.
5372 Level 3 includes extra information, such as all the macro definitions
5373 present in the program. Some debuggers support macro expansion when
5374 you use @option{-g3}.
5376 @option{-gdwarf-2} does not accept a concatenated debug level, because
5377 GCC used to support an option @option{-gdwarf} that meant to generate
5378 debug information in version 1 of the DWARF format (which is very
5379 different from version 2), and it would have been too confusing. That
5380 debug format is long obsolete, but the option cannot be changed now.
5381 Instead use an additional @option{-g@var{level}} option to change the
5382 debug level for DWARF.
5386 Turn off generation of debug info, if leaving out this option
5387 generates it, or turn it on at level 2 otherwise. The position of this
5388 argument in the command line does not matter; it takes effect after all
5389 other options are processed, and it does so only once, no matter how
5390 many times it is given. This is mainly intended to be used with
5391 @option{-fcompare-debug}.
5393 @item -fsanitize=address
5394 @opindex fsanitize=address
5395 Enable AddressSanitizer, a fast memory error detector.
5396 Memory access instructions will be instrumented to detect
5397 out-of-bounds and use-after-free bugs.
5398 See @uref{http://code.google.com/p/address-sanitizer/} for
5399 more details. The run-time behavior can be influenced using the
5400 @env{ASAN_OPTIONS} environment variable; see
5401 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5402 a list of supported options.
5404 @item -fsanitize=kernel-address
5405 @opindex fsanitize=kernel-address
5406 Enable AddressSanitizer for Linux kernel.
5407 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5409 @item -fsanitize=thread
5410 @opindex fsanitize=thread
5411 Enable ThreadSanitizer, a fast data race detector.
5412 Memory access instructions will be instrumented to detect
5413 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5414 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5415 environment variable; see
5416 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5419 @item -fsanitize=leak
5420 @opindex fsanitize=leak
5421 Enable LeakSanitizer, a memory leak detector.
5422 This option only matters for linking of executables and if neither
5423 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5424 case it will link the executable against a library that overrides @code{malloc}
5425 and other allocator functions. See
5426 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5427 details. The run-time behavior can be influenced using the
5428 @env{LSAN_OPTIONS} environment variable.
5430 @item -fsanitize=undefined
5431 @opindex fsanitize=undefined
5432 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5433 Various computations will be instrumented to detect undefined behavior
5434 at runtime. Current suboptions are:
5438 @item -fsanitize=shift
5439 @opindex fsanitize=shift
5440 This option enables checking that the result of a shift operation is
5441 not undefined. Note that what exactly is considered undefined differs
5442 slightly between C and C++, as well as between ISO C90 and C99, etc.
5444 @item -fsanitize=integer-divide-by-zero
5445 @opindex fsanitize=integer-divide-by-zero
5446 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5448 @item -fsanitize=unreachable
5449 @opindex fsanitize=unreachable
5450 With this option, the compiler will turn the @code{__builtin_unreachable}
5451 call into a diagnostics message call instead. When reaching the
5452 @code{__builtin_unreachable} call, the behavior is undefined.
5454 @item -fsanitize=vla-bound
5455 @opindex fsanitize=vla-bound
5456 This option instructs the compiler to check that the size of a variable
5457 length array is positive. This option does not have any effect in
5458 @option{-std=c++1y} mode, as the standard requires the exception be thrown
5461 @item -fsanitize=null
5462 @opindex fsanitize=null
5463 This option enables pointer checking. Particularly, the application
5464 built with this option turned on will issue an error message when it
5465 tries to dereference a NULL pointer, or if a reference (possibly an
5466 rvalue reference) is bound to a NULL pointer.
5468 @item -fsanitize=return
5469 @opindex fsanitize=return
5470 This option enables return statement checking. Programs
5471 built with this option turned on will issue an error message
5472 when the end of a non-void function is reached without actually
5473 returning a value. This option works in C++ only.
5475 @item -fsanitize=signed-integer-overflow
5476 @opindex fsanitize=signed-integer-overflow
5477 This option enables signed integer overflow checking. We check that
5478 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5479 does not overflow in the signed arithmetics. Note, integer promotion
5480 rules must be taken into account. That is, the following is not an
5483 signed char a = SCHAR_MAX;
5487 @item -fsanitize=bounds
5488 @opindex fsanitize=bounds
5489 This option enables instrumentation of array bounds. Various out of bounds
5490 accesses are detected. Flexible array members and initializers of variables
5491 with static storage are not instrumented.
5493 @item -fsanitize=float-divide-by-zero
5494 @opindex fsanitize=float-divide-by-zero
5495 Detect floating-point division by zero. Unlike other similar options,
5496 @option{-fsanitize=float-divide-by-zero} is not enabled by
5497 @option{-fsanitize=undefined}, since floating-point division by zero can
5498 be a legitimate way of obtaining infinities and NaNs.
5500 @item -fsanitize=float-cast-overflow
5501 @opindex fsanitize=float-cast-overflow
5502 This option enables floating-point type to integer conversion checking.
5503 We check that the result of the conversion does not overflow.
5504 This option does not work well with @code{FE_INVALID} exceptions enabled.
5508 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5509 @option{-fsanitize=undefined} gives a diagnostic message.
5510 This currently works only for the C family of languages.
5512 @item -fsanitize-recover
5513 @opindex fsanitize-recover
5514 By default @option{-fsanitize=undefined} sanitization (and its suboptions
5515 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return})
5516 after reporting undefined behavior attempts to continue running the
5517 program as if no undefined behavior happened. This means multiple undefined
5518 behavior runtime errors can be reported in a single program run, and the exit
5519 code of the program may indicate success even when undefined behavior
5520 has been reported. The @option{-fno-sanitize-recover} can be used to alter
5521 this behavior, only the first detected undefined behavior will be reported
5522 and program will exit after that with non-zero exit code.
5524 @item -fsanitize-undefined-trap-on-error
5525 @opindex fsanitize-undefined-trap-on-error
5526 The @option{-fsanitize-undefined-trap-on-error} instructs the compiler to
5527 report undefined behavior using @code{__builtin_trap ()} rather than
5528 a @code{libubsan} library routine. The advantage of this is that the
5529 @code{libubsan} library is not needed and will not be linked in, so this
5530 is usable even for use in freestanding environments.
5532 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5533 @opindex fdump-final-insns
5534 Dump the final internal representation (RTL) to @var{file}. If the
5535 optional argument is omitted (or if @var{file} is @code{.}), the name
5536 of the dump file is determined by appending @code{.gkd} to the
5537 compilation output file name.
5539 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5540 @opindex fcompare-debug
5541 @opindex fno-compare-debug
5542 If no error occurs during compilation, run the compiler a second time,
5543 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5544 passed to the second compilation. Dump the final internal
5545 representation in both compilations, and print an error if they differ.
5547 If the equal sign is omitted, the default @option{-gtoggle} is used.
5549 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5550 and nonzero, implicitly enables @option{-fcompare-debug}. If
5551 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5552 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5555 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5556 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5557 of the final representation and the second compilation, preventing even
5558 @env{GCC_COMPARE_DEBUG} from taking effect.
5560 To verify full coverage during @option{-fcompare-debug} testing, set
5561 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
5562 which GCC rejects as an invalid option in any actual compilation
5563 (rather than preprocessing, assembly or linking). To get just a
5564 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5565 not overridden} will do.
5567 @item -fcompare-debug-second
5568 @opindex fcompare-debug-second
5569 This option is implicitly passed to the compiler for the second
5570 compilation requested by @option{-fcompare-debug}, along with options to
5571 silence warnings, and omitting other options that would cause
5572 side-effect compiler outputs to files or to the standard output. Dump
5573 files and preserved temporary files are renamed so as to contain the
5574 @code{.gk} additional extension during the second compilation, to avoid
5575 overwriting those generated by the first.
5577 When this option is passed to the compiler driver, it causes the
5578 @emph{first} compilation to be skipped, which makes it useful for little
5579 other than debugging the compiler proper.
5581 @item -feliminate-dwarf2-dups
5582 @opindex feliminate-dwarf2-dups
5583 Compress DWARF 2 debugging information by eliminating duplicated
5584 information about each symbol. This option only makes sense when
5585 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5587 @item -femit-struct-debug-baseonly
5588 @opindex femit-struct-debug-baseonly
5589 Emit debug information for struct-like types
5590 only when the base name of the compilation source file
5591 matches the base name of file in which the struct is defined.
5593 This option substantially reduces the size of debugging information,
5594 but at significant potential loss in type information to the debugger.
5595 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5596 See @option{-femit-struct-debug-detailed} for more detailed control.
5598 This option works only with DWARF 2.
5600 @item -femit-struct-debug-reduced
5601 @opindex femit-struct-debug-reduced
5602 Emit debug information for struct-like types
5603 only when the base name of the compilation source file
5604 matches the base name of file in which the type is defined,
5605 unless the struct is a template or defined in a system header.
5607 This option significantly reduces the size of debugging information,
5608 with some potential loss in type information to the debugger.
5609 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5610 See @option{-femit-struct-debug-detailed} for more detailed control.
5612 This option works only with DWARF 2.
5614 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5615 Specify the struct-like types
5616 for which the compiler generates debug information.
5617 The intent is to reduce duplicate struct debug information
5618 between different object files within the same program.
5620 This option is a detailed version of
5621 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5622 which serves for most needs.
5624 A specification has the syntax@*
5625 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5627 The optional first word limits the specification to
5628 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5629 A struct type is used directly when it is the type of a variable, member.
5630 Indirect uses arise through pointers to structs.
5631 That is, when use of an incomplete struct is valid, the use is indirect.
5633 @samp{struct one direct; struct two * indirect;}.
5635 The optional second word limits the specification to
5636 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5637 Generic structs are a bit complicated to explain.
5638 For C++, these are non-explicit specializations of template classes,
5639 or non-template classes within the above.
5640 Other programming languages have generics,
5641 but @option{-femit-struct-debug-detailed} does not yet implement them.
5643 The third word specifies the source files for those
5644 structs for which the compiler should emit debug information.
5645 The values @samp{none} and @samp{any} have the normal meaning.
5646 The value @samp{base} means that
5647 the base of name of the file in which the type declaration appears
5648 must match the base of the name of the main compilation file.
5649 In practice, this means that when compiling @file{foo.c}, debug information
5650 is generated for types declared in that file and @file{foo.h},
5651 but not other header files.
5652 The value @samp{sys} means those types satisfying @samp{base}
5653 or declared in system or compiler headers.
5655 You may need to experiment to determine the best settings for your application.
5657 The default is @option{-femit-struct-debug-detailed=all}.
5659 This option works only with DWARF 2.
5661 @item -fno-merge-debug-strings
5662 @opindex fmerge-debug-strings
5663 @opindex fno-merge-debug-strings
5664 Direct the linker to not merge together strings in the debugging
5665 information that are identical in different object files. Merging is
5666 not supported by all assemblers or linkers. Merging decreases the size
5667 of the debug information in the output file at the cost of increasing
5668 link processing time. Merging is enabled by default.
5670 @item -fdebug-prefix-map=@var{old}=@var{new}
5671 @opindex fdebug-prefix-map
5672 When compiling files in directory @file{@var{old}}, record debugging
5673 information describing them as in @file{@var{new}} instead.
5675 @item -fno-dwarf2-cfi-asm
5676 @opindex fdwarf2-cfi-asm
5677 @opindex fno-dwarf2-cfi-asm
5678 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5679 instead of using GAS @code{.cfi_*} directives.
5681 @cindex @command{prof}
5684 Generate extra code to write profile information suitable for the
5685 analysis program @command{prof}. You must use this option when compiling
5686 the source files you want data about, and you must also use it when
5689 @cindex @command{gprof}
5692 Generate extra code to write profile information suitable for the
5693 analysis program @command{gprof}. You must use this option when compiling
5694 the source files you want data about, and you must also use it when
5699 Makes the compiler print out each function name as it is compiled, and
5700 print some statistics about each pass when it finishes.
5703 @opindex ftime-report
5704 Makes the compiler print some statistics about the time consumed by each
5705 pass when it finishes.
5708 @opindex fmem-report
5709 Makes the compiler print some statistics about permanent memory
5710 allocation when it finishes.
5712 @item -fmem-report-wpa
5713 @opindex fmem-report-wpa
5714 Makes the compiler print some statistics about permanent memory
5715 allocation for the WPA phase only.
5717 @item -fpre-ipa-mem-report
5718 @opindex fpre-ipa-mem-report
5719 @item -fpost-ipa-mem-report
5720 @opindex fpost-ipa-mem-report
5721 Makes the compiler print some statistics about permanent memory
5722 allocation before or after interprocedural optimization.
5724 @item -fprofile-report
5725 @opindex fprofile-report
5726 Makes the compiler print some statistics about consistency of the
5727 (estimated) profile and effect of individual passes.
5730 @opindex fstack-usage
5731 Makes the compiler output stack usage information for the program, on a
5732 per-function basis. The filename for the dump is made by appending
5733 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5734 the output file, if explicitly specified and it is not an executable,
5735 otherwise it is the basename of the source file. An entry is made up
5740 The name of the function.
5744 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5747 The qualifier @code{static} means that the function manipulates the stack
5748 statically: a fixed number of bytes are allocated for the frame on function
5749 entry and released on function exit; no stack adjustments are otherwise made
5750 in the function. The second field is this fixed number of bytes.
5752 The qualifier @code{dynamic} means that the function manipulates the stack
5753 dynamically: in addition to the static allocation described above, stack
5754 adjustments are made in the body of the function, for example to push/pop
5755 arguments around function calls. If the qualifier @code{bounded} is also
5756 present, the amount of these adjustments is bounded at compile time and
5757 the second field is an upper bound of the total amount of stack used by
5758 the function. If it is not present, the amount of these adjustments is
5759 not bounded at compile time and the second field only represents the
5762 @item -fprofile-arcs
5763 @opindex fprofile-arcs
5764 Add code so that program flow @dfn{arcs} are instrumented. During
5765 execution the program records how many times each branch and call is
5766 executed and how many times it is taken or returns. When the compiled
5767 program exits it saves this data to a file called
5768 @file{@var{auxname}.gcda} for each source file. The data may be used for
5769 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5770 test coverage analysis (@option{-ftest-coverage}). Each object file's
5771 @var{auxname} is generated from the name of the output file, if
5772 explicitly specified and it is not the final executable, otherwise it is
5773 the basename of the source file. In both cases any suffix is removed
5774 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5775 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5776 @xref{Cross-profiling}.
5778 @cindex @command{gcov}
5782 This option is used to compile and link code instrumented for coverage
5783 analysis. The option is a synonym for @option{-fprofile-arcs}
5784 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5785 linking). See the documentation for those options for more details.
5790 Compile the source files with @option{-fprofile-arcs} plus optimization
5791 and code generation options. For test coverage analysis, use the
5792 additional @option{-ftest-coverage} option. You do not need to profile
5793 every source file in a program.
5796 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5797 (the latter implies the former).
5800 Run the program on a representative workload to generate the arc profile
5801 information. This may be repeated any number of times. You can run
5802 concurrent instances of your program, and provided that the file system
5803 supports locking, the data files will be correctly updated. Also
5804 @code{fork} calls are detected and correctly handled (double counting
5808 For profile-directed optimizations, compile the source files again with
5809 the same optimization and code generation options plus
5810 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5811 Control Optimization}).
5814 For test coverage analysis, use @command{gcov} to produce human readable
5815 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5816 @command{gcov} documentation for further information.
5820 With @option{-fprofile-arcs}, for each function of your program GCC
5821 creates a program flow graph, then finds a spanning tree for the graph.
5822 Only arcs that are not on the spanning tree have to be instrumented: the
5823 compiler adds code to count the number of times that these arcs are
5824 executed. When an arc is the only exit or only entrance to a block, the
5825 instrumentation code can be added to the block; otherwise, a new basic
5826 block must be created to hold the instrumentation code.
5829 @item -ftest-coverage
5830 @opindex ftest-coverage
5831 Produce a notes file that the @command{gcov} code-coverage utility
5832 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5833 show program coverage. Each source file's note file is called
5834 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5835 above for a description of @var{auxname} and instructions on how to
5836 generate test coverage data. Coverage data matches the source files
5837 more closely if you do not optimize.
5839 @item -fdbg-cnt-list
5840 @opindex fdbg-cnt-list
5841 Print the name and the counter upper bound for all debug counters.
5844 @item -fdbg-cnt=@var{counter-value-list}
5846 Set the internal debug counter upper bound. @var{counter-value-list}
5847 is a comma-separated list of @var{name}:@var{value} pairs
5848 which sets the upper bound of each debug counter @var{name} to @var{value}.
5849 All debug counters have the initial upper bound of @code{UINT_MAX};
5850 thus @code{dbg_cnt()} returns true always unless the upper bound
5851 is set by this option.
5852 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
5853 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
5855 @item -fenable-@var{kind}-@var{pass}
5856 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5860 This is a set of options that are used to explicitly disable/enable
5861 optimization passes. These options are intended for use for debugging GCC.
5862 Compiler users should use regular options for enabling/disabling
5867 @item -fdisable-ipa-@var{pass}
5868 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
5869 statically invoked in the compiler multiple times, the pass name should be
5870 appended with a sequential number starting from 1.
5872 @item -fdisable-rtl-@var{pass}
5873 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
5874 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
5875 statically invoked in the compiler multiple times, the pass name should be
5876 appended with a sequential number starting from 1. @var{range-list} is a
5877 comma-separated list of function ranges or assembler names. Each range is a number
5878 pair separated by a colon. The range is inclusive in both ends. If the range
5879 is trivial, the number pair can be simplified as a single number. If the
5880 function's call graph node's @var{uid} falls within one of the specified ranges,
5881 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5882 function header of a dump file, and the pass names can be dumped by using
5883 option @option{-fdump-passes}.
5885 @item -fdisable-tree-@var{pass}
5886 @itemx -fdisable-tree-@var{pass}=@var{range-list}
5887 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5890 @item -fenable-ipa-@var{pass}
5891 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
5892 statically invoked in the compiler multiple times, the pass name should be
5893 appended with a sequential number starting from 1.
5895 @item -fenable-rtl-@var{pass}
5896 @itemx -fenable-rtl-@var{pass}=@var{range-list}
5897 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
5898 description and examples.
5900 @item -fenable-tree-@var{pass}
5901 @itemx -fenable-tree-@var{pass}=@var{range-list}
5902 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5903 of option arguments.
5907 Here are some examples showing uses of these options.
5911 # disable ccp1 for all functions
5913 # disable complete unroll for function whose cgraph node uid is 1
5914 -fenable-tree-cunroll=1
5915 # disable gcse2 for functions at the following ranges [1,1],
5916 # [300,400], and [400,1000]
5917 # disable gcse2 for functions foo and foo2
5918 -fdisable-rtl-gcse2=foo,foo2
5919 # disable early inlining
5920 -fdisable-tree-einline
5921 # disable ipa inlining
5922 -fdisable-ipa-inline
5923 # enable tree full unroll
5924 -fenable-tree-unroll
5928 @item -d@var{letters}
5929 @itemx -fdump-rtl-@var{pass}
5930 @itemx -fdump-rtl-@var{pass}=@var{filename}
5932 @opindex fdump-rtl-@var{pass}
5933 Says to make debugging dumps during compilation at times specified by
5934 @var{letters}. This is used for debugging the RTL-based passes of the
5935 compiler. The file names for most of the dumps are made by appending
5936 a pass number and a word to the @var{dumpname}, and the files are
5937 created in the directory of the output file. In case of
5938 @option{=@var{filename}} option, the dump is output on the given file
5939 instead of the pass numbered dump files. Note that the pass number is
5940 computed statically as passes get registered into the pass manager.
5941 Thus the numbering is not related to the dynamic order of execution of
5942 passes. In particular, a pass installed by a plugin could have a
5943 number over 200 even if it executed quite early. @var{dumpname} is
5944 generated from the name of the output file, if explicitly specified
5945 and it is not an executable, otherwise it is the basename of the
5946 source file. These switches may have different effects when
5947 @option{-E} is used for preprocessing.
5949 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5950 @option{-d} option @var{letters}. Here are the possible
5951 letters for use in @var{pass} and @var{letters}, and their meanings:
5955 @item -fdump-rtl-alignments
5956 @opindex fdump-rtl-alignments
5957 Dump after branch alignments have been computed.
5959 @item -fdump-rtl-asmcons
5960 @opindex fdump-rtl-asmcons
5961 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5963 @item -fdump-rtl-auto_inc_dec
5964 @opindex fdump-rtl-auto_inc_dec
5965 Dump after auto-inc-dec discovery. This pass is only run on
5966 architectures that have auto inc or auto dec instructions.
5968 @item -fdump-rtl-barriers
5969 @opindex fdump-rtl-barriers
5970 Dump after cleaning up the barrier instructions.
5972 @item -fdump-rtl-bbpart
5973 @opindex fdump-rtl-bbpart
5974 Dump after partitioning hot and cold basic blocks.
5976 @item -fdump-rtl-bbro
5977 @opindex fdump-rtl-bbro
5978 Dump after block reordering.
5980 @item -fdump-rtl-btl1
5981 @itemx -fdump-rtl-btl2
5982 @opindex fdump-rtl-btl2
5983 @opindex fdump-rtl-btl2
5984 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5985 after the two branch
5986 target load optimization passes.
5988 @item -fdump-rtl-bypass
5989 @opindex fdump-rtl-bypass
5990 Dump after jump bypassing and control flow optimizations.
5992 @item -fdump-rtl-combine
5993 @opindex fdump-rtl-combine
5994 Dump after the RTL instruction combination pass.
5996 @item -fdump-rtl-compgotos
5997 @opindex fdump-rtl-compgotos
5998 Dump after duplicating the computed gotos.
6000 @item -fdump-rtl-ce1
6001 @itemx -fdump-rtl-ce2
6002 @itemx -fdump-rtl-ce3
6003 @opindex fdump-rtl-ce1
6004 @opindex fdump-rtl-ce2
6005 @opindex fdump-rtl-ce3
6006 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6007 @option{-fdump-rtl-ce3} enable dumping after the three
6008 if conversion passes.
6010 @item -fdump-rtl-cprop_hardreg
6011 @opindex fdump-rtl-cprop_hardreg
6012 Dump after hard register copy propagation.
6014 @item -fdump-rtl-csa
6015 @opindex fdump-rtl-csa
6016 Dump after combining stack adjustments.
6018 @item -fdump-rtl-cse1
6019 @itemx -fdump-rtl-cse2
6020 @opindex fdump-rtl-cse1
6021 @opindex fdump-rtl-cse2
6022 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6023 the two common subexpression elimination passes.
6025 @item -fdump-rtl-dce
6026 @opindex fdump-rtl-dce
6027 Dump after the standalone dead code elimination passes.
6029 @item -fdump-rtl-dbr
6030 @opindex fdump-rtl-dbr
6031 Dump after delayed branch scheduling.
6033 @item -fdump-rtl-dce1
6034 @itemx -fdump-rtl-dce2
6035 @opindex fdump-rtl-dce1
6036 @opindex fdump-rtl-dce2
6037 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6038 the two dead store elimination passes.
6041 @opindex fdump-rtl-eh
6042 Dump after finalization of EH handling code.
6044 @item -fdump-rtl-eh_ranges
6045 @opindex fdump-rtl-eh_ranges
6046 Dump after conversion of EH handling range regions.
6048 @item -fdump-rtl-expand
6049 @opindex fdump-rtl-expand
6050 Dump after RTL generation.
6052 @item -fdump-rtl-fwprop1
6053 @itemx -fdump-rtl-fwprop2
6054 @opindex fdump-rtl-fwprop1
6055 @opindex fdump-rtl-fwprop2
6056 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6057 dumping after the two forward propagation passes.
6059 @item -fdump-rtl-gcse1
6060 @itemx -fdump-rtl-gcse2
6061 @opindex fdump-rtl-gcse1
6062 @opindex fdump-rtl-gcse2
6063 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6064 after global common subexpression elimination.
6066 @item -fdump-rtl-init-regs
6067 @opindex fdump-rtl-init-regs
6068 Dump after the initialization of the registers.
6070 @item -fdump-rtl-initvals
6071 @opindex fdump-rtl-initvals
6072 Dump after the computation of the initial value sets.
6074 @item -fdump-rtl-into_cfglayout
6075 @opindex fdump-rtl-into_cfglayout
6076 Dump after converting to cfglayout mode.
6078 @item -fdump-rtl-ira
6079 @opindex fdump-rtl-ira
6080 Dump after iterated register allocation.
6082 @item -fdump-rtl-jump
6083 @opindex fdump-rtl-jump
6084 Dump after the second jump optimization.
6086 @item -fdump-rtl-loop2
6087 @opindex fdump-rtl-loop2
6088 @option{-fdump-rtl-loop2} enables dumping after the rtl
6089 loop optimization passes.
6091 @item -fdump-rtl-mach
6092 @opindex fdump-rtl-mach
6093 Dump after performing the machine dependent reorganization pass, if that
6096 @item -fdump-rtl-mode_sw
6097 @opindex fdump-rtl-mode_sw
6098 Dump after removing redundant mode switches.
6100 @item -fdump-rtl-rnreg
6101 @opindex fdump-rtl-rnreg
6102 Dump after register renumbering.
6104 @item -fdump-rtl-outof_cfglayout
6105 @opindex fdump-rtl-outof_cfglayout
6106 Dump after converting from cfglayout mode.
6108 @item -fdump-rtl-peephole2
6109 @opindex fdump-rtl-peephole2
6110 Dump after the peephole pass.
6112 @item -fdump-rtl-postreload
6113 @opindex fdump-rtl-postreload
6114 Dump after post-reload optimizations.
6116 @item -fdump-rtl-pro_and_epilogue
6117 @opindex fdump-rtl-pro_and_epilogue
6118 Dump after generating the function prologues and epilogues.
6120 @item -fdump-rtl-sched1
6121 @itemx -fdump-rtl-sched2
6122 @opindex fdump-rtl-sched1
6123 @opindex fdump-rtl-sched2
6124 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6125 after the basic block scheduling passes.
6127 @item -fdump-rtl-ree
6128 @opindex fdump-rtl-ree
6129 Dump after sign/zero extension elimination.
6131 @item -fdump-rtl-seqabstr
6132 @opindex fdump-rtl-seqabstr
6133 Dump after common sequence discovery.
6135 @item -fdump-rtl-shorten
6136 @opindex fdump-rtl-shorten
6137 Dump after shortening branches.
6139 @item -fdump-rtl-sibling
6140 @opindex fdump-rtl-sibling
6141 Dump after sibling call optimizations.
6143 @item -fdump-rtl-split1
6144 @itemx -fdump-rtl-split2
6145 @itemx -fdump-rtl-split3
6146 @itemx -fdump-rtl-split4
6147 @itemx -fdump-rtl-split5
6148 @opindex fdump-rtl-split1
6149 @opindex fdump-rtl-split2
6150 @opindex fdump-rtl-split3
6151 @opindex fdump-rtl-split4
6152 @opindex fdump-rtl-split5
6153 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
6154 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
6155 @option{-fdump-rtl-split5} enable dumping after five rounds of
6156 instruction splitting.
6158 @item -fdump-rtl-sms
6159 @opindex fdump-rtl-sms
6160 Dump after modulo scheduling. This pass is only run on some
6163 @item -fdump-rtl-stack
6164 @opindex fdump-rtl-stack
6165 Dump after conversion from GCC's ``flat register file'' registers to the
6166 x87's stack-like registers. This pass is only run on x86 variants.
6168 @item -fdump-rtl-subreg1
6169 @itemx -fdump-rtl-subreg2
6170 @opindex fdump-rtl-subreg1
6171 @opindex fdump-rtl-subreg2
6172 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6173 the two subreg expansion passes.
6175 @item -fdump-rtl-unshare
6176 @opindex fdump-rtl-unshare
6177 Dump after all rtl has been unshared.
6179 @item -fdump-rtl-vartrack
6180 @opindex fdump-rtl-vartrack
6181 Dump after variable tracking.
6183 @item -fdump-rtl-vregs
6184 @opindex fdump-rtl-vregs
6185 Dump after converting virtual registers to hard registers.
6187 @item -fdump-rtl-web
6188 @opindex fdump-rtl-web
6189 Dump after live range splitting.
6191 @item -fdump-rtl-regclass
6192 @itemx -fdump-rtl-subregs_of_mode_init
6193 @itemx -fdump-rtl-subregs_of_mode_finish
6194 @itemx -fdump-rtl-dfinit
6195 @itemx -fdump-rtl-dfinish
6196 @opindex fdump-rtl-regclass
6197 @opindex fdump-rtl-subregs_of_mode_init
6198 @opindex fdump-rtl-subregs_of_mode_finish
6199 @opindex fdump-rtl-dfinit
6200 @opindex fdump-rtl-dfinish
6201 These dumps are defined but always produce empty files.
6204 @itemx -fdump-rtl-all
6206 @opindex fdump-rtl-all
6207 Produce all the dumps listed above.
6211 Annotate the assembler output with miscellaneous debugging information.
6215 Dump all macro definitions, at the end of preprocessing, in addition to
6220 Produce a core dump whenever an error occurs.
6224 Annotate the assembler output with a comment indicating which
6225 pattern and alternative is used. The length of each instruction is
6230 Dump the RTL in the assembler output as a comment before each instruction.
6231 Also turns on @option{-dp} annotation.
6235 Just generate RTL for a function instead of compiling it. Usually used
6236 with @option{-fdump-rtl-expand}.
6240 @opindex fdump-noaddr
6241 When doing debugging dumps, suppress address output. This makes it more
6242 feasible to use diff on debugging dumps for compiler invocations with
6243 different compiler binaries and/or different
6244 text / bss / data / heap / stack / dso start locations.
6246 @item -fdump-unnumbered
6247 @opindex fdump-unnumbered
6248 When doing debugging dumps, suppress instruction numbers and address output.
6249 This makes it more feasible to use diff on debugging dumps for compiler
6250 invocations with different options, in particular with and without
6253 @item -fdump-unnumbered-links
6254 @opindex fdump-unnumbered-links
6255 When doing debugging dumps (see @option{-d} option above), suppress
6256 instruction numbers for the links to the previous and next instructions
6259 @item -fdump-translation-unit @r{(C++ only)}
6260 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6261 @opindex fdump-translation-unit
6262 Dump a representation of the tree structure for the entire translation
6263 unit to a file. The file name is made by appending @file{.tu} to the
6264 source file name, and the file is created in the same directory as the
6265 output file. If the @samp{-@var{options}} form is used, @var{options}
6266 controls the details of the dump as described for the
6267 @option{-fdump-tree} options.
6269 @item -fdump-class-hierarchy @r{(C++ only)}
6270 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6271 @opindex fdump-class-hierarchy
6272 Dump a representation of each class's hierarchy and virtual function
6273 table layout to a file. The file name is made by appending
6274 @file{.class} to the source file name, and the file is created in the
6275 same directory as the output file. If the @samp{-@var{options}} form
6276 is used, @var{options} controls the details of the dump as described
6277 for the @option{-fdump-tree} options.
6279 @item -fdump-ipa-@var{switch}
6281 Control the dumping at various stages of inter-procedural analysis
6282 language tree to a file. The file name is generated by appending a
6283 switch specific suffix to the source file name, and the file is created
6284 in the same directory as the output file. The following dumps are
6289 Enables all inter-procedural analysis dumps.
6292 Dumps information about call-graph optimization, unused function removal,
6293 and inlining decisions.
6296 Dump after function inlining.
6301 @opindex fdump-passes
6302 Dump the list of optimization passes that are turned on and off by
6303 the current command-line options.
6305 @item -fdump-statistics-@var{option}
6306 @opindex fdump-statistics
6307 Enable and control dumping of pass statistics in a separate file. The
6308 file name is generated by appending a suffix ending in
6309 @samp{.statistics} to the source file name, and the file is created in
6310 the same directory as the output file. If the @samp{-@var{option}}
6311 form is used, @samp{-stats} causes counters to be summed over the
6312 whole compilation unit while @samp{-details} dumps every event as
6313 the passes generate them. The default with no option is to sum
6314 counters for each function compiled.
6316 @item -fdump-tree-@var{switch}
6317 @itemx -fdump-tree-@var{switch}-@var{options}
6318 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6320 Control the dumping at various stages of processing the intermediate
6321 language tree to a file. The file name is generated by appending a
6322 switch-specific suffix to the source file name, and the file is
6323 created in the same directory as the output file. In case of
6324 @option{=@var{filename}} option, the dump is output on the given file
6325 instead of the auto named dump files. If the @samp{-@var{options}}
6326 form is used, @var{options} is a list of @samp{-} separated options
6327 which control the details of the dump. Not all options are applicable
6328 to all dumps; those that are not meaningful are ignored. The
6329 following options are available
6333 Print the address of each node. Usually this is not meaningful as it
6334 changes according to the environment and source file. Its primary use
6335 is for tying up a dump file with a debug environment.
6337 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6338 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6339 use working backward from mangled names in the assembly file.
6341 When dumping front-end intermediate representations, inhibit dumping
6342 of members of a scope or body of a function merely because that scope
6343 has been reached. Only dump such items when they are directly reachable
6346 When dumping pretty-printed trees, this option inhibits dumping the
6347 bodies of control structures.
6349 When dumping RTL, print the RTL in slim (condensed) form instead of
6350 the default LISP-like representation.
6352 Print a raw representation of the tree. By default, trees are
6353 pretty-printed into a C-like representation.
6355 Enable more detailed dumps (not honored by every dump option). Also
6356 include information from the optimization passes.
6358 Enable dumping various statistics about the pass (not honored by every dump
6361 Enable showing basic block boundaries (disabled in raw dumps).
6363 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6364 dump a representation of the control flow graph suitable for viewing with
6365 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6366 the file is pretty-printed as a subgraph, so that GraphViz can render them
6367 all in a single plot.
6369 This option currently only works for RTL dumps, and the RTL is always
6370 dumped in slim form.
6372 Enable showing virtual operands for every statement.
6374 Enable showing line numbers for statements.
6376 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6378 Enable showing the tree dump for each statement.
6380 Enable showing the EH region number holding each statement.
6382 Enable showing scalar evolution analysis details.
6384 Enable showing optimization information (only available in certain
6387 Enable showing missed optimization information (only available in certain
6390 Enable other detailed optimization information (only available in
6392 @item =@var{filename}
6393 Instead of an auto named dump file, output into the given file
6394 name. The file names @file{stdout} and @file{stderr} are treated
6395 specially and are considered already open standard streams. For
6399 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6400 -fdump-tree-pre=stderr file.c
6403 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6404 output on to @file{stderr}. If two conflicting dump filenames are
6405 given for the same pass, then the latter option overrides the earlier
6409 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6410 and @option{lineno}.
6413 Turn on all optimization options, i.e., @option{optimized},
6414 @option{missed}, and @option{note}.
6417 The following tree dumps are possible:
6421 @opindex fdump-tree-original
6422 Dump before any tree based optimization, to @file{@var{file}.original}.
6425 @opindex fdump-tree-optimized
6426 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6429 @opindex fdump-tree-gimple
6430 Dump each function before and after the gimplification pass to a file. The
6431 file name is made by appending @file{.gimple} to the source file name.
6434 @opindex fdump-tree-cfg
6435 Dump the control flow graph of each function to a file. The file name is
6436 made by appending @file{.cfg} to the source file name.
6439 @opindex fdump-tree-ch
6440 Dump each function after copying loop headers. The file name is made by
6441 appending @file{.ch} to the source file name.
6444 @opindex fdump-tree-ssa
6445 Dump SSA related information to a file. The file name is made by appending
6446 @file{.ssa} to the source file name.
6449 @opindex fdump-tree-alias
6450 Dump aliasing information for each function. The file name is made by
6451 appending @file{.alias} to the source file name.
6454 @opindex fdump-tree-ccp
6455 Dump each function after CCP@. The file name is made by appending
6456 @file{.ccp} to the source file name.
6459 @opindex fdump-tree-storeccp
6460 Dump each function after STORE-CCP@. The file name is made by appending
6461 @file{.storeccp} to the source file name.
6464 @opindex fdump-tree-pre
6465 Dump trees after partial redundancy elimination. The file name is made
6466 by appending @file{.pre} to the source file name.
6469 @opindex fdump-tree-fre
6470 Dump trees after full redundancy elimination. The file name is made
6471 by appending @file{.fre} to the source file name.
6474 @opindex fdump-tree-copyprop
6475 Dump trees after copy propagation. The file name is made
6476 by appending @file{.copyprop} to the source file name.
6478 @item store_copyprop
6479 @opindex fdump-tree-store_copyprop
6480 Dump trees after store copy-propagation. The file name is made
6481 by appending @file{.store_copyprop} to the source file name.
6484 @opindex fdump-tree-dce
6485 Dump each function after dead code elimination. The file name is made by
6486 appending @file{.dce} to the source file name.
6489 @opindex fdump-tree-sra
6490 Dump each function after performing scalar replacement of aggregates. The
6491 file name is made by appending @file{.sra} to the source file name.
6494 @opindex fdump-tree-sink
6495 Dump each function after performing code sinking. The file name is made
6496 by appending @file{.sink} to the source file name.
6499 @opindex fdump-tree-dom
6500 Dump each function after applying dominator tree optimizations. The file
6501 name is made by appending @file{.dom} to the source file name.
6504 @opindex fdump-tree-dse
6505 Dump each function after applying dead store elimination. The file
6506 name is made by appending @file{.dse} to the source file name.
6509 @opindex fdump-tree-phiopt
6510 Dump each function after optimizing PHI nodes into straightline code. The file
6511 name is made by appending @file{.phiopt} to the source file name.
6514 @opindex fdump-tree-forwprop
6515 Dump each function after forward propagating single use variables. The file
6516 name is made by appending @file{.forwprop} to the source file name.
6519 @opindex fdump-tree-copyrename
6520 Dump each function after applying the copy rename optimization. The file
6521 name is made by appending @file{.copyrename} to the source file name.
6524 @opindex fdump-tree-nrv
6525 Dump each function after applying the named return value optimization on
6526 generic trees. The file name is made by appending @file{.nrv} to the source
6530 @opindex fdump-tree-vect
6531 Dump each function after applying vectorization of loops. The file name is
6532 made by appending @file{.vect} to the source file name.
6535 @opindex fdump-tree-slp
6536 Dump each function after applying vectorization of basic blocks. The file name
6537 is made by appending @file{.slp} to the source file name.
6540 @opindex fdump-tree-vrp
6541 Dump each function after Value Range Propagation (VRP). The file name
6542 is made by appending @file{.vrp} to the source file name.
6545 @opindex fdump-tree-all
6546 Enable all the available tree dumps with the flags provided in this option.
6550 @itemx -fopt-info-@var{options}
6551 @itemx -fopt-info-@var{options}=@var{filename}
6553 Controls optimization dumps from various optimization passes. If the
6554 @samp{-@var{options}} form is used, @var{options} is a list of
6555 @samp{-} separated options to select the dump details and
6556 optimizations. If @var{options} is not specified, it defaults to
6557 @option{optimized} for details and @option{optall} for optimization
6558 groups. If the @var{filename} is not specified, it defaults to
6559 @file{stderr}. Note that the output @var{filename} will be overwritten
6560 in case of multiple translation units. If a combined output from
6561 multiple translation units is desired, @file{stderr} should be used
6564 The options can be divided into two groups, 1) options describing the
6565 verbosity of the dump, and 2) options describing which optimizations
6566 should be included. The options from both the groups can be freely
6567 mixed as they are non-overlapping. However, in case of any conflicts,
6568 the latter options override the earlier options on the command
6569 line. Though multiple -fopt-info options are accepted, only one of
6570 them can have @option{=filename}. If other filenames are provided then
6571 all but the first one are ignored.
6573 The dump verbosity has the following options
6577 Print information when an optimization is successfully applied. It is
6578 up to a pass to decide which information is relevant. For example, the
6579 vectorizer passes print the source location of loops which got
6580 successfully vectorized.
6582 Print information about missed optimizations. Individual passes
6583 control which information to include in the output. For example,
6586 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
6589 will print information about missed optimization opportunities from
6590 vectorization passes on stderr.
6592 Print verbose information about optimizations, such as certain
6593 transformations, more detailed messages about decisions etc.
6595 Print detailed optimization information. This includes
6596 @var{optimized}, @var{missed}, and @var{note}.
6599 The second set of options describes a group of optimizations and may
6600 include one or more of the following.
6604 Enable dumps from all interprocedural optimizations.
6606 Enable dumps from all loop optimizations.
6608 Enable dumps from all inlining optimizations.
6610 Enable dumps from all vectorization optimizations.
6612 Enable dumps from all optimizations. This is a superset of
6613 the optimization groups listed above.
6618 gcc -O3 -fopt-info-missed=missed.all
6621 outputs missed optimization report from all the passes into
6626 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
6629 will output information about missed optimizations as well as
6630 optimized locations from all the inlining passes into
6633 If the @var{filename} is provided, then the dumps from all the
6634 applicable optimizations are concatenated into the @file{filename}.
6635 Otherwise the dump is output onto @file{stderr}. If @var{options} is
6636 omitted, it defaults to @option{all-optall}, which means dump all
6637 available optimization info from all the passes. In the following
6638 example, all optimization info is output on to @file{stderr}.
6644 Note that @option{-fopt-info-vec-missed} behaves the same as
6645 @option{-fopt-info-missed-vec}.
6647 As another example, consider
6650 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
6653 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
6654 in conflict since only one output file is allowed. In this case, only
6655 the first option takes effect and the subsequent options are
6656 ignored. Thus only the @file{vec.miss} is produced which contains
6657 dumps from the vectorizer about missed opportunities.
6659 @item -frandom-seed=@var{string}
6660 @opindex frandom-seed
6661 This option provides a seed that GCC uses in place of
6662 random numbers in generating certain symbol names
6663 that have to be different in every compiled file. It is also used to
6664 place unique stamps in coverage data files and the object files that
6665 produce them. You can use the @option{-frandom-seed} option to produce
6666 reproducibly identical object files.
6668 The @var{string} should be different for every file you compile.
6670 @item -fsched-verbose=@var{n}
6671 @opindex fsched-verbose
6672 On targets that use instruction scheduling, this option controls the
6673 amount of debugging output the scheduler prints. This information is
6674 written to standard error, unless @option{-fdump-rtl-sched1} or
6675 @option{-fdump-rtl-sched2} is specified, in which case it is output
6676 to the usual dump listing file, @file{.sched1} or @file{.sched2}
6677 respectively. However for @var{n} greater than nine, the output is
6678 always printed to standard error.
6680 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
6681 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
6682 For @var{n} greater than one, it also output basic block probabilities,
6683 detailed ready list information and unit/insn info. For @var{n} greater
6684 than two, it includes RTL at abort point, control-flow and regions info.
6685 And for @var{n} over four, @option{-fsched-verbose} also includes
6689 @itemx -save-temps=cwd
6691 Store the usual ``temporary'' intermediate files permanently; place them
6692 in the current directory and name them based on the source file. Thus,
6693 compiling @file{foo.c} with @option{-c -save-temps} produces files
6694 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
6695 preprocessed @file{foo.i} output file even though the compiler now
6696 normally uses an integrated preprocessor.
6698 When used in combination with the @option{-x} command-line option,
6699 @option{-save-temps} is sensible enough to avoid over writing an
6700 input source file with the same extension as an intermediate file.
6701 The corresponding intermediate file may be obtained by renaming the
6702 source file before using @option{-save-temps}.
6704 If you invoke GCC in parallel, compiling several different source
6705 files that share a common base name in different subdirectories or the
6706 same source file compiled for multiple output destinations, it is
6707 likely that the different parallel compilers will interfere with each
6708 other, and overwrite the temporary files. For instance:
6711 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
6712 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
6715 may result in @file{foo.i} and @file{foo.o} being written to
6716 simultaneously by both compilers.
6718 @item -save-temps=obj
6719 @opindex save-temps=obj
6720 Store the usual ``temporary'' intermediate files permanently. If the
6721 @option{-o} option is used, the temporary files are based on the
6722 object file. If the @option{-o} option is not used, the
6723 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
6728 gcc -save-temps=obj -c foo.c
6729 gcc -save-temps=obj -c bar.c -o dir/xbar.o
6730 gcc -save-temps=obj foobar.c -o dir2/yfoobar
6734 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
6735 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
6736 @file{dir2/yfoobar.o}.
6738 @item -time@r{[}=@var{file}@r{]}
6740 Report the CPU time taken by each subprocess in the compilation
6741 sequence. For C source files, this is the compiler proper and assembler
6742 (plus the linker if linking is done).
6744 Without the specification of an output file, the output looks like this:
6751 The first number on each line is the ``user time'', that is time spent
6752 executing the program itself. The second number is ``system time'',
6753 time spent executing operating system routines on behalf of the program.
6754 Both numbers are in seconds.
6756 With the specification of an output file, the output is appended to the
6757 named file, and it looks like this:
6760 0.12 0.01 cc1 @var{options}
6761 0.00 0.01 as @var{options}
6764 The ``user time'' and the ``system time'' are moved before the program
6765 name, and the options passed to the program are displayed, so that one
6766 can later tell what file was being compiled, and with which options.
6768 @item -fvar-tracking
6769 @opindex fvar-tracking
6770 Run variable tracking pass. It computes where variables are stored at each
6771 position in code. Better debugging information is then generated
6772 (if the debugging information format supports this information).
6774 It is enabled by default when compiling with optimization (@option{-Os},
6775 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6776 the debug info format supports it.
6778 @item -fvar-tracking-assignments
6779 @opindex fvar-tracking-assignments
6780 @opindex fno-var-tracking-assignments
6781 Annotate assignments to user variables early in the compilation and
6782 attempt to carry the annotations over throughout the compilation all the
6783 way to the end, in an attempt to improve debug information while
6784 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6786 It can be enabled even if var-tracking is disabled, in which case
6787 annotations are created and maintained, but discarded at the end.
6789 @item -fvar-tracking-assignments-toggle
6790 @opindex fvar-tracking-assignments-toggle
6791 @opindex fno-var-tracking-assignments-toggle
6792 Toggle @option{-fvar-tracking-assignments}, in the same way that
6793 @option{-gtoggle} toggles @option{-g}.
6795 @item -print-file-name=@var{library}
6796 @opindex print-file-name
6797 Print the full absolute name of the library file @var{library} that
6798 would be used when linking---and don't do anything else. With this
6799 option, GCC does not compile or link anything; it just prints the
6802 @item -print-multi-directory
6803 @opindex print-multi-directory
6804 Print the directory name corresponding to the multilib selected by any
6805 other switches present in the command line. This directory is supposed
6806 to exist in @env{GCC_EXEC_PREFIX}.
6808 @item -print-multi-lib
6809 @opindex print-multi-lib
6810 Print the mapping from multilib directory names to compiler switches
6811 that enable them. The directory name is separated from the switches by
6812 @samp{;}, and each switch starts with an @samp{@@} instead of the
6813 @samp{-}, without spaces between multiple switches. This is supposed to
6814 ease shell processing.
6816 @item -print-multi-os-directory
6817 @opindex print-multi-os-directory
6818 Print the path to OS libraries for the selected
6819 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6820 present in the @file{lib} subdirectory and no multilibs are used, this is
6821 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6822 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6823 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6824 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6826 @item -print-multiarch
6827 @opindex print-multiarch
6828 Print the path to OS libraries for the selected multiarch,
6829 relative to some @file{lib} subdirectory.
6831 @item -print-prog-name=@var{program}
6832 @opindex print-prog-name
6833 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6835 @item -print-libgcc-file-name
6836 @opindex print-libgcc-file-name
6837 Same as @option{-print-file-name=libgcc.a}.
6839 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6840 but you do want to link with @file{libgcc.a}. You can do:
6843 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6846 @item -print-search-dirs
6847 @opindex print-search-dirs
6848 Print the name of the configured installation directory and a list of
6849 program and library directories @command{gcc} searches---and don't do anything else.
6851 This is useful when @command{gcc} prints the error message
6852 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6853 To resolve this you either need to put @file{cpp0} and the other compiler
6854 components where @command{gcc} expects to find them, or you can set the environment
6855 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6856 Don't forget the trailing @samp{/}.
6857 @xref{Environment Variables}.
6859 @item -print-sysroot
6860 @opindex print-sysroot
6861 Print the target sysroot directory that is used during
6862 compilation. This is the target sysroot specified either at configure
6863 time or using the @option{--sysroot} option, possibly with an extra
6864 suffix that depends on compilation options. If no target sysroot is
6865 specified, the option prints nothing.
6867 @item -print-sysroot-headers-suffix
6868 @opindex print-sysroot-headers-suffix
6869 Print the suffix added to the target sysroot when searching for
6870 headers, or give an error if the compiler is not configured with such
6871 a suffix---and don't do anything else.
6874 @opindex dumpmachine
6875 Print the compiler's target machine (for example,
6876 @samp{i686-pc-linux-gnu})---and don't do anything else.
6879 @opindex dumpversion
6880 Print the compiler version (for example, @samp{3.0})---and don't do
6885 Print the compiler's built-in specs---and don't do anything else. (This
6886 is used when GCC itself is being built.) @xref{Spec Files}.
6888 @item -fno-eliminate-unused-debug-types
6889 @opindex feliminate-unused-debug-types
6890 @opindex fno-eliminate-unused-debug-types
6891 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
6892 output for types that are nowhere used in the source file being compiled.
6893 Sometimes it is useful to have GCC emit debugging
6894 information for all types declared in a compilation
6895 unit, regardless of whether or not they are actually used
6896 in that compilation unit, for example
6897 if, in the debugger, you want to cast a value to a type that is
6898 not actually used in your program (but is declared). More often,
6899 however, this results in a significant amount of wasted space.
6902 @node Optimize Options
6903 @section Options That Control Optimization
6904 @cindex optimize options
6905 @cindex options, optimization
6907 These options control various sorts of optimizations.
6909 Without any optimization option, the compiler's goal is to reduce the
6910 cost of compilation and to make debugging produce the expected
6911 results. Statements are independent: if you stop the program with a
6912 breakpoint between statements, you can then assign a new value to any
6913 variable or change the program counter to any other statement in the
6914 function and get exactly the results you expect from the source
6917 Turning on optimization flags makes the compiler attempt to improve
6918 the performance and/or code size at the expense of compilation time
6919 and possibly the ability to debug the program.
6921 The compiler performs optimization based on the knowledge it has of the
6922 program. Compiling multiple files at once to a single output file mode allows
6923 the compiler to use information gained from all of the files when compiling
6926 Not all optimizations are controlled directly by a flag. Only
6927 optimizations that have a flag are listed in this section.
6929 Most optimizations are only enabled if an @option{-O} level is set on
6930 the command line. Otherwise they are disabled, even if individual
6931 optimization flags are specified.
6933 Depending on the target and how GCC was configured, a slightly different
6934 set of optimizations may be enabled at each @option{-O} level than
6935 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6936 to find out the exact set of optimizations that are enabled at each level.
6937 @xref{Overall Options}, for examples.
6944 Optimize. Optimizing compilation takes somewhat more time, and a lot
6945 more memory for a large function.
6947 With @option{-O}, the compiler tries to reduce code size and execution
6948 time, without performing any optimizations that take a great deal of
6951 @option{-O} turns on the following optimization flags:
6955 -fcprop-registers @gol
6958 -fdelayed-branch @gol
6960 -fguess-branch-probability @gol
6961 -fif-conversion2 @gol
6962 -fif-conversion @gol
6963 -fipa-pure-const @gol
6965 -fipa-reference @gol
6967 -fsplit-wide-types @gol
6969 -ftree-builtin-call-dce @gol
6973 -ftree-copyrename @gol
6975 -ftree-dominator-opts @gol
6977 -ftree-forwprop @gol
6986 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6987 where doing so does not interfere with debugging.
6991 Optimize even more. GCC performs nearly all supported optimizations
6992 that do not involve a space-speed tradeoff.
6993 As compared to @option{-O}, this option increases both compilation time
6994 and the performance of the generated code.
6996 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6997 also turns on the following optimization flags:
6998 @gccoptlist{-fthread-jumps @gol
6999 -falign-functions -falign-jumps @gol
7000 -falign-loops -falign-labels @gol
7003 -fcse-follow-jumps -fcse-skip-blocks @gol
7004 -fdelete-null-pointer-checks @gol
7005 -fdevirtualize -fdevirtualize-speculatively @gol
7006 -fexpensive-optimizations @gol
7007 -fgcse -fgcse-lm @gol
7008 -fhoist-adjacent-loads @gol
7009 -finline-small-functions @gol
7010 -findirect-inlining @gol
7012 -fisolate-erroneous-paths-dereference @gol
7013 -foptimize-sibling-calls @gol
7014 -fpartial-inlining @gol
7016 -freorder-blocks -freorder-functions @gol
7017 -frerun-cse-after-loop @gol
7018 -fsched-interblock -fsched-spec @gol
7019 -fschedule-insns -fschedule-insns2 @gol
7020 -fstrict-aliasing -fstrict-overflow @gol
7021 -ftree-switch-conversion -ftree-tail-merge @gol
7025 Please note the warning under @option{-fgcse} about
7026 invoking @option{-O2} on programs that use computed gotos.
7030 Optimize yet more. @option{-O3} turns on all optimizations specified
7031 by @option{-O2} and also turns on the @option{-finline-functions},
7032 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7033 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7034 @option{-ftree-loop-distribute-patterns},
7035 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7036 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7040 Reduce compilation time and make debugging produce the expected
7041 results. This is the default.
7045 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7046 do not typically increase code size. It also performs further
7047 optimizations designed to reduce code size.
7049 @option{-Os} disables the following optimization flags:
7050 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7051 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7052 -fprefetch-loop-arrays}
7056 Disregard strict standards compliance. @option{-Ofast} enables all
7057 @option{-O3} optimizations. It also enables optimizations that are not
7058 valid for all standard-compliant programs.
7059 It turns on @option{-ffast-math} and the Fortran-specific
7060 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7064 Optimize debugging experience. @option{-Og} enables optimizations
7065 that do not interfere with debugging. It should be the optimization
7066 level of choice for the standard edit-compile-debug cycle, offering
7067 a reasonable level of optimization while maintaining fast compilation
7068 and a good debugging experience.
7070 If you use multiple @option{-O} options, with or without level numbers,
7071 the last such option is the one that is effective.
7074 Options of the form @option{-f@var{flag}} specify machine-independent
7075 flags. Most flags have both positive and negative forms; the negative
7076 form of @option{-ffoo} is @option{-fno-foo}. In the table
7077 below, only one of the forms is listed---the one you typically
7078 use. You can figure out the other form by either removing @samp{no-}
7081 The following options control specific optimizations. They are either
7082 activated by @option{-O} options or are related to ones that are. You
7083 can use the following flags in the rare cases when ``fine-tuning'' of
7084 optimizations to be performed is desired.
7087 @item -fno-defer-pop
7088 @opindex fno-defer-pop
7089 Always pop the arguments to each function call as soon as that function
7090 returns. For machines that must pop arguments after a function call,
7091 the compiler normally lets arguments accumulate on the stack for several
7092 function calls and pops them all at once.
7094 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7096 @item -fforward-propagate
7097 @opindex fforward-propagate
7098 Perform a forward propagation pass on RTL@. The pass tries to combine two
7099 instructions and checks if the result can be simplified. If loop unrolling
7100 is active, two passes are performed and the second is scheduled after
7103 This option is enabled by default at optimization levels @option{-O},
7104 @option{-O2}, @option{-O3}, @option{-Os}.
7106 @item -ffp-contract=@var{style}
7107 @opindex ffp-contract
7108 @option{-ffp-contract=off} disables floating-point expression contraction.
7109 @option{-ffp-contract=fast} enables floating-point expression contraction
7110 such as forming of fused multiply-add operations if the target has
7111 native support for them.
7112 @option{-ffp-contract=on} enables floating-point expression contraction
7113 if allowed by the language standard. This is currently not implemented
7114 and treated equal to @option{-ffp-contract=off}.
7116 The default is @option{-ffp-contract=fast}.
7118 @item -fomit-frame-pointer
7119 @opindex fomit-frame-pointer
7120 Don't keep the frame pointer in a register for functions that
7121 don't need one. This avoids the instructions to save, set up and
7122 restore frame pointers; it also makes an extra register available
7123 in many functions. @strong{It also makes debugging impossible on
7126 On some machines, such as the VAX, this flag has no effect, because
7127 the standard calling sequence automatically handles the frame pointer
7128 and nothing is saved by pretending it doesn't exist. The
7129 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7130 whether a target machine supports this flag. @xref{Registers,,Register
7131 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7133 Starting with GCC version 4.6, the default setting (when not optimizing for
7134 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
7135 @option{-fomit-frame-pointer}. The default can be reverted to
7136 @option{-fno-omit-frame-pointer} by configuring GCC with the
7137 @option{--enable-frame-pointer} configure option.
7139 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7141 @item -foptimize-sibling-calls
7142 @opindex foptimize-sibling-calls
7143 Optimize sibling and tail recursive calls.
7145 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7149 Do not expand any functions inline apart from those marked with
7150 the @code{always_inline} attribute. This is the default when not
7153 Single functions can be exempted from inlining by marking them
7154 with the @code{noinline} attribute.
7156 @item -finline-small-functions
7157 @opindex finline-small-functions
7158 Integrate functions into their callers when their body is smaller than expected
7159 function call code (so overall size of program gets smaller). The compiler
7160 heuristically decides which functions are simple enough to be worth integrating
7161 in this way. This inlining applies to all functions, even those not declared
7164 Enabled at level @option{-O2}.
7166 @item -findirect-inlining
7167 @opindex findirect-inlining
7168 Inline also indirect calls that are discovered to be known at compile
7169 time thanks to previous inlining. This option has any effect only
7170 when inlining itself is turned on by the @option{-finline-functions}
7171 or @option{-finline-small-functions} options.
7173 Enabled at level @option{-O2}.
7175 @item -finline-functions
7176 @opindex finline-functions
7177 Consider all functions for inlining, even if they are not declared inline.
7178 The compiler heuristically decides which functions are worth integrating
7181 If all calls to a given function are integrated, and the function is
7182 declared @code{static}, then the function is normally not output as
7183 assembler code in its own right.
7185 Enabled at level @option{-O3}.
7187 @item -finline-functions-called-once
7188 @opindex finline-functions-called-once
7189 Consider all @code{static} functions called once for inlining into their
7190 caller even if they are not marked @code{inline}. If a call to a given
7191 function is integrated, then the function is not output as assembler code
7194 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7196 @item -fearly-inlining
7197 @opindex fearly-inlining
7198 Inline functions marked by @code{always_inline} and functions whose body seems
7199 smaller than the function call overhead early before doing
7200 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7201 makes profiling significantly cheaper and usually inlining faster on programs
7202 having large chains of nested wrapper functions.
7208 Perform interprocedural scalar replacement of aggregates, removal of
7209 unused parameters and replacement of parameters passed by reference
7210 by parameters passed by value.
7212 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7214 @item -finline-limit=@var{n}
7215 @opindex finline-limit
7216 By default, GCC limits the size of functions that can be inlined. This flag
7217 allows coarse control of this limit. @var{n} is the size of functions that
7218 can be inlined in number of pseudo instructions.
7220 Inlining is actually controlled by a number of parameters, which may be
7221 specified individually by using @option{--param @var{name}=@var{value}}.
7222 The @option{-finline-limit=@var{n}} option sets some of these parameters
7226 @item max-inline-insns-single
7227 is set to @var{n}/2.
7228 @item max-inline-insns-auto
7229 is set to @var{n}/2.
7232 See below for a documentation of the individual
7233 parameters controlling inlining and for the defaults of these parameters.
7235 @emph{Note:} there may be no value to @option{-finline-limit} that results
7236 in default behavior.
7238 @emph{Note:} pseudo instruction represents, in this particular context, an
7239 abstract measurement of function's size. In no way does it represent a count
7240 of assembly instructions and as such its exact meaning might change from one
7241 release to an another.
7243 @item -fno-keep-inline-dllexport
7244 @opindex -fno-keep-inline-dllexport
7245 This is a more fine-grained version of @option{-fkeep-inline-functions},
7246 which applies only to functions that are declared using the @code{dllexport}
7247 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7250 @item -fkeep-inline-functions
7251 @opindex fkeep-inline-functions
7252 In C, emit @code{static} functions that are declared @code{inline}
7253 into the object file, even if the function has been inlined into all
7254 of its callers. This switch does not affect functions using the
7255 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7256 inline functions into the object file.
7258 @item -fkeep-static-consts
7259 @opindex fkeep-static-consts
7260 Emit variables declared @code{static const} when optimization isn't turned
7261 on, even if the variables aren't referenced.
7263 GCC enables this option by default. If you want to force the compiler to
7264 check if a variable is referenced, regardless of whether or not
7265 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7267 @item -fmerge-constants
7268 @opindex fmerge-constants
7269 Attempt to merge identical constants (string constants and floating-point
7270 constants) across compilation units.
7272 This option is the default for optimized compilation if the assembler and
7273 linker support it. Use @option{-fno-merge-constants} to inhibit this
7276 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7278 @item -fmerge-all-constants
7279 @opindex fmerge-all-constants
7280 Attempt to merge identical constants and identical variables.
7282 This option implies @option{-fmerge-constants}. In addition to
7283 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7284 arrays or initialized constant variables with integral or floating-point
7285 types. Languages like C or C++ require each variable, including multiple
7286 instances of the same variable in recursive calls, to have distinct locations,
7287 so using this option results in non-conforming
7290 @item -fmodulo-sched
7291 @opindex fmodulo-sched
7292 Perform swing modulo scheduling immediately before the first scheduling
7293 pass. This pass looks at innermost loops and reorders their
7294 instructions by overlapping different iterations.
7296 @item -fmodulo-sched-allow-regmoves
7297 @opindex fmodulo-sched-allow-regmoves
7298 Perform more aggressive SMS-based modulo scheduling with register moves
7299 allowed. By setting this flag certain anti-dependences edges are
7300 deleted, which triggers the generation of reg-moves based on the
7301 life-range analysis. This option is effective only with
7302 @option{-fmodulo-sched} enabled.
7304 @item -fno-branch-count-reg
7305 @opindex fno-branch-count-reg
7306 Do not use ``decrement and branch'' instructions on a count register,
7307 but instead generate a sequence of instructions that decrement a
7308 register, compare it against zero, then branch based upon the result.
7309 This option is only meaningful on architectures that support such
7310 instructions, which include x86, PowerPC, IA-64 and S/390.
7312 The default is @option{-fbranch-count-reg}.
7314 @item -fno-function-cse
7315 @opindex fno-function-cse
7316 Do not put function addresses in registers; make each instruction that
7317 calls a constant function contain the function's address explicitly.
7319 This option results in less efficient code, but some strange hacks
7320 that alter the assembler output may be confused by the optimizations
7321 performed when this option is not used.
7323 The default is @option{-ffunction-cse}
7325 @item -fno-zero-initialized-in-bss
7326 @opindex fno-zero-initialized-in-bss
7327 If the target supports a BSS section, GCC by default puts variables that
7328 are initialized to zero into BSS@. This can save space in the resulting
7331 This option turns off this behavior because some programs explicitly
7332 rely on variables going to the data section---e.g., so that the
7333 resulting executable can find the beginning of that section and/or make
7334 assumptions based on that.
7336 The default is @option{-fzero-initialized-in-bss}.
7338 @item -fthread-jumps
7339 @opindex fthread-jumps
7340 Perform optimizations that check to see if a jump branches to a
7341 location where another comparison subsumed by the first is found. If
7342 so, the first branch is redirected to either the destination of the
7343 second branch or a point immediately following it, depending on whether
7344 the condition is known to be true or false.
7346 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7348 @item -fsplit-wide-types
7349 @opindex fsplit-wide-types
7350 When using a type that occupies multiple registers, such as @code{long
7351 long} on a 32-bit system, split the registers apart and allocate them
7352 independently. This normally generates better code for those types,
7353 but may make debugging more difficult.
7355 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7358 @item -fcse-follow-jumps
7359 @opindex fcse-follow-jumps
7360 In common subexpression elimination (CSE), scan through jump instructions
7361 when the target of the jump is not reached by any other path. For
7362 example, when CSE encounters an @code{if} statement with an
7363 @code{else} clause, CSE follows the jump when the condition
7366 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7368 @item -fcse-skip-blocks
7369 @opindex fcse-skip-blocks
7370 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7371 follow jumps that conditionally skip over blocks. When CSE
7372 encounters a simple @code{if} statement with no else clause,
7373 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7374 body of the @code{if}.
7376 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7378 @item -frerun-cse-after-loop
7379 @opindex frerun-cse-after-loop
7380 Re-run common subexpression elimination after loop optimizations are
7383 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7387 Perform a global common subexpression elimination pass.
7388 This pass also performs global constant and copy propagation.
7390 @emph{Note:} When compiling a program using computed gotos, a GCC
7391 extension, you may get better run-time performance if you disable
7392 the global common subexpression elimination pass by adding
7393 @option{-fno-gcse} to the command line.
7395 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7399 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7400 attempts to move loads that are only killed by stores into themselves. This
7401 allows a loop containing a load/store sequence to be changed to a load outside
7402 the loop, and a copy/store within the loop.
7404 Enabled by default when @option{-fgcse} is enabled.
7408 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7409 global common subexpression elimination. This pass attempts to move
7410 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7411 loops containing a load/store sequence can be changed to a load before
7412 the loop and a store after the loop.
7414 Not enabled at any optimization level.
7418 When @option{-fgcse-las} is enabled, the global common subexpression
7419 elimination pass eliminates redundant loads that come after stores to the
7420 same memory location (both partial and full redundancies).
7422 Not enabled at any optimization level.
7424 @item -fgcse-after-reload
7425 @opindex fgcse-after-reload
7426 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7427 pass is performed after reload. The purpose of this pass is to clean up
7430 @item -faggressive-loop-optimizations
7431 @opindex faggressive-loop-optimizations
7432 This option tells the loop optimizer to use language constraints to
7433 derive bounds for the number of iterations of a loop. This assumes that
7434 loop code does not invoke undefined behavior by for example causing signed
7435 integer overflows or out-of-bound array accesses. The bounds for the
7436 number of iterations of a loop are used to guide loop unrolling and peeling
7437 and loop exit test optimizations.
7438 This option is enabled by default.
7440 @item -funsafe-loop-optimizations
7441 @opindex funsafe-loop-optimizations
7442 This option tells the loop optimizer to assume that loop indices do not
7443 overflow, and that loops with nontrivial exit condition are not
7444 infinite. This enables a wider range of loop optimizations even if
7445 the loop optimizer itself cannot prove that these assumptions are valid.
7446 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7447 if it finds this kind of loop.
7449 @item -fcrossjumping
7450 @opindex fcrossjumping
7451 Perform cross-jumping transformation.
7452 This transformation unifies equivalent code and saves code size. The
7453 resulting code may or may not perform better than without cross-jumping.
7455 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7457 @item -fauto-inc-dec
7458 @opindex fauto-inc-dec
7459 Combine increments or decrements of addresses with memory accesses.
7460 This pass is always skipped on architectures that do not have
7461 instructions to support this. Enabled by default at @option{-O} and
7462 higher on architectures that support this.
7466 Perform dead code elimination (DCE) on RTL@.
7467 Enabled by default at @option{-O} and higher.
7471 Perform dead store elimination (DSE) on RTL@.
7472 Enabled by default at @option{-O} and higher.
7474 @item -fif-conversion
7475 @opindex fif-conversion
7476 Attempt to transform conditional jumps into branch-less equivalents. This
7477 includes use of conditional moves, min, max, set flags and abs instructions, and
7478 some tricks doable by standard arithmetics. The use of conditional execution
7479 on chips where it is available is controlled by @code{if-conversion2}.
7481 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7483 @item -fif-conversion2
7484 @opindex fif-conversion2
7485 Use conditional execution (where available) to transform conditional jumps into
7486 branch-less equivalents.
7488 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7490 @item -fdeclone-ctor-dtor
7491 @opindex fdeclone-ctor-dtor
7492 The C++ ABI requires multiple entry points for constructors and
7493 destructors: one for a base subobject, one for a complete object, and
7494 one for a virtual destructor that calls operator delete afterwards.
7495 For a hierarchy with virtual bases, the base and complete variants are
7496 clones, which means two copies of the function. With this option, the
7497 base and complete variants are changed to be thunks that call a common
7500 Enabled by @option{-Os}.
7502 @item -fdelete-null-pointer-checks
7503 @opindex fdelete-null-pointer-checks
7504 Assume that programs cannot safely dereference null pointers, and that
7505 no code or data element resides there. This enables simple constant
7506 folding optimizations at all optimization levels. In addition, other
7507 optimization passes in GCC use this flag to control global dataflow
7508 analyses that eliminate useless checks for null pointers; these assume
7509 that if a pointer is checked after it has already been dereferenced,
7512 Note however that in some environments this assumption is not true.
7513 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7514 for programs that depend on that behavior.
7516 Some targets, especially embedded ones, disable this option at all levels.
7517 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7518 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7519 are enabled independently at different optimization levels.
7521 @item -fdevirtualize
7522 @opindex fdevirtualize
7523 Attempt to convert calls to virtual functions to direct calls. This
7524 is done both within a procedure and interprocedurally as part of
7525 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
7526 propagation (@option{-fipa-cp}).
7527 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7529 @item -fdevirtualize-speculatively
7530 @opindex fdevirtualize-speculatively
7531 Attempt to convert calls to virtual functions to speculative direct calls.
7532 Based on the analysis of the type inheritance graph, determine for a given call
7533 the set of likely targets. If the set is small, preferably of size 1, change
7534 the call into an conditional deciding on direct and indirect call. The
7535 speculative calls enable more optimizations, such as inlining. When they seem
7536 useless after further optimization, they are converted back into original form.
7538 @item -fexpensive-optimizations
7539 @opindex fexpensive-optimizations
7540 Perform a number of minor optimizations that are relatively expensive.
7542 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7546 Attempt to remove redundant extension instructions. This is especially
7547 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7548 registers after writing to their lower 32-bit half.
7550 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7551 @option{-O3}, @option{-Os}.
7553 @item -flive-range-shrinkage
7554 @opindex flive-range-shrinkage
7555 Attempt to decrease register pressure through register live range
7556 shrinkage. This is helpful for fast processors with small or moderate
7559 @item -fira-algorithm=@var{algorithm}
7560 Use the specified coloring algorithm for the integrated register
7561 allocator. The @var{algorithm} argument can be @samp{priority}, which
7562 specifies Chow's priority coloring, or @samp{CB}, which specifies
7563 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7564 for all architectures, but for those targets that do support it, it is
7565 the default because it generates better code.
7567 @item -fira-region=@var{region}
7568 Use specified regions for the integrated register allocator. The
7569 @var{region} argument should be one of the following:
7574 Use all loops as register allocation regions.
7575 This can give the best results for machines with a small and/or
7576 irregular register set.
7579 Use all loops except for loops with small register pressure
7580 as the regions. This value usually gives
7581 the best results in most cases and for most architectures,
7582 and is enabled by default when compiling with optimization for speed
7583 (@option{-O}, @option{-O2}, @dots{}).
7586 Use all functions as a single region.
7587 This typically results in the smallest code size, and is enabled by default for
7588 @option{-Os} or @option{-O0}.
7592 @item -fira-hoist-pressure
7593 @opindex fira-hoist-pressure
7594 Use IRA to evaluate register pressure in the code hoisting pass for
7595 decisions to hoist expressions. This option usually results in smaller
7596 code, but it can slow the compiler down.
7598 This option is enabled at level @option{-Os} for all targets.
7600 @item -fira-loop-pressure
7601 @opindex fira-loop-pressure
7602 Use IRA to evaluate register pressure in loops for decisions to move
7603 loop invariants. This option usually results in generation
7604 of faster and smaller code on machines with large register files (>= 32
7605 registers), but it can slow the compiler down.
7607 This option is enabled at level @option{-O3} for some targets.
7609 @item -fno-ira-share-save-slots
7610 @opindex fno-ira-share-save-slots
7611 Disable sharing of stack slots used for saving call-used hard
7612 registers living through a call. Each hard register gets a
7613 separate stack slot, and as a result function stack frames are
7616 @item -fno-ira-share-spill-slots
7617 @opindex fno-ira-share-spill-slots
7618 Disable sharing of stack slots allocated for pseudo-registers. Each
7619 pseudo-register that does not get a hard register gets a separate
7620 stack slot, and as a result function stack frames are larger.
7622 @item -fira-verbose=@var{n}
7623 @opindex fira-verbose
7624 Control the verbosity of the dump file for the integrated register allocator.
7625 The default value is 5. If the value @var{n} is greater or equal to 10,
7626 the dump output is sent to stderr using the same format as @var{n} minus 10.
7628 @item -fdelayed-branch
7629 @opindex fdelayed-branch
7630 If supported for the target machine, attempt to reorder instructions
7631 to exploit instruction slots available after delayed branch
7634 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7636 @item -fschedule-insns
7637 @opindex fschedule-insns
7638 If supported for the target machine, attempt to reorder instructions to
7639 eliminate execution stalls due to required data being unavailable. This
7640 helps machines that have slow floating point or memory load instructions
7641 by allowing other instructions to be issued until the result of the load
7642 or floating-point instruction is required.
7644 Enabled at levels @option{-O2}, @option{-O3}.
7646 @item -fschedule-insns2
7647 @opindex fschedule-insns2
7648 Similar to @option{-fschedule-insns}, but requests an additional pass of
7649 instruction scheduling after register allocation has been done. This is
7650 especially useful on machines with a relatively small number of
7651 registers and where memory load instructions take more than one cycle.
7653 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7655 @item -fno-sched-interblock
7656 @opindex fno-sched-interblock
7657 Don't schedule instructions across basic blocks. This is normally
7658 enabled by default when scheduling before register allocation, i.e.@:
7659 with @option{-fschedule-insns} or at @option{-O2} or higher.
7661 @item -fno-sched-spec
7662 @opindex fno-sched-spec
7663 Don't allow speculative motion of non-load instructions. This is normally
7664 enabled by default when scheduling before register allocation, i.e.@:
7665 with @option{-fschedule-insns} or at @option{-O2} or higher.
7667 @item -fsched-pressure
7668 @opindex fsched-pressure
7669 Enable register pressure sensitive insn scheduling before register
7670 allocation. This only makes sense when scheduling before register
7671 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7672 @option{-O2} or higher. Usage of this option can improve the
7673 generated code and decrease its size by preventing register pressure
7674 increase above the number of available hard registers and subsequent
7675 spills in register allocation.
7677 @item -fsched-spec-load
7678 @opindex fsched-spec-load
7679 Allow speculative motion of some load instructions. This only makes
7680 sense when scheduling before register allocation, i.e.@: with
7681 @option{-fschedule-insns} or at @option{-O2} or higher.
7683 @item -fsched-spec-load-dangerous
7684 @opindex fsched-spec-load-dangerous
7685 Allow speculative motion of more load instructions. This only makes
7686 sense when scheduling before register allocation, i.e.@: with
7687 @option{-fschedule-insns} or at @option{-O2} or higher.
7689 @item -fsched-stalled-insns
7690 @itemx -fsched-stalled-insns=@var{n}
7691 @opindex fsched-stalled-insns
7692 Define how many insns (if any) can be moved prematurely from the queue
7693 of stalled insns into the ready list during the second scheduling pass.
7694 @option{-fno-sched-stalled-insns} means that no insns are moved
7695 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7696 on how many queued insns can be moved prematurely.
7697 @option{-fsched-stalled-insns} without a value is equivalent to
7698 @option{-fsched-stalled-insns=1}.
7700 @item -fsched-stalled-insns-dep
7701 @itemx -fsched-stalled-insns-dep=@var{n}
7702 @opindex fsched-stalled-insns-dep
7703 Define how many insn groups (cycles) are examined for a dependency
7704 on a stalled insn that is a candidate for premature removal from the queue
7705 of stalled insns. This has an effect only during the second scheduling pass,
7706 and only if @option{-fsched-stalled-insns} is used.
7707 @option{-fno-sched-stalled-insns-dep} is equivalent to
7708 @option{-fsched-stalled-insns-dep=0}.
7709 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7710 @option{-fsched-stalled-insns-dep=1}.
7712 @item -fsched2-use-superblocks
7713 @opindex fsched2-use-superblocks
7714 When scheduling after register allocation, use superblock scheduling.
7715 This allows motion across basic block boundaries,
7716 resulting in faster schedules. This option is experimental, as not all machine
7717 descriptions used by GCC model the CPU closely enough to avoid unreliable
7718 results from the algorithm.
7720 This only makes sense when scheduling after register allocation, i.e.@: with
7721 @option{-fschedule-insns2} or at @option{-O2} or higher.
7723 @item -fsched-group-heuristic
7724 @opindex fsched-group-heuristic
7725 Enable the group heuristic in the scheduler. This heuristic favors
7726 the instruction that belongs to a schedule group. This is enabled
7727 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7728 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7730 @item -fsched-critical-path-heuristic
7731 @opindex fsched-critical-path-heuristic
7732 Enable the critical-path heuristic in the scheduler. This heuristic favors
7733 instructions on the critical path. This is enabled by default when
7734 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7735 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7737 @item -fsched-spec-insn-heuristic
7738 @opindex fsched-spec-insn-heuristic
7739 Enable the speculative instruction heuristic in the scheduler. This
7740 heuristic favors speculative instructions with greater dependency weakness.
7741 This is enabled by default when scheduling is enabled, i.e.@:
7742 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7743 or at @option{-O2} or higher.
7745 @item -fsched-rank-heuristic
7746 @opindex fsched-rank-heuristic
7747 Enable the rank heuristic in the scheduler. This heuristic favors
7748 the instruction belonging to a basic block with greater size or frequency.
7749 This is enabled by default when scheduling is enabled, i.e.@:
7750 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7751 at @option{-O2} or higher.
7753 @item -fsched-last-insn-heuristic
7754 @opindex fsched-last-insn-heuristic
7755 Enable the last-instruction heuristic in the scheduler. This heuristic
7756 favors the instruction that is less dependent on the last instruction
7757 scheduled. This is enabled by default when scheduling is enabled,
7758 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7759 at @option{-O2} or higher.
7761 @item -fsched-dep-count-heuristic
7762 @opindex fsched-dep-count-heuristic
7763 Enable the dependent-count heuristic in the scheduler. This heuristic
7764 favors the instruction that has more instructions depending on it.
7765 This is enabled by default when scheduling is enabled, i.e.@:
7766 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7767 at @option{-O2} or higher.
7769 @item -freschedule-modulo-scheduled-loops
7770 @opindex freschedule-modulo-scheduled-loops
7771 Modulo scheduling is performed before traditional scheduling. If a loop
7772 is modulo scheduled, later scheduling passes may change its schedule.
7773 Use this option to control that behavior.
7775 @item -fselective-scheduling
7776 @opindex fselective-scheduling
7777 Schedule instructions using selective scheduling algorithm. Selective
7778 scheduling runs instead of the first scheduler pass.
7780 @item -fselective-scheduling2
7781 @opindex fselective-scheduling2
7782 Schedule instructions using selective scheduling algorithm. Selective
7783 scheduling runs instead of the second scheduler pass.
7785 @item -fsel-sched-pipelining
7786 @opindex fsel-sched-pipelining
7787 Enable software pipelining of innermost loops during selective scheduling.
7788 This option has no effect unless one of @option{-fselective-scheduling} or
7789 @option{-fselective-scheduling2} is turned on.
7791 @item -fsel-sched-pipelining-outer-loops
7792 @opindex fsel-sched-pipelining-outer-loops
7793 When pipelining loops during selective scheduling, also pipeline outer loops.
7794 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7796 @item -fsemantic-interposition
7797 @opindex fsemantic-interposition
7798 Some object formats, like ELF, allow interposing of symbols by dynamic linker.
7799 This means that for symbols exported from the DSO compiler can not perform
7800 inter-procedural propagation, inlining and other optimizations in anticipation
7801 that the function or variable in question may change. While this feature is
7802 useful, for example, to rewrite memory allocation functions by a debugging
7803 implementation, it is expensive in the terms of code quality.
7804 With @option{-fno-semantic-inteposition} compiler assumest that if interposition
7805 happens for functions the overwritting function will have
7806 precisely same semantics (and side effects). Similarly if interposition happens
7807 for variables, the constructor of the variable will be the same. The flag
7808 has no effect for functions explicitly declared inline, where
7809 interposition changing semantic is never allowed and for symbols explicitly
7813 @opindex fshrink-wrap
7814 Emit function prologues only before parts of the function that need it,
7815 rather than at the top of the function. This flag is enabled by default at
7816 @option{-O} and higher.
7818 @item -fcaller-saves
7819 @opindex fcaller-saves
7820 Enable allocation of values to registers that are clobbered by
7821 function calls, by emitting extra instructions to save and restore the
7822 registers around such calls. Such allocation is done only when it
7823 seems to result in better code.
7825 This option is always enabled by default on certain machines, usually
7826 those which have no call-preserved registers to use instead.
7828 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7830 @item -fcombine-stack-adjustments
7831 @opindex fcombine-stack-adjustments
7832 Tracks stack adjustments (pushes and pops) and stack memory references
7833 and then tries to find ways to combine them.
7835 Enabled by default at @option{-O1} and higher.
7837 @item -fuse-caller-save
7838 Use caller save registers for allocation if those registers are not used by
7839 any called function. In that case it is not necessary to save and restore
7840 them around calls. This is only possible if called functions are part of
7841 same compilation unit as current function and they are compiled before it.
7843 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7845 @item -fconserve-stack
7846 @opindex fconserve-stack
7847 Attempt to minimize stack usage. The compiler attempts to use less
7848 stack space, even if that makes the program slower. This option
7849 implies setting the @option{large-stack-frame} parameter to 100
7850 and the @option{large-stack-frame-growth} parameter to 400.
7852 @item -ftree-reassoc
7853 @opindex ftree-reassoc
7854 Perform reassociation on trees. This flag is enabled by default
7855 at @option{-O} and higher.
7859 Perform partial redundancy elimination (PRE) on trees. This flag is
7860 enabled by default at @option{-O2} and @option{-O3}.
7862 @item -ftree-partial-pre
7863 @opindex ftree-partial-pre
7864 Make partial redundancy elimination (PRE) more aggressive. This flag is
7865 enabled by default at @option{-O3}.
7867 @item -ftree-forwprop
7868 @opindex ftree-forwprop
7869 Perform forward propagation on trees. This flag is enabled by default
7870 at @option{-O} and higher.
7874 Perform full redundancy elimination (FRE) on trees. The difference
7875 between FRE and PRE is that FRE only considers expressions
7876 that are computed on all paths leading to the redundant computation.
7877 This analysis is faster than PRE, though it exposes fewer redundancies.
7878 This flag is enabled by default at @option{-O} and higher.
7880 @item -ftree-phiprop
7881 @opindex ftree-phiprop
7882 Perform hoisting of loads from conditional pointers on trees. This
7883 pass is enabled by default at @option{-O} and higher.
7885 @item -fhoist-adjacent-loads
7886 @opindex hoist-adjacent-loads
7887 Speculatively hoist loads from both branches of an if-then-else if the
7888 loads are from adjacent locations in the same structure and the target
7889 architecture has a conditional move instruction. This flag is enabled
7890 by default at @option{-O2} and higher.
7892 @item -ftree-copy-prop
7893 @opindex ftree-copy-prop
7894 Perform copy propagation on trees. This pass eliminates unnecessary
7895 copy operations. This flag is enabled by default at @option{-O} and
7898 @item -fipa-pure-const
7899 @opindex fipa-pure-const
7900 Discover which functions are pure or constant.
7901 Enabled by default at @option{-O} and higher.
7903 @item -fipa-reference
7904 @opindex fipa-reference
7905 Discover which static variables do not escape the
7907 Enabled by default at @option{-O} and higher.
7911 Perform interprocedural pointer analysis and interprocedural modification
7912 and reference analysis. This option can cause excessive memory and
7913 compile-time usage on large compilation units. It is not enabled by
7914 default at any optimization level.
7917 @opindex fipa-profile
7918 Perform interprocedural profile propagation. The functions called only from
7919 cold functions are marked as cold. Also functions executed once (such as
7920 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7921 functions and loop less parts of functions executed once are then optimized for
7923 Enabled by default at @option{-O} and higher.
7927 Perform interprocedural constant propagation.
7928 This optimization analyzes the program to determine when values passed
7929 to functions are constants and then optimizes accordingly.
7930 This optimization can substantially increase performance
7931 if the application has constants passed to functions.
7932 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7934 @item -fipa-cp-clone
7935 @opindex fipa-cp-clone
7936 Perform function cloning to make interprocedural constant propagation stronger.
7937 When enabled, interprocedural constant propagation performs function cloning
7938 when externally visible function can be called with constant arguments.
7939 Because this optimization can create multiple copies of functions,
7940 it may significantly increase code size
7941 (see @option{--param ipcp-unit-growth=@var{value}}).
7942 This flag is enabled by default at @option{-O3}.
7944 @item -fisolate-erroneous-paths-dereference
7945 Detect paths which trigger erroneous or undefined behaviour due to
7946 dereferencing a NULL pointer. Isolate those paths from the main control
7947 flow and turn the statement with erroneous or undefined behaviour into a trap.
7949 @item -fisolate-erroneous-paths-attribute
7950 Detect paths which trigger erroneous or undefined behaviour due a NULL value
7951 being used in a way which is forbidden by a @code{returns_nonnull} or @code{nonnull}
7952 attribute. Isolate those paths from the main control flow and turn the
7953 statement with erroneous or undefined behaviour into a trap. This is not
7954 currently enabled, but may be enabled by @code{-O2} in the future.
7958 Perform forward store motion on trees. This flag is
7959 enabled by default at @option{-O} and higher.
7961 @item -ftree-bit-ccp
7962 @opindex ftree-bit-ccp
7963 Perform sparse conditional bit constant propagation on trees and propagate
7964 pointer alignment information.
7965 This pass only operates on local scalar variables and is enabled by default
7966 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7970 Perform sparse conditional constant propagation (CCP) on trees. This
7971 pass only operates on local scalar variables and is enabled by default
7972 at @option{-O} and higher.
7975 @opindex fssa-phiopt
7976 Perform pattern matching on SSA PHI nodes to optimize conditional
7977 code. This pass is enabled by default at @option{-O} and higher.
7979 @item -ftree-switch-conversion
7980 Perform conversion of simple initializations in a switch to
7981 initializations from a scalar array. This flag is enabled by default
7982 at @option{-O2} and higher.
7984 @item -ftree-tail-merge
7985 Look for identical code sequences. When found, replace one with a jump to the
7986 other. This optimization is known as tail merging or cross jumping. This flag
7987 is enabled by default at @option{-O2} and higher. The compilation time
7989 be limited using @option{max-tail-merge-comparisons} parameter and
7990 @option{max-tail-merge-iterations} parameter.
7994 Perform dead code elimination (DCE) on trees. This flag is enabled by
7995 default at @option{-O} and higher.
7997 @item -ftree-builtin-call-dce
7998 @opindex ftree-builtin-call-dce
7999 Perform conditional dead code elimination (DCE) for calls to built-in functions
8000 that may set @code{errno} but are otherwise side-effect free. This flag is
8001 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8004 @item -ftree-dominator-opts
8005 @opindex ftree-dominator-opts
8006 Perform a variety of simple scalar cleanups (constant/copy
8007 propagation, redundancy elimination, range propagation and expression
8008 simplification) based on a dominator tree traversal. This also
8009 performs jump threading (to reduce jumps to jumps). This flag is
8010 enabled by default at @option{-O} and higher.
8014 Perform dead store elimination (DSE) on trees. A dead store is a store into
8015 a memory location that is later overwritten by another store without
8016 any intervening loads. In this case the earlier store can be deleted. This
8017 flag is enabled by default at @option{-O} and higher.
8021 Perform loop header copying on trees. This is beneficial since it increases
8022 effectiveness of code motion optimizations. It also saves one jump. This flag
8023 is enabled by default at @option{-O} and higher. It is not enabled
8024 for @option{-Os}, since it usually increases code size.
8026 @item -ftree-loop-optimize
8027 @opindex ftree-loop-optimize
8028 Perform loop optimizations on trees. This flag is enabled by default
8029 at @option{-O} and higher.
8031 @item -ftree-loop-linear
8032 @opindex ftree-loop-linear
8033 Perform loop interchange transformations on tree. Same as
8034 @option{-floop-interchange}. To use this code transformation, GCC has
8035 to be configured with @option{--with-ppl} and @option{--with-cloog} to
8036 enable the Graphite loop transformation infrastructure.
8038 @item -floop-interchange
8039 @opindex floop-interchange
8040 Perform loop interchange transformations on loops. Interchanging two
8041 nested loops switches the inner and outer loops. For example, given a
8046 A(J, I) = A(J, I) * C
8050 loop interchange transforms the loop as if it were written:
8054 A(J, I) = A(J, I) * C
8058 which can be beneficial when @code{N} is larger than the caches,
8059 because in Fortran, the elements of an array are stored in memory
8060 contiguously by column, and the original loop iterates over rows,
8061 potentially creating at each access a cache miss. This optimization
8062 applies to all the languages supported by GCC and is not limited to
8063 Fortran. To use this code transformation, GCC has to be configured
8064 with @option{--with-ppl} and @option{--with-cloog} to enable the
8065 Graphite loop transformation infrastructure.
8067 @item -floop-strip-mine
8068 @opindex floop-strip-mine
8069 Perform loop strip mining transformations on loops. Strip mining
8070 splits a loop into two nested loops. The outer loop has strides
8071 equal to the strip size and the inner loop has strides of the
8072 original loop within a strip. The strip length can be changed
8073 using the @option{loop-block-tile-size} parameter. For example,
8080 loop strip mining transforms the loop as if it were written:
8083 DO I = II, min (II + 50, N)
8088 This optimization applies to all the languages supported by GCC and is
8089 not limited to Fortran. To use this code transformation, GCC has to
8090 be configured with @option{--with-ppl} and @option{--with-cloog} to
8091 enable the Graphite loop transformation infrastructure.
8094 @opindex floop-block
8095 Perform loop blocking transformations on loops. Blocking strip mines
8096 each loop in the loop nest such that the memory accesses of the
8097 element loops fit inside caches. The strip length can be changed
8098 using the @option{loop-block-tile-size} parameter. For example, given
8103 A(J, I) = B(I) + C(J)
8107 loop blocking transforms the loop as if it were written:
8111 DO I = II, min (II + 50, N)
8112 DO J = JJ, min (JJ + 50, M)
8113 A(J, I) = B(I) + C(J)
8119 which can be beneficial when @code{M} is larger than the caches,
8120 because the innermost loop iterates over a smaller amount of data
8121 which can be kept in the caches. This optimization applies to all the
8122 languages supported by GCC and is not limited to Fortran. To use this
8123 code transformation, GCC has to be configured with @option{--with-ppl}
8124 and @option{--with-cloog} to enable the Graphite loop transformation
8127 @item -fgraphite-identity
8128 @opindex fgraphite-identity
8129 Enable the identity transformation for graphite. For every SCoP we generate
8130 the polyhedral representation and transform it back to gimple. Using
8131 @option{-fgraphite-identity} we can check the costs or benefits of the
8132 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8133 are also performed by the code generator CLooG, like index splitting and
8134 dead code elimination in loops.
8136 @item -floop-nest-optimize
8137 @opindex floop-nest-optimize
8138 Enable the ISL based loop nest optimizer. This is a generic loop nest
8139 optimizer based on the Pluto optimization algorithms. It calculates a loop
8140 structure optimized for data-locality and parallelism. This option
8143 @item -floop-parallelize-all
8144 @opindex floop-parallelize-all
8145 Use the Graphite data dependence analysis to identify loops that can
8146 be parallelized. Parallelize all the loops that can be analyzed to
8147 not contain loop carried dependences without checking that it is
8148 profitable to parallelize the loops.
8150 @item -fcheck-data-deps
8151 @opindex fcheck-data-deps
8152 Compare the results of several data dependence analyzers. This option
8153 is used for debugging the data dependence analyzers.
8155 @item -ftree-loop-if-convert
8156 Attempt to transform conditional jumps in the innermost loops to
8157 branch-less equivalents. The intent is to remove control-flow from
8158 the innermost loops in order to improve the ability of the
8159 vectorization pass to handle these loops. This is enabled by default
8160 if vectorization is enabled.
8162 @item -ftree-loop-if-convert-stores
8163 Attempt to also if-convert conditional jumps containing memory writes.
8164 This transformation can be unsafe for multi-threaded programs as it
8165 transforms conditional memory writes into unconditional memory writes.
8168 for (i = 0; i < N; i++)
8174 for (i = 0; i < N; i++)
8175 A[i] = cond ? expr : A[i];
8177 potentially producing data races.
8179 @item -ftree-loop-distribution
8180 Perform loop distribution. This flag can improve cache performance on
8181 big loop bodies and allow further loop optimizations, like
8182 parallelization or vectorization, to take place. For example, the loop
8199 @item -ftree-loop-distribute-patterns
8200 Perform loop distribution of patterns that can be code generated with
8201 calls to a library. This flag is enabled by default at @option{-O3}.
8203 This pass distributes the initialization loops and generates a call to
8204 memset zero. For example, the loop
8220 and the initialization loop is transformed into a call to memset zero.
8222 @item -ftree-loop-im
8223 @opindex ftree-loop-im
8224 Perform loop invariant motion on trees. This pass moves only invariants that
8225 are hard to handle at RTL level (function calls, operations that expand to
8226 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8227 operands of conditions that are invariant out of the loop, so that we can use
8228 just trivial invariantness analysis in loop unswitching. The pass also includes
8231 @item -ftree-loop-ivcanon
8232 @opindex ftree-loop-ivcanon
8233 Create a canonical counter for number of iterations in loops for which
8234 determining number of iterations requires complicated analysis. Later
8235 optimizations then may determine the number easily. Useful especially
8236 in connection with unrolling.
8240 Perform induction variable optimizations (strength reduction, induction
8241 variable merging and induction variable elimination) on trees.
8243 @item -ftree-parallelize-loops=n
8244 @opindex ftree-parallelize-loops
8245 Parallelize loops, i.e., split their iteration space to run in n threads.
8246 This is only possible for loops whose iterations are independent
8247 and can be arbitrarily reordered. The optimization is only
8248 profitable on multiprocessor machines, for loops that are CPU-intensive,
8249 rather than constrained e.g.@: by memory bandwidth. This option
8250 implies @option{-pthread}, and thus is only supported on targets
8251 that have support for @option{-pthread}.
8255 Perform function-local points-to analysis on trees. This flag is
8256 enabled by default at @option{-O} and higher.
8260 Perform scalar replacement of aggregates. This pass replaces structure
8261 references with scalars to prevent committing structures to memory too
8262 early. This flag is enabled by default at @option{-O} and higher.
8264 @item -ftree-copyrename
8265 @opindex ftree-copyrename
8266 Perform copy renaming on trees. This pass attempts to rename compiler
8267 temporaries to other variables at copy locations, usually resulting in
8268 variable names which more closely resemble the original variables. This flag
8269 is enabled by default at @option{-O} and higher.
8271 @item -ftree-coalesce-inlined-vars
8272 @opindex ftree-coalesce-inlined-vars
8273 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8274 combine small user-defined variables too, but only if they were inlined
8275 from other functions. It is a more limited form of
8276 @option{-ftree-coalesce-vars}. This may harm debug information of such
8277 inlined variables, but it will keep variables of the inlined-into
8278 function apart from each other, such that they are more likely to
8279 contain the expected values in a debugging session. This was the
8280 default in GCC versions older than 4.7.
8282 @item -ftree-coalesce-vars
8283 @opindex ftree-coalesce-vars
8284 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8285 combine small user-defined variables too, instead of just compiler
8286 temporaries. This may severely limit the ability to debug an optimized
8287 program compiled with @option{-fno-var-tracking-assignments}. In the
8288 negated form, this flag prevents SSA coalescing of user variables,
8289 including inlined ones. This option is enabled by default.
8293 Perform temporary expression replacement during the SSA->normal phase. Single
8294 use/single def temporaries are replaced at their use location with their
8295 defining expression. This results in non-GIMPLE code, but gives the expanders
8296 much more complex trees to work on resulting in better RTL generation. This is
8297 enabled by default at @option{-O} and higher.
8301 Perform straight-line strength reduction on trees. This recognizes related
8302 expressions involving multiplications and replaces them by less expensive
8303 calculations when possible. This is enabled by default at @option{-O} and
8306 @item -ftree-vectorize
8307 @opindex ftree-vectorize
8308 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8309 and @option{-ftree-slp-vectorize} if not explicitly specified.
8311 @item -ftree-loop-vectorize
8312 @opindex ftree-loop-vectorize
8313 Perform loop vectorization on trees. This flag is enabled by default at
8314 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8316 @item -ftree-slp-vectorize
8317 @opindex ftree-slp-vectorize
8318 Perform basic block vectorization on trees. This flag is enabled by default at
8319 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8321 @item -fvect-cost-model=@var{model}
8322 @opindex fvect-cost-model
8323 Alter the cost model used for vectorization. The @var{model} argument
8324 should be one of @code{unlimited}, @code{dynamic} or @code{cheap}.
8325 With the @code{unlimited} model the vectorized code-path is assumed
8326 to be profitable while with the @code{dynamic} model a runtime check
8327 will guard the vectorized code-path to enable it only for iteration
8328 counts that will likely execute faster than when executing the original
8329 scalar loop. The @code{cheap} model will disable vectorization of
8330 loops where doing so would be cost prohibitive for example due to
8331 required runtime checks for data dependence or alignment but otherwise
8332 is equal to the @code{dynamic} model.
8333 The default cost model depends on other optimization flags and is
8334 either @code{dynamic} or @code{cheap}.
8336 @item -fsimd-cost-model=@var{model}
8337 @opindex fsimd-cost-model
8338 Alter the cost model used for vectorization of loops marked with the OpenMP
8339 or Cilk Plus simd directive. The @var{model} argument should be one of
8340 @code{unlimited}, @code{dynamic}, @code{cheap}. All values of @var{model}
8341 have the same meaning as described in @option{-fvect-cost-model} and by
8342 default a cost model defined with @option{-fvect-cost-model} is used.
8346 Perform Value Range Propagation on trees. This is similar to the
8347 constant propagation pass, but instead of values, ranges of values are
8348 propagated. This allows the optimizers to remove unnecessary range
8349 checks like array bound checks and null pointer checks. This is
8350 enabled by default at @option{-O2} and higher. Null pointer check
8351 elimination is only done if @option{-fdelete-null-pointer-checks} is
8356 Perform tail duplication to enlarge superblock size. This transformation
8357 simplifies the control flow of the function allowing other optimizations to do
8360 @item -funroll-loops
8361 @opindex funroll-loops
8362 Unroll loops whose number of iterations can be determined at compile
8363 time or upon entry to the loop. @option{-funroll-loops} implies
8364 @option{-frerun-cse-after-loop}. This option makes code larger,
8365 and may or may not make it run faster.
8367 @item -funroll-all-loops
8368 @opindex funroll-all-loops
8369 Unroll all loops, even if their number of iterations is uncertain when
8370 the loop is entered. This usually makes programs run more slowly.
8371 @option{-funroll-all-loops} implies the same options as
8372 @option{-funroll-loops},
8374 @item -fsplit-ivs-in-unroller
8375 @opindex fsplit-ivs-in-unroller
8376 Enables expression of values of induction variables in later iterations
8377 of the unrolled loop using the value in the first iteration. This breaks
8378 long dependency chains, thus improving efficiency of the scheduling passes.
8380 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8381 same effect. However, that is not reliable in cases where the loop body
8382 is more complicated than a single basic block. It also does not work at all
8383 on some architectures due to restrictions in the CSE pass.
8385 This optimization is enabled by default.
8387 @item -fvariable-expansion-in-unroller
8388 @opindex fvariable-expansion-in-unroller
8389 With this option, the compiler creates multiple copies of some
8390 local variables when unrolling a loop, which can result in superior code.
8392 @item -fpartial-inlining
8393 @opindex fpartial-inlining
8394 Inline parts of functions. This option has any effect only
8395 when inlining itself is turned on by the @option{-finline-functions}
8396 or @option{-finline-small-functions} options.
8398 Enabled at level @option{-O2}.
8400 @item -fpredictive-commoning
8401 @opindex fpredictive-commoning
8402 Perform predictive commoning optimization, i.e., reusing computations
8403 (especially memory loads and stores) performed in previous
8404 iterations of loops.
8406 This option is enabled at level @option{-O3}.
8408 @item -fprefetch-loop-arrays
8409 @opindex fprefetch-loop-arrays
8410 If supported by the target machine, generate instructions to prefetch
8411 memory to improve the performance of loops that access large arrays.
8413 This option may generate better or worse code; results are highly
8414 dependent on the structure of loops within the source code.
8416 Disabled at level @option{-Os}.
8419 @itemx -fno-peephole2
8420 @opindex fno-peephole
8421 @opindex fno-peephole2
8422 Disable any machine-specific peephole optimizations. The difference
8423 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8424 are implemented in the compiler; some targets use one, some use the
8425 other, a few use both.
8427 @option{-fpeephole} is enabled by default.
8428 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8430 @item -fno-guess-branch-probability
8431 @opindex fno-guess-branch-probability
8432 Do not guess branch probabilities using heuristics.
8434 GCC uses heuristics to guess branch probabilities if they are
8435 not provided by profiling feedback (@option{-fprofile-arcs}). These
8436 heuristics are based on the control flow graph. If some branch probabilities
8437 are specified by @samp{__builtin_expect}, then the heuristics are
8438 used to guess branch probabilities for the rest of the control flow graph,
8439 taking the @samp{__builtin_expect} info into account. The interactions
8440 between the heuristics and @samp{__builtin_expect} can be complex, and in
8441 some cases, it may be useful to disable the heuristics so that the effects
8442 of @samp{__builtin_expect} are easier to understand.
8444 The default is @option{-fguess-branch-probability} at levels
8445 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8447 @item -freorder-blocks
8448 @opindex freorder-blocks
8449 Reorder basic blocks in the compiled function in order to reduce number of
8450 taken branches and improve code locality.
8452 Enabled at levels @option{-O2}, @option{-O3}.
8454 @item -freorder-blocks-and-partition
8455 @opindex freorder-blocks-and-partition
8456 In addition to reordering basic blocks in the compiled function, in order
8457 to reduce number of taken branches, partitions hot and cold basic blocks
8458 into separate sections of the assembly and .o files, to improve
8459 paging and cache locality performance.
8461 This optimization is automatically turned off in the presence of
8462 exception handling, for linkonce sections, for functions with a user-defined
8463 section attribute and on any architecture that does not support named
8466 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8468 @item -freorder-functions
8469 @opindex freorder-functions
8470 Reorder functions in the object file in order to
8471 improve code locality. This is implemented by using special
8472 subsections @code{.text.hot} for most frequently executed functions and
8473 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8474 the linker so object file format must support named sections and linker must
8475 place them in a reasonable way.
8477 Also profile feedback must be available to make this option effective. See
8478 @option{-fprofile-arcs} for details.
8480 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8482 @item -fstrict-aliasing
8483 @opindex fstrict-aliasing
8484 Allow the compiler to assume the strictest aliasing rules applicable to
8485 the language being compiled. For C (and C++), this activates
8486 optimizations based on the type of expressions. In particular, an
8487 object of one type is assumed never to reside at the same address as an
8488 object of a different type, unless the types are almost the same. For
8489 example, an @code{unsigned int} can alias an @code{int}, but not a
8490 @code{void*} or a @code{double}. A character type may alias any other
8493 @anchor{Type-punning}Pay special attention to code like this:
8506 The practice of reading from a different union member than the one most
8507 recently written to (called ``type-punning'') is common. Even with
8508 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8509 is accessed through the union type. So, the code above works as
8510 expected. @xref{Structures unions enumerations and bit-fields
8511 implementation}. However, this code might not:
8522 Similarly, access by taking the address, casting the resulting pointer
8523 and dereferencing the result has undefined behavior, even if the cast
8524 uses a union type, e.g.:
8528 return ((union a_union *) &d)->i;
8532 The @option{-fstrict-aliasing} option is enabled at levels
8533 @option{-O2}, @option{-O3}, @option{-Os}.
8535 @item -fstrict-overflow
8536 @opindex fstrict-overflow
8537 Allow the compiler to assume strict signed overflow rules, depending
8538 on the language being compiled. For C (and C++) this means that
8539 overflow when doing arithmetic with signed numbers is undefined, which
8540 means that the compiler may assume that it does not happen. This
8541 permits various optimizations. For example, the compiler assumes
8542 that an expression like @code{i + 10 > i} is always true for
8543 signed @code{i}. This assumption is only valid if signed overflow is
8544 undefined, as the expression is false if @code{i + 10} overflows when
8545 using twos complement arithmetic. When this option is in effect any
8546 attempt to determine whether an operation on signed numbers
8547 overflows must be written carefully to not actually involve overflow.
8549 This option also allows the compiler to assume strict pointer
8550 semantics: given a pointer to an object, if adding an offset to that
8551 pointer does not produce a pointer to the same object, the addition is
8552 undefined. This permits the compiler to conclude that @code{p + u >
8553 p} is always true for a pointer @code{p} and unsigned integer
8554 @code{u}. This assumption is only valid because pointer wraparound is
8555 undefined, as the expression is false if @code{p + u} overflows using
8556 twos complement arithmetic.
8558 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8559 that integer signed overflow is fully defined: it wraps. When
8560 @option{-fwrapv} is used, there is no difference between
8561 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8562 integers. With @option{-fwrapv} certain types of overflow are
8563 permitted. For example, if the compiler gets an overflow when doing
8564 arithmetic on constants, the overflowed value can still be used with
8565 @option{-fwrapv}, but not otherwise.
8567 The @option{-fstrict-overflow} option is enabled at levels
8568 @option{-O2}, @option{-O3}, @option{-Os}.
8570 @item -falign-functions
8571 @itemx -falign-functions=@var{n}
8572 @opindex falign-functions
8573 Align the start of functions to the next power-of-two greater than
8574 @var{n}, skipping up to @var{n} bytes. For instance,
8575 @option{-falign-functions=32} aligns functions to the next 32-byte
8576 boundary, but @option{-falign-functions=24} aligns to the next
8577 32-byte boundary only if this can be done by skipping 23 bytes or less.
8579 @option{-fno-align-functions} and @option{-falign-functions=1} are
8580 equivalent and mean that functions are not aligned.
8582 Some assemblers only support this flag when @var{n} is a power of two;
8583 in that case, it is rounded up.
8585 If @var{n} is not specified or is zero, use a machine-dependent default.
8587 Enabled at levels @option{-O2}, @option{-O3}.
8589 @item -falign-labels
8590 @itemx -falign-labels=@var{n}
8591 @opindex falign-labels
8592 Align all branch targets to a power-of-two boundary, skipping up to
8593 @var{n} bytes like @option{-falign-functions}. This option can easily
8594 make code slower, because it must insert dummy operations for when the
8595 branch target is reached in the usual flow of the code.
8597 @option{-fno-align-labels} and @option{-falign-labels=1} are
8598 equivalent and mean that labels are not aligned.
8600 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8601 are greater than this value, then their values are used instead.
8603 If @var{n} is not specified or is zero, use a machine-dependent default
8604 which is very likely to be @samp{1}, meaning no alignment.
8606 Enabled at levels @option{-O2}, @option{-O3}.
8609 @itemx -falign-loops=@var{n}
8610 @opindex falign-loops
8611 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8612 like @option{-falign-functions}. If the loops are
8613 executed many times, this makes up for any execution of the dummy
8616 @option{-fno-align-loops} and @option{-falign-loops=1} are
8617 equivalent and mean that loops are not aligned.
8619 If @var{n} is not specified or is zero, use a machine-dependent default.
8621 Enabled at levels @option{-O2}, @option{-O3}.
8624 @itemx -falign-jumps=@var{n}
8625 @opindex falign-jumps
8626 Align branch targets to a power-of-two boundary, for branch targets
8627 where the targets can only be reached by jumping, skipping up to @var{n}
8628 bytes like @option{-falign-functions}. In this case, no dummy operations
8631 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8632 equivalent and mean that loops are not aligned.
8634 If @var{n} is not specified or is zero, use a machine-dependent default.
8636 Enabled at levels @option{-O2}, @option{-O3}.
8638 @item -funit-at-a-time
8639 @opindex funit-at-a-time
8640 This option is left for compatibility reasons. @option{-funit-at-a-time}
8641 has no effect, while @option{-fno-unit-at-a-time} implies
8642 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
8646 @item -fno-toplevel-reorder
8647 @opindex fno-toplevel-reorder
8648 Do not reorder top-level functions, variables, and @code{asm}
8649 statements. Output them in the same order that they appear in the
8650 input file. When this option is used, unreferenced static variables
8651 are not removed. This option is intended to support existing code
8652 that relies on a particular ordering. For new code, it is better to
8653 use attributes when possible.
8655 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8656 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8661 Constructs webs as commonly used for register allocation purposes and assign
8662 each web individual pseudo register. This allows the register allocation pass
8663 to operate on pseudos directly, but also strengthens several other optimization
8664 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8665 however, make debugging impossible, since variables no longer stay in a
8668 Enabled by default with @option{-funroll-loops}.
8670 @item -fwhole-program
8671 @opindex fwhole-program
8672 Assume that the current compilation unit represents the whole program being
8673 compiled. All public functions and variables with the exception of @code{main}
8674 and those merged by attribute @code{externally_visible} become static functions
8675 and in effect are optimized more aggressively by interprocedural optimizers.
8677 This option should not be used in combination with @code{-flto}.
8678 Instead relying on a linker plugin should provide safer and more precise
8681 @item -flto[=@var{n}]
8683 This option runs the standard link-time optimizer. When invoked
8684 with source code, it generates GIMPLE (one of GCC's internal
8685 representations) and writes it to special ELF sections in the object
8686 file. When the object files are linked together, all the function
8687 bodies are read from these ELF sections and instantiated as if they
8688 had been part of the same translation unit.
8690 To use the link-time optimizer, @option{-flto} and optimization
8691 options should be specified at compile time and during the final link.
8695 gcc -c -O2 -flto foo.c
8696 gcc -c -O2 -flto bar.c
8697 gcc -o myprog -flto -O2 foo.o bar.o
8700 The first two invocations to GCC save a bytecode representation
8701 of GIMPLE into special ELF sections inside @file{foo.o} and
8702 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8703 @file{foo.o} and @file{bar.o}, merges the two files into a single
8704 internal image, and compiles the result as usual. Since both
8705 @file{foo.o} and @file{bar.o} are merged into a single image, this
8706 causes all the interprocedural analyses and optimizations in GCC to
8707 work across the two files as if they were a single one. This means,
8708 for example, that the inliner is able to inline functions in
8709 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8711 Another (simpler) way to enable link-time optimization is:
8714 gcc -o myprog -flto -O2 foo.c bar.c
8717 The above generates bytecode for @file{foo.c} and @file{bar.c},
8718 merges them together into a single GIMPLE representation and optimizes
8719 them as usual to produce @file{myprog}.
8721 The only important thing to keep in mind is that to enable link-time
8722 optimizations you need to use the GCC driver to perform the link-step.
8723 GCC then automatically performs link-time optimization if any of the
8724 objects involved were compiled with the @option{-flto}. You generally
8725 should specify the optimization options to be used for link-time
8726 optimization though GCC will try to be clever at guessing an
8727 optimization level to use from the options used at compile-time
8728 if you fail to specify one at link-time. You can always override
8729 the automatic decision to do link-time optimization at link-time
8730 by passing @option{-fno-lto} to the link command.
8732 To make whole program optimization effective, it is necessary to make
8733 certain whole program assumptions. The compiler needs to know
8734 what functions and variables can be accessed by libraries and runtime
8735 outside of the link-time optimized unit. When supported by the linker,
8736 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8737 to the compiler about used and externally visible symbols. When
8738 the linker plugin is not available, @option{-fwhole-program} should be
8739 used to allow the compiler to make these assumptions, which leads
8740 to more aggressive optimization decisions.
8742 When @option{-fuse-linker-plugin} is not enabled then, when a file is
8743 compiled with @option{-flto}, the generated object file is larger than
8744 a regular object file because it contains GIMPLE bytecodes and the usual
8745 final code (see @option{-ffat-lto-objects}. This means that
8746 object files with LTO information can be linked as normal object
8747 files; if @option{-fno-lto} is passed to the linker, no
8748 interprocedural optimizations are applied. Note that when
8749 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
8750 but you cannot perform a regular, non-LTO link on them.
8752 Additionally, the optimization flags used to compile individual files
8753 are not necessarily related to those used at link time. For instance,
8756 gcc -c -O0 -ffat-lto-objects -flto foo.c
8757 gcc -c -O0 -ffat-lto-objects -flto bar.c
8758 gcc -o myprog -O3 foo.o bar.o
8761 This produces individual object files with unoptimized assembler
8762 code, but the resulting binary @file{myprog} is optimized at
8763 @option{-O3}. If, instead, the final binary is generated with
8764 @option{-fno-lto}, then @file{myprog} is not optimized.
8766 When producing the final binary, GCC only
8767 applies link-time optimizations to those files that contain bytecode.
8768 Therefore, you can mix and match object files and libraries with
8769 GIMPLE bytecodes and final object code. GCC automatically selects
8770 which files to optimize in LTO mode and which files to link without
8773 There are some code generation flags preserved by GCC when
8774 generating bytecodes, as they need to be used during the final link
8775 stage. Generally options specified at link-time override those
8776 specified at compile-time.
8778 If you do not specify an optimization level option @option{-O} at
8779 link-time then GCC will compute one based on the optimization levels
8780 used when compiling the object files. The highest optimization
8781 level will win here.
8783 Currently, the following options and their setting are take from
8784 the first object file that explicitely specified it:
8785 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8786 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8787 and all the @option{-m} target flags.
8789 Certain ABI changing flags are required to match in all compilation-units
8790 and trying to override this at link-time with a conflicting value
8791 is ignored. This includes options such as @option{-freg-struct-return}
8792 and @option{-fpcc-struct-return}.
8794 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8795 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8796 are passed through to the link stage and merged conservatively for
8797 conflicting translation units. Specifically
8798 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8799 precedence and for example @option{-ffp-contract=off} takes precedence
8800 over @option{-ffp-contract=fast}. You can override them at linke-time.
8802 It is recommended that you compile all the files participating in the
8803 same link with the same options and also specify those options at
8806 If LTO encounters objects with C linkage declared with incompatible
8807 types in separate translation units to be linked together (undefined
8808 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8809 issued. The behavior is still undefined at run time. Similar
8810 diagnostics may be raised for other languages.
8812 Another feature of LTO is that it is possible to apply interprocedural
8813 optimizations on files written in different languages:
8818 gfortran -c -flto baz.f90
8819 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8822 Notice that the final link is done with @command{g++} to get the C++
8823 runtime libraries and @option{-lgfortran} is added to get the Fortran
8824 runtime libraries. In general, when mixing languages in LTO mode, you
8825 should use the same link command options as when mixing languages in a
8826 regular (non-LTO) compilation.
8828 If object files containing GIMPLE bytecode are stored in a library archive, say
8829 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8830 are using a linker with plugin support. To create static libraries suitable
8831 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8832 and @code{ranlib}; to show the symbols of object files with GIMPLE bytecode, use
8833 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8834 and @command{nm} have been compiled with plugin support. At link time, use the the
8835 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8836 the LTO optimization process:
8839 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8842 With the linker plugin enabled, the linker extracts the needed
8843 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8844 to make them part of the aggregated GIMPLE image to be optimized.
8846 If you are not using a linker with plugin support and/or do not
8847 enable the linker plugin, then the objects inside @file{libfoo.a}
8848 are extracted and linked as usual, but they do not participate
8849 in the LTO optimization process. In order to make a static library suitable
8850 for both LTO optimization and usual linkage, compile its object files with
8851 @option{-flto} @code{-ffat-lto-objects}.
8853 Link-time optimizations do not require the presence of the whole program to
8854 operate. If the program does not require any symbols to be exported, it is
8855 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8856 the interprocedural optimizers to use more aggressive assumptions which may
8857 lead to improved optimization opportunities.
8858 Use of @option{-fwhole-program} is not needed when linker plugin is
8859 active (see @option{-fuse-linker-plugin}).
8861 The current implementation of LTO makes no
8862 attempt to generate bytecode that is portable between different
8863 types of hosts. The bytecode files are versioned and there is a
8864 strict version check, so bytecode files generated in one version of
8865 GCC will not work with an older or newer version of GCC.
8867 Link-time optimization does not work well with generation of debugging
8868 information. Combining @option{-flto} with
8869 @option{-g} is currently experimental and expected to produce unexpected
8872 If you specify the optional @var{n}, the optimization and code
8873 generation done at link time is executed in parallel using @var{n}
8874 parallel jobs by utilizing an installed @command{make} program. The
8875 environment variable @env{MAKE} may be used to override the program
8876 used. The default value for @var{n} is 1.
8878 You can also specify @option{-flto=jobserver} to use GNU make's
8879 job server mode to determine the number of parallel jobs. This
8880 is useful when the Makefile calling GCC is already executing in parallel.
8881 You must prepend a @samp{+} to the command recipe in the parent Makefile
8882 for this to work. This option likely only works if @env{MAKE} is
8885 @item -flto-partition=@var{alg}
8886 @opindex flto-partition
8887 Specify the partitioning algorithm used by the link-time optimizer.
8888 The value is either @code{1to1} to specify a partitioning mirroring
8889 the original source files or @code{balanced} to specify partitioning
8890 into equally sized chunks (whenever possible) or @code{max} to create
8891 new partition for every symbol where possible. Specifying @code{none}
8892 as an algorithm disables partitioning and streaming completely.
8893 The default value is @code{balanced}. While @code{1to1} can be used
8894 as an workaround for various code ordering issues, the @code{max}
8895 partitioning is intended for internal testing only.
8896 The value @code{one} specifies that exactly one partition should be
8897 used while the value @code{none} bypasses partitioning and executes
8898 the link-time optimization step directly from the WPA phase.
8900 @item -flto-compression-level=@var{n}
8901 This option specifies the level of compression used for intermediate
8902 language written to LTO object files, and is only meaningful in
8903 conjunction with LTO mode (@option{-flto}). Valid
8904 values are 0 (no compression) to 9 (maximum compression). Values
8905 outside this range are clamped to either 0 or 9. If the option is not
8906 given, a default balanced compression setting is used.
8909 Prints a report with internal details on the workings of the link-time
8910 optimizer. The contents of this report vary from version to version.
8911 It is meant to be useful to GCC developers when processing object
8912 files in LTO mode (via @option{-flto}).
8914 Disabled by default.
8916 @item -flto-report-wpa
8917 Like @option{-flto-report}, but only print for the WPA phase of Link
8920 @item -fuse-linker-plugin
8921 Enables the use of a linker plugin during link-time optimization. This
8922 option relies on plugin support in the linker, which is available in gold
8923 or in GNU ld 2.21 or newer.
8925 This option enables the extraction of object files with GIMPLE bytecode out
8926 of library archives. This improves the quality of optimization by exposing
8927 more code to the link-time optimizer. This information specifies what
8928 symbols can be accessed externally (by non-LTO object or during dynamic
8929 linking). Resulting code quality improvements on binaries (and shared
8930 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
8931 See @option{-flto} for a description of the effect of this flag and how to
8934 This option is enabled by default when LTO support in GCC is enabled
8935 and GCC was configured for use with
8936 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8938 @item -ffat-lto-objects
8939 @opindex ffat-lto-objects
8940 Fat LTO objects are object files that contain both the intermediate language
8941 and the object code. This makes them usable for both LTO linking and normal
8942 linking. This option is effective only when compiling with @option{-flto}
8943 and is ignored at link time.
8945 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8946 requires the complete toolchain to be aware of LTO. It requires a linker with
8947 linker plugin support for basic functionality. Additionally,
8948 @command{nm}, @command{ar} and @command{ranlib}
8949 need to support linker plugins to allow a full-featured build environment
8950 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8951 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8952 to these tools. With non fat LTO makefiles need to be modified to use them.
8954 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8957 @item -fcompare-elim
8958 @opindex fcompare-elim
8959 After register allocation and post-register allocation instruction splitting,
8960 identify arithmetic instructions that compute processor flags similar to a
8961 comparison operation based on that arithmetic. If possible, eliminate the
8962 explicit comparison operation.
8964 This pass only applies to certain targets that cannot explicitly represent
8965 the comparison operation before register allocation is complete.
8967 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8970 @opindex fuse-ld=bfd
8971 Use the @command{bfd} linker instead of the default linker.
8974 @opindex fuse-ld=gold
8975 Use the @command{gold} linker instead of the default linker.
8977 @item -fcprop-registers
8978 @opindex fcprop-registers
8979 After register allocation and post-register allocation instruction splitting,
8980 perform a copy-propagation pass to try to reduce scheduling dependencies
8981 and occasionally eliminate the copy.
8983 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8985 @item -fprofile-correction
8986 @opindex fprofile-correction
8987 Profiles collected using an instrumented binary for multi-threaded programs may
8988 be inconsistent due to missed counter updates. When this option is specified,
8989 GCC uses heuristics to correct or smooth out such inconsistencies. By
8990 default, GCC emits an error message when an inconsistent profile is detected.
8992 @item -fprofile-dir=@var{path}
8993 @opindex fprofile-dir
8995 Set the directory to search for the profile data files in to @var{path}.
8996 This option affects only the profile data generated by
8997 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8998 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8999 and its related options. Both absolute and relative paths can be used.
9000 By default, GCC uses the current directory as @var{path}, thus the
9001 profile data file appears in the same directory as the object file.
9003 @item -fprofile-generate
9004 @itemx -fprofile-generate=@var{path}
9005 @opindex fprofile-generate
9007 Enable options usually used for instrumenting application to produce
9008 profile useful for later recompilation with profile feedback based
9009 optimization. You must use @option{-fprofile-generate} both when
9010 compiling and when linking your program.
9012 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
9014 If @var{path} is specified, GCC looks at the @var{path} to find
9015 the profile feedback data files. See @option{-fprofile-dir}.
9018 @itemx -fprofile-use=@var{path}
9019 @opindex fprofile-use
9020 Enable profile feedback directed optimizations, and optimizations
9021 generally profitable only with profile feedback available.
9023 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
9024 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}, @code{-ftree-vectorize},
9025 @code{ftree-loop-distribute-patterns}
9027 By default, GCC emits an error message if the feedback profiles do not
9028 match the source code. This error can be turned into a warning by using
9029 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9032 If @var{path} is specified, GCC looks at the @var{path} to find
9033 the profile feedback data files. See @option{-fprofile-dir}.
9036 The following options control compiler behavior regarding floating-point
9037 arithmetic. These options trade off between speed and
9038 correctness. All must be specifically enabled.
9042 @opindex ffloat-store
9043 Do not store floating-point variables in registers, and inhibit other
9044 options that might change whether a floating-point value is taken from a
9047 @cindex floating-point precision
9048 This option prevents undesirable excess precision on machines such as
9049 the 68000 where the floating registers (of the 68881) keep more
9050 precision than a @code{double} is supposed to have. Similarly for the
9051 x86 architecture. For most programs, the excess precision does only
9052 good, but a few programs rely on the precise definition of IEEE floating
9053 point. Use @option{-ffloat-store} for such programs, after modifying
9054 them to store all pertinent intermediate computations into variables.
9056 @item -fexcess-precision=@var{style}
9057 @opindex fexcess-precision
9058 This option allows further control over excess precision on machines
9059 where floating-point registers have more precision than the IEEE
9060 @code{float} and @code{double} types and the processor does not
9061 support operations rounding to those types. By default,
9062 @option{-fexcess-precision=fast} is in effect; this means that
9063 operations are carried out in the precision of the registers and that
9064 it is unpredictable when rounding to the types specified in the source
9065 code takes place. When compiling C, if
9066 @option{-fexcess-precision=standard} is specified then excess
9067 precision follows the rules specified in ISO C99; in particular,
9068 both casts and assignments cause values to be rounded to their
9069 semantic types (whereas @option{-ffloat-store} only affects
9070 assignments). This option is enabled by default for C if a strict
9071 conformance option such as @option{-std=c99} is used.
9074 @option{-fexcess-precision=standard} is not implemented for languages
9075 other than C, and has no effect if
9076 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9077 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9078 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9079 semantics apply without excess precision, and in the latter, rounding
9084 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9085 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9086 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9088 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9090 This option is not turned on by any @option{-O} option besides
9091 @option{-Ofast} since it can result in incorrect output for programs
9092 that depend on an exact implementation of IEEE or ISO rules/specifications
9093 for math functions. It may, however, yield faster code for programs
9094 that do not require the guarantees of these specifications.
9096 @item -fno-math-errno
9097 @opindex fno-math-errno
9098 Do not set @code{errno} after calling math functions that are executed
9099 with a single instruction, e.g., @code{sqrt}. A program that relies on
9100 IEEE exceptions for math error handling may want to use this flag
9101 for speed while maintaining IEEE arithmetic compatibility.
9103 This option is not turned on by any @option{-O} option since
9104 it can result in incorrect output for programs that depend on
9105 an exact implementation of IEEE or ISO rules/specifications for
9106 math functions. It may, however, yield faster code for programs
9107 that do not require the guarantees of these specifications.
9109 The default is @option{-fmath-errno}.
9111 On Darwin systems, the math library never sets @code{errno}. There is
9112 therefore no reason for the compiler to consider the possibility that
9113 it might, and @option{-fno-math-errno} is the default.
9115 @item -funsafe-math-optimizations
9116 @opindex funsafe-math-optimizations
9118 Allow optimizations for floating-point arithmetic that (a) assume
9119 that arguments and results are valid and (b) may violate IEEE or
9120 ANSI standards. When used at link-time, it may include libraries
9121 or startup files that change the default FPU control word or other
9122 similar optimizations.
9124 This option is not turned on by any @option{-O} option since
9125 it can result in incorrect output for programs that depend on
9126 an exact implementation of IEEE or ISO rules/specifications for
9127 math functions. It may, however, yield faster code for programs
9128 that do not require the guarantees of these specifications.
9129 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9130 @option{-fassociative-math} and @option{-freciprocal-math}.
9132 The default is @option{-fno-unsafe-math-optimizations}.
9134 @item -fassociative-math
9135 @opindex fassociative-math
9137 Allow re-association of operands in series of floating-point operations.
9138 This violates the ISO C and C++ language standard by possibly changing
9139 computation result. NOTE: re-ordering may change the sign of zero as
9140 well as ignore NaNs and inhibit or create underflow or overflow (and
9141 thus cannot be used on code that relies on rounding behavior like
9142 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9143 and thus may not be used when ordered comparisons are required.
9144 This option requires that both @option{-fno-signed-zeros} and
9145 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9146 much sense with @option{-frounding-math}. For Fortran the option
9147 is automatically enabled when both @option{-fno-signed-zeros} and
9148 @option{-fno-trapping-math} are in effect.
9150 The default is @option{-fno-associative-math}.
9152 @item -freciprocal-math
9153 @opindex freciprocal-math
9155 Allow the reciprocal of a value to be used instead of dividing by
9156 the value if this enables optimizations. For example @code{x / y}
9157 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9158 is subject to common subexpression elimination. Note that this loses
9159 precision and increases the number of flops operating on the value.
9161 The default is @option{-fno-reciprocal-math}.
9163 @item -ffinite-math-only
9164 @opindex ffinite-math-only
9165 Allow optimizations for floating-point arithmetic that assume
9166 that arguments and results are not NaNs or +-Infs.
9168 This option is not turned on by any @option{-O} option since
9169 it can result in incorrect output for programs that depend on
9170 an exact implementation of IEEE or ISO rules/specifications for
9171 math functions. It may, however, yield faster code for programs
9172 that do not require the guarantees of these specifications.
9174 The default is @option{-fno-finite-math-only}.
9176 @item -fno-signed-zeros
9177 @opindex fno-signed-zeros
9178 Allow optimizations for floating-point arithmetic that ignore the
9179 signedness of zero. IEEE arithmetic specifies the behavior of
9180 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9181 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9182 This option implies that the sign of a zero result isn't significant.
9184 The default is @option{-fsigned-zeros}.
9186 @item -fno-trapping-math
9187 @opindex fno-trapping-math
9188 Compile code assuming that floating-point operations cannot generate
9189 user-visible traps. These traps include division by zero, overflow,
9190 underflow, inexact result and invalid operation. This option requires
9191 that @option{-fno-signaling-nans} be in effect. Setting this option may
9192 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9194 This option should never be turned on by any @option{-O} option since
9195 it can result in incorrect output for programs that depend on
9196 an exact implementation of IEEE or ISO rules/specifications for
9199 The default is @option{-ftrapping-math}.
9201 @item -frounding-math
9202 @opindex frounding-math
9203 Disable transformations and optimizations that assume default floating-point
9204 rounding behavior. This is round-to-zero for all floating point
9205 to integer conversions, and round-to-nearest for all other arithmetic
9206 truncations. This option should be specified for programs that change
9207 the FP rounding mode dynamically, or that may be executed with a
9208 non-default rounding mode. This option disables constant folding of
9209 floating-point expressions at compile time (which may be affected by
9210 rounding mode) and arithmetic transformations that are unsafe in the
9211 presence of sign-dependent rounding modes.
9213 The default is @option{-fno-rounding-math}.
9215 This option is experimental and does not currently guarantee to
9216 disable all GCC optimizations that are affected by rounding mode.
9217 Future versions of GCC may provide finer control of this setting
9218 using C99's @code{FENV_ACCESS} pragma. This command-line option
9219 will be used to specify the default state for @code{FENV_ACCESS}.
9221 @item -fsignaling-nans
9222 @opindex fsignaling-nans
9223 Compile code assuming that IEEE signaling NaNs may generate user-visible
9224 traps during floating-point operations. Setting this option disables
9225 optimizations that may change the number of exceptions visible with
9226 signaling NaNs. This option implies @option{-ftrapping-math}.
9228 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9231 The default is @option{-fno-signaling-nans}.
9233 This option is experimental and does not currently guarantee to
9234 disable all GCC optimizations that affect signaling NaN behavior.
9236 @item -fsingle-precision-constant
9237 @opindex fsingle-precision-constant
9238 Treat floating-point constants as single precision instead of
9239 implicitly converting them to double-precision constants.
9241 @item -fcx-limited-range
9242 @opindex fcx-limited-range
9243 When enabled, this option states that a range reduction step is not
9244 needed when performing complex division. Also, there is no checking
9245 whether the result of a complex multiplication or division is @code{NaN
9246 + I*NaN}, with an attempt to rescue the situation in that case. The
9247 default is @option{-fno-cx-limited-range}, but is enabled by
9248 @option{-ffast-math}.
9250 This option controls the default setting of the ISO C99
9251 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9254 @item -fcx-fortran-rules
9255 @opindex fcx-fortran-rules
9256 Complex multiplication and division follow Fortran rules. Range
9257 reduction is done as part of complex division, but there is no checking
9258 whether the result of a complex multiplication or division is @code{NaN
9259 + I*NaN}, with an attempt to rescue the situation in that case.
9261 The default is @option{-fno-cx-fortran-rules}.
9265 The following options control optimizations that may improve
9266 performance, but are not enabled by any @option{-O} options. This
9267 section includes experimental options that may produce broken code.
9270 @item -fbranch-probabilities
9271 @opindex fbranch-probabilities
9272 After running a program compiled with @option{-fprofile-arcs}
9273 (@pxref{Debugging Options,, Options for Debugging Your Program or
9274 @command{gcc}}), you can compile it a second time using
9275 @option{-fbranch-probabilities}, to improve optimizations based on
9276 the number of times each branch was taken. When a program
9277 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9278 counts to a file called @file{@var{sourcename}.gcda} for each source
9279 file. The information in this data file is very dependent on the
9280 structure of the generated code, so you must use the same source code
9281 and the same optimization options for both compilations.
9283 With @option{-fbranch-probabilities}, GCC puts a
9284 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9285 These can be used to improve optimization. Currently, they are only
9286 used in one place: in @file{reorg.c}, instead of guessing which path a
9287 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9288 exactly determine which path is taken more often.
9290 @item -fprofile-values
9291 @opindex fprofile-values
9292 If combined with @option{-fprofile-arcs}, it adds code so that some
9293 data about values of expressions in the program is gathered.
9295 With @option{-fbranch-probabilities}, it reads back the data gathered
9296 from profiling values of expressions for usage in optimizations.
9298 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9300 @item -fprofile-reorder-functions
9301 @opindex fprofile-reorder-functions
9302 Function reordering based on profile instrumentation collects
9303 first time of execution of a function and orders these functions
9306 Enabled with @option{-fprofile-use}.
9310 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9311 to add code to gather information about values of expressions.
9313 With @option{-fbranch-probabilities}, it reads back the data gathered
9314 and actually performs the optimizations based on them.
9315 Currently the optimizations include specialization of division operations
9316 using the knowledge about the value of the denominator.
9318 @item -frename-registers
9319 @opindex frename-registers
9320 Attempt to avoid false dependencies in scheduled code by making use
9321 of registers left over after register allocation. This optimization
9322 most benefits processors with lots of registers. Depending on the
9323 debug information format adopted by the target, however, it can
9324 make debugging impossible, since variables no longer stay in
9325 a ``home register''.
9327 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9331 Perform tail duplication to enlarge superblock size. This transformation
9332 simplifies the control flow of the function allowing other optimizations to do
9335 Enabled with @option{-fprofile-use}.
9337 @item -funroll-loops
9338 @opindex funroll-loops
9339 Unroll loops whose number of iterations can be determined at compile time or
9340 upon entry to the loop. @option{-funroll-loops} implies
9341 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9342 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9343 a small constant number of iterations). This option makes code larger, and may
9344 or may not make it run faster.
9346 Enabled with @option{-fprofile-use}.
9348 @item -funroll-all-loops
9349 @opindex funroll-all-loops
9350 Unroll all loops, even if their number of iterations is uncertain when
9351 the loop is entered. This usually makes programs run more slowly.
9352 @option{-funroll-all-loops} implies the same options as
9353 @option{-funroll-loops}.
9356 @opindex fpeel-loops
9357 Peels loops for which there is enough information that they do not
9358 roll much (from profile feedback). It also turns on complete loop peeling
9359 (i.e.@: complete removal of loops with small constant number of iterations).
9361 Enabled with @option{-fprofile-use}.
9363 @item -fmove-loop-invariants
9364 @opindex fmove-loop-invariants
9365 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9366 at level @option{-O1}
9368 @item -funswitch-loops
9369 @opindex funswitch-loops
9370 Move branches with loop invariant conditions out of the loop, with duplicates
9371 of the loop on both branches (modified according to result of the condition).
9373 @item -ffunction-sections
9374 @itemx -fdata-sections
9375 @opindex ffunction-sections
9376 @opindex fdata-sections
9377 Place each function or data item into its own section in the output
9378 file if the target supports arbitrary sections. The name of the
9379 function or the name of the data item determines the section's name
9382 Use these options on systems where the linker can perform optimizations
9383 to improve locality of reference in the instruction space. Most systems
9384 using the ELF object format and SPARC processors running Solaris 2 have
9385 linkers with such optimizations. AIX may have these optimizations in
9388 Only use these options when there are significant benefits from doing
9389 so. When you specify these options, the assembler and linker
9390 create larger object and executable files and are also slower.
9391 You cannot use @code{gprof} on all systems if you
9392 specify this option, and you may have problems with debugging if
9393 you specify both this option and @option{-g}.
9395 @item -fbranch-target-load-optimize
9396 @opindex fbranch-target-load-optimize
9397 Perform branch target register load optimization before prologue / epilogue
9399 The use of target registers can typically be exposed only during reload,
9400 thus hoisting loads out of loops and doing inter-block scheduling needs
9401 a separate optimization pass.
9403 @item -fbranch-target-load-optimize2
9404 @opindex fbranch-target-load-optimize2
9405 Perform branch target register load optimization after prologue / epilogue
9408 @item -fbtr-bb-exclusive
9409 @opindex fbtr-bb-exclusive
9410 When performing branch target register load optimization, don't reuse
9411 branch target registers within any basic block.
9413 @item -fstack-protector
9414 @opindex fstack-protector
9415 Emit extra code to check for buffer overflows, such as stack smashing
9416 attacks. This is done by adding a guard variable to functions with
9417 vulnerable objects. This includes functions that call @code{alloca}, and
9418 functions with buffers larger than 8 bytes. The guards are initialized
9419 when a function is entered and then checked when the function exits.
9420 If a guard check fails, an error message is printed and the program exits.
9422 @item -fstack-protector-all
9423 @opindex fstack-protector-all
9424 Like @option{-fstack-protector} except that all functions are protected.
9426 @item -fstack-protector-strong
9427 @opindex fstack-protector-strong
9428 Like @option{-fstack-protector} but includes additional functions to
9429 be protected --- those that have local array definitions, or have
9430 references to local frame addresses.
9432 @item -fsection-anchors
9433 @opindex fsection-anchors
9434 Try to reduce the number of symbolic address calculations by using
9435 shared ``anchor'' symbols to address nearby objects. This transformation
9436 can help to reduce the number of GOT entries and GOT accesses on some
9439 For example, the implementation of the following function @code{foo}:
9443 int foo (void) @{ return a + b + c; @}
9447 usually calculates the addresses of all three variables, but if you
9448 compile it with @option{-fsection-anchors}, it accesses the variables
9449 from a common anchor point instead. The effect is similar to the
9450 following pseudocode (which isn't valid C):
9455 register int *xr = &x;
9456 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9460 Not all targets support this option.
9462 @item --param @var{name}=@var{value}
9464 In some places, GCC uses various constants to control the amount of
9465 optimization that is done. For example, GCC does not inline functions
9466 that contain more than a certain number of instructions. You can
9467 control some of these constants on the command line using the
9468 @option{--param} option.
9470 The names of specific parameters, and the meaning of the values, are
9471 tied to the internals of the compiler, and are subject to change
9472 without notice in future releases.
9474 In each case, the @var{value} is an integer. The allowable choices for
9478 @item predictable-branch-outcome
9479 When branch is predicted to be taken with probability lower than this threshold
9480 (in percent), then it is considered well predictable. The default is 10.
9482 @item max-crossjump-edges
9483 The maximum number of incoming edges to consider for cross-jumping.
9484 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9485 the number of edges incoming to each block. Increasing values mean
9486 more aggressive optimization, making the compilation time increase with
9487 probably small improvement in executable size.
9489 @item min-crossjump-insns
9490 The minimum number of instructions that must be matched at the end
9491 of two blocks before cross-jumping is performed on them. This
9492 value is ignored in the case where all instructions in the block being
9493 cross-jumped from are matched. The default value is 5.
9495 @item max-grow-copy-bb-insns
9496 The maximum code size expansion factor when copying basic blocks
9497 instead of jumping. The expansion is relative to a jump instruction.
9498 The default value is 8.
9500 @item max-goto-duplication-insns
9501 The maximum number of instructions to duplicate to a block that jumps
9502 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9503 passes, GCC factors computed gotos early in the compilation process,
9504 and unfactors them as late as possible. Only computed jumps at the
9505 end of a basic blocks with no more than max-goto-duplication-insns are
9506 unfactored. The default value is 8.
9508 @item max-delay-slot-insn-search
9509 The maximum number of instructions to consider when looking for an
9510 instruction to fill a delay slot. If more than this arbitrary number of
9511 instructions are searched, the time savings from filling the delay slot
9512 are minimal, so stop searching. Increasing values mean more
9513 aggressive optimization, making the compilation time increase with probably
9514 small improvement in execution time.
9516 @item max-delay-slot-live-search
9517 When trying to fill delay slots, the maximum number of instructions to
9518 consider when searching for a block with valid live register
9519 information. Increasing this arbitrarily chosen value means more
9520 aggressive optimization, increasing the compilation time. This parameter
9521 should be removed when the delay slot code is rewritten to maintain the
9524 @item max-gcse-memory
9525 The approximate maximum amount of memory that can be allocated in
9526 order to perform the global common subexpression elimination
9527 optimization. If more memory than specified is required, the
9528 optimization is not done.
9530 @item max-gcse-insertion-ratio
9531 If the ratio of expression insertions to deletions is larger than this value
9532 for any expression, then RTL PRE inserts or removes the expression and thus
9533 leaves partially redundant computations in the instruction stream. The default value is 20.
9535 @item max-pending-list-length
9536 The maximum number of pending dependencies scheduling allows
9537 before flushing the current state and starting over. Large functions
9538 with few branches or calls can create excessively large lists which
9539 needlessly consume memory and resources.
9541 @item max-modulo-backtrack-attempts
9542 The maximum number of backtrack attempts the scheduler should make
9543 when modulo scheduling a loop. Larger values can exponentially increase
9546 @item max-inline-insns-single
9547 Several parameters control the tree inliner used in GCC@.
9548 This number sets the maximum number of instructions (counted in GCC's
9549 internal representation) in a single function that the tree inliner
9550 considers for inlining. This only affects functions declared
9551 inline and methods implemented in a class declaration (C++).
9552 The default value is 400.
9554 @item max-inline-insns-auto
9555 When you use @option{-finline-functions} (included in @option{-O3}),
9556 a lot of functions that would otherwise not be considered for inlining
9557 by the compiler are investigated. To those functions, a different
9558 (more restrictive) limit compared to functions declared inline can
9560 The default value is 40.
9562 @item inline-min-speedup
9563 When estimated performance improvement of caller + callee runtime exceeds this
9564 threshold (in precent), the function can be inlined regardless the limit on
9565 @option{--param max-inline-insns-single} and @option{--param
9566 max-inline-insns-auto}.
9568 @item large-function-insns
9569 The limit specifying really large functions. For functions larger than this
9570 limit after inlining, inlining is constrained by
9571 @option{--param large-function-growth}. This parameter is useful primarily
9572 to avoid extreme compilation time caused by non-linear algorithms used by the
9574 The default value is 2700.
9576 @item large-function-growth
9577 Specifies maximal growth of large function caused by inlining in percents.
9578 The default value is 100 which limits large function growth to 2.0 times
9581 @item large-unit-insns
9582 The limit specifying large translation unit. Growth caused by inlining of
9583 units larger than this limit is limited by @option{--param inline-unit-growth}.
9584 For small units this might be too tight.
9585 For example, consider a unit consisting of function A
9586 that is inline and B that just calls A three times. If B is small relative to
9587 A, the growth of unit is 300\% and yet such inlining is very sane. For very
9588 large units consisting of small inlineable functions, however, the overall unit
9589 growth limit is needed to avoid exponential explosion of code size. Thus for
9590 smaller units, the size is increased to @option{--param large-unit-insns}
9591 before applying @option{--param inline-unit-growth}. The default is 10000.
9593 @item inline-unit-growth
9594 Specifies maximal overall growth of the compilation unit caused by inlining.
9595 The default value is 30 which limits unit growth to 1.3 times the original
9596 size. Cold functions (either marked cold via an attribibute or by profile
9597 feedback) are not accounted into the unit size.
9599 @item ipcp-unit-growth
9600 Specifies maximal overall growth of the compilation unit caused by
9601 interprocedural constant propagation. The default value is 10 which limits
9602 unit growth to 1.1 times the original size.
9604 @item large-stack-frame
9605 The limit specifying large stack frames. While inlining the algorithm is trying
9606 to not grow past this limit too much. The default value is 256 bytes.
9608 @item large-stack-frame-growth
9609 Specifies maximal growth of large stack frames caused by inlining in percents.
9610 The default value is 1000 which limits large stack frame growth to 11 times
9613 @item max-inline-insns-recursive
9614 @itemx max-inline-insns-recursive-auto
9615 Specifies the maximum number of instructions an out-of-line copy of a
9616 self-recursive inline
9617 function can grow into by performing recursive inlining.
9619 For functions declared inline, @option{--param max-inline-insns-recursive} is
9620 taken into account. For functions not declared inline, recursive inlining
9621 happens only when @option{-finline-functions} (included in @option{-O3}) is
9622 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
9623 default value is 450.
9625 @item max-inline-recursive-depth
9626 @itemx max-inline-recursive-depth-auto
9627 Specifies the maximum recursion depth used for recursive inlining.
9629 For functions declared inline, @option{--param max-inline-recursive-depth} is
9630 taken into account. For functions not declared inline, recursive inlining
9631 happens only when @option{-finline-functions} (included in @option{-O3}) is
9632 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
9635 @item min-inline-recursive-probability
9636 Recursive inlining is profitable only for function having deep recursion
9637 in average and can hurt for function having little recursion depth by
9638 increasing the prologue size or complexity of function body to other
9641 When profile feedback is available (see @option{-fprofile-generate}) the actual
9642 recursion depth can be guessed from probability that function recurses via a
9643 given call expression. This parameter limits inlining only to call expressions
9644 whose probability exceeds the given threshold (in percents).
9645 The default value is 10.
9647 @item early-inlining-insns
9648 Specify growth that the early inliner can make. In effect it increases
9649 the amount of inlining for code having a large abstraction penalty.
9650 The default value is 10.
9652 @item max-early-inliner-iterations
9653 @itemx max-early-inliner-iterations
9654 Limit of iterations of the early inliner. This basically bounds
9655 the number of nested indirect calls the early inliner can resolve.
9656 Deeper chains are still handled by late inlining.
9658 @item comdat-sharing-probability
9659 @itemx comdat-sharing-probability
9660 Probability (in percent) that C++ inline function with comdat visibility
9661 are shared across multiple compilation units. The default value is 20.
9663 @item profile-func-internal-id
9664 @itemx profile-func-internal-id
9665 A parameter to control whether to use function internal id in profile
9666 database lookup. If the value is 0, the compiler will use id that
9667 is based on function assembler name and filename, which makes old profile
9668 data more tolerant to source changes such as function reordering etc.
9669 The default value is 0.
9671 @item min-vect-loop-bound
9672 The minimum number of iterations under which loops are not vectorized
9673 when @option{-ftree-vectorize} is used. The number of iterations after
9674 vectorization needs to be greater than the value specified by this option
9675 to allow vectorization. The default value is 0.
9677 @item gcse-cost-distance-ratio
9678 Scaling factor in calculation of maximum distance an expression
9679 can be moved by GCSE optimizations. This is currently supported only in the
9680 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9681 is with simple expressions, i.e., the expressions that have cost
9682 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9683 hoisting of simple expressions. The default value is 10.
9685 @item gcse-unrestricted-cost
9686 Cost, roughly measured as the cost of a single typical machine
9687 instruction, at which GCSE optimizations do not constrain
9688 the distance an expression can travel. This is currently
9689 supported only in the code hoisting pass. The lesser the cost,
9690 the more aggressive code hoisting is. Specifying 0
9691 allows all expressions to travel unrestricted distances.
9692 The default value is 3.
9694 @item max-hoist-depth
9695 The depth of search in the dominator tree for expressions to hoist.
9696 This is used to avoid quadratic behavior in hoisting algorithm.
9697 The value of 0 does not limit on the search, but may slow down compilation
9698 of huge functions. The default value is 30.
9700 @item max-tail-merge-comparisons
9701 The maximum amount of similar bbs to compare a bb with. This is used to
9702 avoid quadratic behavior in tree tail merging. The default value is 10.
9704 @item max-tail-merge-iterations
9705 The maximum amount of iterations of the pass over the function. This is used to
9706 limit compilation time in tree tail merging. The default value is 2.
9708 @item max-unrolled-insns
9709 The maximum number of instructions that a loop may have to be unrolled.
9710 If a loop is unrolled, this parameter also determines how many times
9711 the loop code is unrolled.
9713 @item max-average-unrolled-insns
9714 The maximum number of instructions biased by probabilities of their execution
9715 that a loop may have to be unrolled. If a loop is unrolled,
9716 this parameter also determines how many times the loop code is unrolled.
9718 @item max-unroll-times
9719 The maximum number of unrollings of a single loop.
9721 @item max-peeled-insns
9722 The maximum number of instructions that a loop may have to be peeled.
9723 If a loop is peeled, this parameter also determines how many times
9724 the loop code is peeled.
9726 @item max-peel-times
9727 The maximum number of peelings of a single loop.
9729 @item max-peel-branches
9730 The maximum number of branches on the hot path through the peeled sequence.
9732 @item max-completely-peeled-insns
9733 The maximum number of insns of a completely peeled loop.
9735 @item max-completely-peel-times
9736 The maximum number of iterations of a loop to be suitable for complete peeling.
9738 @item max-completely-peel-loop-nest-depth
9739 The maximum depth of a loop nest suitable for complete peeling.
9741 @item max-unswitch-insns
9742 The maximum number of insns of an unswitched loop.
9744 @item max-unswitch-level
9745 The maximum number of branches unswitched in a single loop.
9748 The minimum cost of an expensive expression in the loop invariant motion.
9750 @item iv-consider-all-candidates-bound
9751 Bound on number of candidates for induction variables, below which
9752 all candidates are considered for each use in induction variable
9753 optimizations. If there are more candidates than this,
9754 only the most relevant ones are considered to avoid quadratic time complexity.
9756 @item iv-max-considered-uses
9757 The induction variable optimizations give up on loops that contain more
9758 induction variable uses.
9760 @item iv-always-prune-cand-set-bound
9761 If the number of candidates in the set is smaller than this value,
9762 always try to remove unnecessary ivs from the set
9763 when adding a new one.
9765 @item scev-max-expr-size
9766 Bound on size of expressions used in the scalar evolutions analyzer.
9767 Large expressions slow the analyzer.
9769 @item scev-max-expr-complexity
9770 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9771 Complex expressions slow the analyzer.
9773 @item omega-max-vars
9774 The maximum number of variables in an Omega constraint system.
9775 The default value is 128.
9777 @item omega-max-geqs
9778 The maximum number of inequalities in an Omega constraint system.
9779 The default value is 256.
9782 The maximum number of equalities in an Omega constraint system.
9783 The default value is 128.
9785 @item omega-max-wild-cards
9786 The maximum number of wildcard variables that the Omega solver is
9787 able to insert. The default value is 18.
9789 @item omega-hash-table-size
9790 The size of the hash table in the Omega solver. The default value is
9793 @item omega-max-keys
9794 The maximal number of keys used by the Omega solver. The default
9797 @item omega-eliminate-redundant-constraints
9798 When set to 1, use expensive methods to eliminate all redundant
9799 constraints. The default value is 0.
9801 @item vect-max-version-for-alignment-checks
9802 The maximum number of run-time checks that can be performed when
9803 doing loop versioning for alignment in the vectorizer.
9805 @item vect-max-version-for-alias-checks
9806 The maximum number of run-time checks that can be performed when
9807 doing loop versioning for alias in the vectorizer.
9809 @item vect-max-peeling-for-alignment
9810 The maximum number of loop peels to enhance access alignment
9811 for vectorizer. Value -1 means 'no limit'.
9813 @item max-iterations-to-track
9814 The maximum number of iterations of a loop the brute-force algorithm
9815 for analysis of the number of iterations of the loop tries to evaluate.
9817 @item hot-bb-count-ws-permille
9818 A basic block profile count is considered hot if it contributes to
9819 the given permillage (i.e. 0...1000) of the entire profiled execution.
9821 @item hot-bb-frequency-fraction
9822 Select fraction of the entry block frequency of executions of basic block in
9823 function given basic block needs to have to be considered hot.
9825 @item max-predicted-iterations
9826 The maximum number of loop iterations we predict statically. This is useful
9827 in cases where a function contains a single loop with known bound and
9828 another loop with unknown bound.
9829 The known number of iterations is predicted correctly, while
9830 the unknown number of iterations average to roughly 10. This means that the
9831 loop without bounds appears artificially cold relative to the other one.
9833 @item builtin-expect-probability
9834 Control the probability of the expression having the specified value. This
9835 parameter takes a percentage (i.e. 0 ... 100) as input.
9836 The default probability of 90 is obtained empirically.
9838 @item align-threshold
9840 Select fraction of the maximal frequency of executions of a basic block in
9841 a function to align the basic block.
9843 @item align-loop-iterations
9845 A loop expected to iterate at least the selected number of iterations is
9848 @item tracer-dynamic-coverage
9849 @itemx tracer-dynamic-coverage-feedback
9851 This value is used to limit superblock formation once the given percentage of
9852 executed instructions is covered. This limits unnecessary code size
9855 The @option{tracer-dynamic-coverage-feedback} is used only when profile
9856 feedback is available. The real profiles (as opposed to statically estimated
9857 ones) are much less balanced allowing the threshold to be larger value.
9859 @item tracer-max-code-growth
9860 Stop tail duplication once code growth has reached given percentage. This is
9861 a rather artificial limit, as most of the duplicates are eliminated later in
9862 cross jumping, so it may be set to much higher values than is the desired code
9865 @item tracer-min-branch-ratio
9867 Stop reverse growth when the reverse probability of best edge is less than this
9868 threshold (in percent).
9870 @item tracer-min-branch-ratio
9871 @itemx tracer-min-branch-ratio-feedback
9873 Stop forward growth if the best edge has probability lower than this
9876 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
9877 compilation for profile feedback and one for compilation without. The value
9878 for compilation with profile feedback needs to be more conservative (higher) in
9879 order to make tracer effective.
9881 @item max-cse-path-length
9883 The maximum number of basic blocks on path that CSE considers.
9887 The maximum number of instructions CSE processes before flushing.
9888 The default is 1000.
9890 @item ggc-min-expand
9892 GCC uses a garbage collector to manage its own memory allocation. This
9893 parameter specifies the minimum percentage by which the garbage
9894 collector's heap should be allowed to expand between collections.
9895 Tuning this may improve compilation speed; it has no effect on code
9898 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9899 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9900 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9901 GCC is not able to calculate RAM on a particular platform, the lower
9902 bound of 30% is used. Setting this parameter and
9903 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9904 every opportunity. This is extremely slow, but can be useful for
9907 @item ggc-min-heapsize
9909 Minimum size of the garbage collector's heap before it begins bothering
9910 to collect garbage. The first collection occurs after the heap expands
9911 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9912 tuning this may improve compilation speed, and has no effect on code
9915 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9916 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9917 with a lower bound of 4096 (four megabytes) and an upper bound of
9918 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9919 particular platform, the lower bound is used. Setting this parameter
9920 very large effectively disables garbage collection. Setting this
9921 parameter and @option{ggc-min-expand} to zero causes a full collection
9922 to occur at every opportunity.
9924 @item max-reload-search-insns
9925 The maximum number of instruction reload should look backward for equivalent
9926 register. Increasing values mean more aggressive optimization, making the
9927 compilation time increase with probably slightly better performance.
9928 The default value is 100.
9930 @item max-cselib-memory-locations
9931 The maximum number of memory locations cselib should take into account.
9932 Increasing values mean more aggressive optimization, making the compilation time
9933 increase with probably slightly better performance. The default value is 500.
9935 @item reorder-blocks-duplicate
9936 @itemx reorder-blocks-duplicate-feedback
9938 Used by the basic block reordering pass to decide whether to use unconditional
9939 branch or duplicate the code on its destination. Code is duplicated when its
9940 estimated size is smaller than this value multiplied by the estimated size of
9941 unconditional jump in the hot spots of the program.
9943 The @option{reorder-block-duplicate-feedback} is used only when profile
9944 feedback is available. It may be set to higher values than
9945 @option{reorder-block-duplicate} since information about the hot spots is more
9948 @item max-sched-ready-insns
9949 The maximum number of instructions ready to be issued the scheduler should
9950 consider at any given time during the first scheduling pass. Increasing
9951 values mean more thorough searches, making the compilation time increase
9952 with probably little benefit. The default value is 100.
9954 @item max-sched-region-blocks
9955 The maximum number of blocks in a region to be considered for
9956 interblock scheduling. The default value is 10.
9958 @item max-pipeline-region-blocks
9959 The maximum number of blocks in a region to be considered for
9960 pipelining in the selective scheduler. The default value is 15.
9962 @item max-sched-region-insns
9963 The maximum number of insns in a region to be considered for
9964 interblock scheduling. The default value is 100.
9966 @item max-pipeline-region-insns
9967 The maximum number of insns in a region to be considered for
9968 pipelining in the selective scheduler. The default value is 200.
9971 The minimum probability (in percents) of reaching a source block
9972 for interblock speculative scheduling. The default value is 40.
9974 @item max-sched-extend-regions-iters
9975 The maximum number of iterations through CFG to extend regions.
9976 A value of 0 (the default) disables region extensions.
9978 @item max-sched-insn-conflict-delay
9979 The maximum conflict delay for an insn to be considered for speculative motion.
9980 The default value is 3.
9982 @item sched-spec-prob-cutoff
9983 The minimal probability of speculation success (in percents), so that
9984 speculative insns are scheduled.
9985 The default value is 40.
9987 @item sched-spec-state-edge-prob-cutoff
9988 The minimum probability an edge must have for the scheduler to save its
9990 The default value is 10.
9992 @item sched-mem-true-dep-cost
9993 Minimal distance (in CPU cycles) between store and load targeting same
9994 memory locations. The default value is 1.
9996 @item selsched-max-lookahead
9997 The maximum size of the lookahead window of selective scheduling. It is a
9998 depth of search for available instructions.
9999 The default value is 50.
10001 @item selsched-max-sched-times
10002 The maximum number of times that an instruction is scheduled during
10003 selective scheduling. This is the limit on the number of iterations
10004 through which the instruction may be pipelined. The default value is 2.
10006 @item selsched-max-insns-to-rename
10007 The maximum number of best instructions in the ready list that are considered
10008 for renaming in the selective scheduler. The default value is 2.
10011 The minimum value of stage count that swing modulo scheduler
10012 generates. The default value is 2.
10014 @item max-last-value-rtl
10015 The maximum size measured as number of RTLs that can be recorded in an expression
10016 in combiner for a pseudo register as last known value of that register. The default
10019 @item max-combine-insns
10020 The maximum number of instructions the RTL combiner tries to combine.
10021 The default value is 2 at @option{-Og} and 4 otherwise.
10023 @item integer-share-limit
10024 Small integer constants can use a shared data structure, reducing the
10025 compiler's memory usage and increasing its speed. This sets the maximum
10026 value of a shared integer constant. The default value is 256.
10028 @item ssp-buffer-size
10029 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10030 protection when @option{-fstack-protection} is used.
10032 @item min-size-for-stack-sharing
10033 The minimum size of variables taking part in stack slot sharing when not
10034 optimizing. The default value is 32.
10036 @item max-jump-thread-duplication-stmts
10037 Maximum number of statements allowed in a block that needs to be
10038 duplicated when threading jumps.
10040 @item max-fields-for-field-sensitive
10041 Maximum number of fields in a structure treated in
10042 a field sensitive manner during pointer analysis. The default is zero
10043 for @option{-O0} and @option{-O1},
10044 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10046 @item prefetch-latency
10047 Estimate on average number of instructions that are executed before
10048 prefetch finishes. The distance prefetched ahead is proportional
10049 to this constant. Increasing this number may also lead to less
10050 streams being prefetched (see @option{simultaneous-prefetches}).
10052 @item simultaneous-prefetches
10053 Maximum number of prefetches that can run at the same time.
10055 @item l1-cache-line-size
10056 The size of cache line in L1 cache, in bytes.
10058 @item l1-cache-size
10059 The size of L1 cache, in kilobytes.
10061 @item l2-cache-size
10062 The size of L2 cache, in kilobytes.
10064 @item min-insn-to-prefetch-ratio
10065 The minimum ratio between the number of instructions and the
10066 number of prefetches to enable prefetching in a loop.
10068 @item prefetch-min-insn-to-mem-ratio
10069 The minimum ratio between the number of instructions and the
10070 number of memory references to enable prefetching in a loop.
10072 @item use-canonical-types
10073 Whether the compiler should use the ``canonical'' type system. By
10074 default, this should always be 1, which uses a more efficient internal
10075 mechanism for comparing types in C++ and Objective-C++. However, if
10076 bugs in the canonical type system are causing compilation failures,
10077 set this value to 0 to disable canonical types.
10079 @item switch-conversion-max-branch-ratio
10080 Switch initialization conversion refuses to create arrays that are
10081 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10082 branches in the switch.
10084 @item max-partial-antic-length
10085 Maximum length of the partial antic set computed during the tree
10086 partial redundancy elimination optimization (@option{-ftree-pre}) when
10087 optimizing at @option{-O3} and above. For some sorts of source code
10088 the enhanced partial redundancy elimination optimization can run away,
10089 consuming all of the memory available on the host machine. This
10090 parameter sets a limit on the length of the sets that are computed,
10091 which prevents the runaway behavior. Setting a value of 0 for
10092 this parameter allows an unlimited set length.
10094 @item sccvn-max-scc-size
10095 Maximum size of a strongly connected component (SCC) during SCCVN
10096 processing. If this limit is hit, SCCVN processing for the whole
10097 function is not done and optimizations depending on it are
10098 disabled. The default maximum SCC size is 10000.
10100 @item sccvn-max-alias-queries-per-access
10101 Maximum number of alias-oracle queries we perform when looking for
10102 redundancies for loads and stores. If this limit is hit the search
10103 is aborted and the load or store is not considered redundant. The
10104 number of queries is algorithmically limited to the number of
10105 stores on all paths from the load to the function entry.
10106 The default maxmimum number of queries is 1000.
10108 @item ira-max-loops-num
10109 IRA uses regional register allocation by default. If a function
10110 contains more loops than the number given by this parameter, only at most
10111 the given number of the most frequently-executed loops form regions
10112 for regional register allocation. The default value of the
10115 @item ira-max-conflict-table-size
10116 Although IRA uses a sophisticated algorithm to compress the conflict
10117 table, the table can still require excessive amounts of memory for
10118 huge functions. If the conflict table for a function could be more
10119 than the size in MB given by this parameter, the register allocator
10120 instead uses a faster, simpler, and lower-quality
10121 algorithm that does not require building a pseudo-register conflict table.
10122 The default value of the parameter is 2000.
10124 @item ira-loop-reserved-regs
10125 IRA can be used to evaluate more accurate register pressure in loops
10126 for decisions to move loop invariants (see @option{-O3}). The number
10127 of available registers reserved for some other purposes is given
10128 by this parameter. The default value of the parameter is 2, which is
10129 the minimal number of registers needed by typical instructions.
10130 This value is the best found from numerous experiments.
10132 @item loop-invariant-max-bbs-in-loop
10133 Loop invariant motion can be very expensive, both in compilation time and
10134 in amount of needed compile-time memory, with very large loops. Loops
10135 with more basic blocks than this parameter won't have loop invariant
10136 motion optimization performed on them. The default value of the
10137 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10139 @item loop-max-datarefs-for-datadeps
10140 Building data dapendencies is expensive for very large loops. This
10141 parameter limits the number of data references in loops that are
10142 considered for data dependence analysis. These large loops are no
10143 handled by the optimizations using loop data dependencies.
10144 The default value is 1000.
10146 @item max-vartrack-size
10147 Sets a maximum number of hash table slots to use during variable
10148 tracking dataflow analysis of any function. If this limit is exceeded
10149 with variable tracking at assignments enabled, analysis for that
10150 function is retried without it, after removing all debug insns from
10151 the function. If the limit is exceeded even without debug insns, var
10152 tracking analysis is completely disabled for the function. Setting
10153 the parameter to zero makes it unlimited.
10155 @item max-vartrack-expr-depth
10156 Sets a maximum number of recursion levels when attempting to map
10157 variable names or debug temporaries to value expressions. This trades
10158 compilation time for more complete debug information. If this is set too
10159 low, value expressions that are available and could be represented in
10160 debug information may end up not being used; setting this higher may
10161 enable the compiler to find more complex debug expressions, but compile
10162 time and memory use may grow. The default is 12.
10164 @item min-nondebug-insn-uid
10165 Use uids starting at this parameter for nondebug insns. The range below
10166 the parameter is reserved exclusively for debug insns created by
10167 @option{-fvar-tracking-assignments}, but debug insns may get
10168 (non-overlapping) uids above it if the reserved range is exhausted.
10170 @item ipa-sra-ptr-growth-factor
10171 IPA-SRA replaces a pointer to an aggregate with one or more new
10172 parameters only when their cumulative size is less or equal to
10173 @option{ipa-sra-ptr-growth-factor} times the size of the original
10176 @item tm-max-aggregate-size
10177 When making copies of thread-local variables in a transaction, this
10178 parameter specifies the size in bytes after which variables are
10179 saved with the logging functions as opposed to save/restore code
10180 sequence pairs. This option only applies when using
10183 @item graphite-max-nb-scop-params
10184 To avoid exponential effects in the Graphite loop transforms, the
10185 number of parameters in a Static Control Part (SCoP) is bounded. The
10186 default value is 10 parameters. A variable whose value is unknown at
10187 compilation time and defined outside a SCoP is a parameter of the SCoP.
10189 @item graphite-max-bbs-per-function
10190 To avoid exponential effects in the detection of SCoPs, the size of
10191 the functions analyzed by Graphite is bounded. The default value is
10194 @item loop-block-tile-size
10195 Loop blocking or strip mining transforms, enabled with
10196 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10197 loop in the loop nest by a given number of iterations. The strip
10198 length can be changed using the @option{loop-block-tile-size}
10199 parameter. The default value is 51 iterations.
10201 @item ipa-cp-value-list-size
10202 IPA-CP attempts to track all possible values and types passed to a function's
10203 parameter in order to propagate them and perform devirtualization.
10204 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10205 stores per one formal parameter of a function.
10207 @item ipa-cp-eval-threshold
10208 IPA-CP calculates its own score of cloning profitability heuristics
10209 and performs those cloning opportunities with scores that exceed
10210 @option{ipa-cp-eval-threshold}.
10212 @item ipa-max-agg-items
10213 IPA-CP is also capable to propagate a number of scalar values passed
10214 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10215 number of such values per one parameter.
10217 @item ipa-cp-loop-hint-bonus
10218 When IPA-CP determines that a cloning candidate would make the number
10219 of iterations of a loop known, it adds a bonus of
10220 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10223 @item ipa-cp-array-index-hint-bonus
10224 When IPA-CP determines that a cloning candidate would make the index of
10225 an array access known, it adds a bonus of
10226 @option{ipa-cp-array-index-hint-bonus} to the profitability
10227 score of the candidate.
10229 @item ipa-max-aa-steps
10230 During its analysis of function bodies, IPA-CP employs alias analysis
10231 in order to track values pointed to by function parameters. In order
10232 not spend too much time analyzing huge functions, it will give up and
10233 consider all memory clobbered after examining
10234 @option{ipa-max-aa-steps} statements modifying memory.
10236 @item lto-partitions
10237 Specify desired number of partitions produced during WHOPR compilation.
10238 The number of partitions should exceed the number of CPUs used for compilation.
10239 The default value is 32.
10241 @item lto-minpartition
10242 Size of minimal partition for WHOPR (in estimated instructions).
10243 This prevents expenses of splitting very small programs into too many
10246 @item cxx-max-namespaces-for-diagnostic-help
10247 The maximum number of namespaces to consult for suggestions when C++
10248 name lookup fails for an identifier. The default is 1000.
10250 @item sink-frequency-threshold
10251 The maximum relative execution frequency (in percents) of the target block
10252 relative to a statement's original block to allow statement sinking of a
10253 statement. Larger numbers result in more aggressive statement sinking.
10254 The default value is 75. A small positive adjustment is applied for
10255 statements with memory operands as those are even more profitable so sink.
10257 @item max-stores-to-sink
10258 The maximum number of conditional stores paires that can be sunk. Set to 0
10259 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10260 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10262 @item allow-store-data-races
10263 Allow optimizers to introduce new data races on stores.
10264 Set to 1 to allow, otherwise to 0. This option is enabled by default
10265 at optimization level @option{-Ofast}.
10267 @item case-values-threshold
10268 The smallest number of different values for which it is best to use a
10269 jump-table instead of a tree of conditional branches. If the value is
10270 0, use the default for the machine. The default is 0.
10272 @item tree-reassoc-width
10273 Set the maximum number of instructions executed in parallel in
10274 reassociated tree. This parameter overrides target dependent
10275 heuristics used by default if has non zero value.
10277 @item sched-pressure-algorithm
10278 Choose between the two available implementations of
10279 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10280 and is the more likely to prevent instructions from being reordered.
10281 Algorithm 2 was designed to be a compromise between the relatively
10282 conservative approach taken by algorithm 1 and the rather aggressive
10283 approach taken by the default scheduler. It relies more heavily on
10284 having a regular register file and accurate register pressure classes.
10285 See @file{haifa-sched.c} in the GCC sources for more details.
10287 The default choice depends on the target.
10289 @item max-slsr-cand-scan
10290 Set the maximum number of existing candidates that will be considered when
10291 seeking a basis for a new straight-line strength reduction candidate.
10294 Enable buffer overflow detection for global objects. This kind
10295 of protection is enabled by default if you are using
10296 @option{-fsanitize=address} option.
10297 To disable global objects protection use @option{--param asan-globals=0}.
10300 Enable buffer overflow detection for stack objects. This kind of
10301 protection is enabled by default when using@option{-fsanitize=address}.
10302 To disable stack protection use @option{--param asan-stack=0} option.
10304 @item asan-instrument-reads
10305 Enable buffer overflow detection for memory reads. This kind of
10306 protection is enabled by default when using @option{-fsanitize=address}.
10307 To disable memory reads protection use
10308 @option{--param asan-instrument-reads=0}.
10310 @item asan-instrument-writes
10311 Enable buffer overflow detection for memory writes. This kind of
10312 protection is enabled by default when using @option{-fsanitize=address}.
10313 To disable memory writes protection use
10314 @option{--param asan-instrument-writes=0} option.
10316 @item asan-memintrin
10317 Enable detection for built-in functions. This kind of protection
10318 is enabled by default when using @option{-fsanitize=address}.
10319 To disable built-in functions protection use
10320 @option{--param asan-memintrin=0}.
10322 @item asan-use-after-return
10323 Enable detection of use-after-return. This kind of protection
10324 is enabled by default when using @option{-fsanitize=address} option.
10325 To disable use-after-return detection use
10326 @option{--param asan-use-after-return=0}.
10328 @item asan-instrumentation-with-call-threshold
10329 Once number of memory accesses in function becomes greater
10330 or equal than this number, use callbacks instead of
10331 generating inline code. E.g. to disable inline code use
10332 @option{--param asan-instrumentation-with-call-threshold=0}.
10337 @node Preprocessor Options
10338 @section Options Controlling the Preprocessor
10339 @cindex preprocessor options
10340 @cindex options, preprocessor
10342 These options control the C preprocessor, which is run on each C source
10343 file before actual compilation.
10345 If you use the @option{-E} option, nothing is done except preprocessing.
10346 Some of these options make sense only together with @option{-E} because
10347 they cause the preprocessor output to be unsuitable for actual
10351 @item -Wp,@var{option}
10353 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10354 and pass @var{option} directly through to the preprocessor. If
10355 @var{option} contains commas, it is split into multiple options at the
10356 commas. However, many options are modified, translated or interpreted
10357 by the compiler driver before being passed to the preprocessor, and
10358 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10359 interface is undocumented and subject to change, so whenever possible
10360 you should avoid using @option{-Wp} and let the driver handle the
10363 @item -Xpreprocessor @var{option}
10364 @opindex Xpreprocessor
10365 Pass @var{option} as an option to the preprocessor. You can use this to
10366 supply system-specific preprocessor options that GCC does not
10369 If you want to pass an option that takes an argument, you must use
10370 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10372 @item -no-integrated-cpp
10373 @opindex no-integrated-cpp
10374 Perform preprocessing as a separate pass before compilation.
10375 By default, GCC performs preprocessing as an integrated part of
10376 input tokenization and parsing.
10377 If this option is provided, the appropriate language front end
10378 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10379 and Objective-C, respectively) is instead invoked twice,
10380 once for preprocessing only and once for actual compilation
10381 of the preprocessed input.
10382 This option may be useful in conjunction with the @option{-B} or
10383 @option{-wrapper} options to specify an alternate preprocessor or
10384 perform additional processing of the program source between
10385 normal preprocessing and compilation.
10388 @include cppopts.texi
10390 @node Assembler Options
10391 @section Passing Options to the Assembler
10393 @c prevent bad page break with this line
10394 You can pass options to the assembler.
10397 @item -Wa,@var{option}
10399 Pass @var{option} as an option to the assembler. If @var{option}
10400 contains commas, it is split into multiple options at the commas.
10402 @item -Xassembler @var{option}
10403 @opindex Xassembler
10404 Pass @var{option} as an option to the assembler. You can use this to
10405 supply system-specific assembler options that GCC does not
10408 If you want to pass an option that takes an argument, you must use
10409 @option{-Xassembler} twice, once for the option and once for the argument.
10414 @section Options for Linking
10415 @cindex link options
10416 @cindex options, linking
10418 These options come into play when the compiler links object files into
10419 an executable output file. They are meaningless if the compiler is
10420 not doing a link step.
10424 @item @var{object-file-name}
10425 A file name that does not end in a special recognized suffix is
10426 considered to name an object file or library. (Object files are
10427 distinguished from libraries by the linker according to the file
10428 contents.) If linking is done, these object files are used as input
10437 If any of these options is used, then the linker is not run, and
10438 object file names should not be used as arguments. @xref{Overall
10442 @item -l@var{library}
10443 @itemx -l @var{library}
10445 Search the library named @var{library} when linking. (The second
10446 alternative with the library as a separate argument is only for
10447 POSIX compliance and is not recommended.)
10449 It makes a difference where in the command you write this option; the
10450 linker searches and processes libraries and object files in the order they
10451 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10452 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10453 to functions in @samp{z}, those functions may not be loaded.
10455 The linker searches a standard list of directories for the library,
10456 which is actually a file named @file{lib@var{library}.a}. The linker
10457 then uses this file as if it had been specified precisely by name.
10459 The directories searched include several standard system directories
10460 plus any that you specify with @option{-L}.
10462 Normally the files found this way are library files---archive files
10463 whose members are object files. The linker handles an archive file by
10464 scanning through it for members which define symbols that have so far
10465 been referenced but not defined. But if the file that is found is an
10466 ordinary object file, it is linked in the usual fashion. The only
10467 difference between using an @option{-l} option and specifying a file name
10468 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10469 and searches several directories.
10473 You need this special case of the @option{-l} option in order to
10474 link an Objective-C or Objective-C++ program.
10476 @item -nostartfiles
10477 @opindex nostartfiles
10478 Do not use the standard system startup files when linking.
10479 The standard system libraries are used normally, unless @option{-nostdlib}
10480 or @option{-nodefaultlibs} is used.
10482 @item -nodefaultlibs
10483 @opindex nodefaultlibs
10484 Do not use the standard system libraries when linking.
10485 Only the libraries you specify are passed to the linker, and options
10486 specifying linkage of the system libraries, such as @code{-static-libgcc}
10487 or @code{-shared-libgcc}, are ignored.
10488 The standard startup files are used normally, unless @option{-nostartfiles}
10491 The compiler may generate calls to @code{memcmp},
10492 @code{memset}, @code{memcpy} and @code{memmove}.
10493 These entries are usually resolved by entries in
10494 libc. These entry points should be supplied through some other
10495 mechanism when this option is specified.
10499 Do not use the standard system startup files or libraries when linking.
10500 No startup files and only the libraries you specify are passed to
10501 the linker, and options specifying linkage of the system libraries, such as
10502 @code{-static-libgcc} or @code{-shared-libgcc}, are ignored.
10504 The compiler may generate calls to @code{memcmp}, @code{memset},
10505 @code{memcpy} and @code{memmove}.
10506 These entries are usually resolved by entries in
10507 libc. These entry points should be supplied through some other
10508 mechanism when this option is specified.
10510 @cindex @option{-lgcc}, use with @option{-nostdlib}
10511 @cindex @option{-nostdlib} and unresolved references
10512 @cindex unresolved references and @option{-nostdlib}
10513 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10514 @cindex @option{-nodefaultlibs} and unresolved references
10515 @cindex unresolved references and @option{-nodefaultlibs}
10516 One of the standard libraries bypassed by @option{-nostdlib} and
10517 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10518 which GCC uses to overcome shortcomings of particular machines, or special
10519 needs for some languages.
10520 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10521 Collection (GCC) Internals},
10522 for more discussion of @file{libgcc.a}.)
10523 In most cases, you need @file{libgcc.a} even when you want to avoid
10524 other standard libraries. In other words, when you specify @option{-nostdlib}
10525 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10526 This ensures that you have no unresolved references to internal GCC
10527 library subroutines.
10528 (An example of such an internal subroutine is @samp{__main}, used to ensure C++
10529 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10530 GNU Compiler Collection (GCC) Internals}.)
10534 Produce a position independent executable on targets that support it.
10535 For predictable results, you must also specify the same set of options
10536 used for compilation (@option{-fpie}, @option{-fPIE},
10537 or model suboptions) when you specify this linker option.
10541 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10542 that support it. This instructs the linker to add all symbols, not
10543 only used ones, to the dynamic symbol table. This option is needed
10544 for some uses of @code{dlopen} or to allow obtaining backtraces
10545 from within a program.
10549 Remove all symbol table and relocation information from the executable.
10553 On systems that support dynamic linking, this prevents linking with the shared
10554 libraries. On other systems, this option has no effect.
10558 Produce a shared object which can then be linked with other objects to
10559 form an executable. Not all systems support this option. For predictable
10560 results, you must also specify the same set of options used for compilation
10561 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
10562 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
10563 needs to build supplementary stub code for constructors to work. On
10564 multi-libbed systems, @samp{gcc -shared} must select the correct support
10565 libraries to link against. Failing to supply the correct flags may lead
10566 to subtle defects. Supplying them in cases where they are not necessary
10569 @item -shared-libgcc
10570 @itemx -static-libgcc
10571 @opindex shared-libgcc
10572 @opindex static-libgcc
10573 On systems that provide @file{libgcc} as a shared library, these options
10574 force the use of either the shared or static version, respectively.
10575 If no shared version of @file{libgcc} was built when the compiler was
10576 configured, these options have no effect.
10578 There are several situations in which an application should use the
10579 shared @file{libgcc} instead of the static version. The most common
10580 of these is when the application wishes to throw and catch exceptions
10581 across different shared libraries. In that case, each of the libraries
10582 as well as the application itself should use the shared @file{libgcc}.
10584 Therefore, the G++ and GCJ drivers automatically add
10585 @option{-shared-libgcc} whenever you build a shared library or a main
10586 executable, because C++ and Java programs typically use exceptions, so
10587 this is the right thing to do.
10589 If, instead, you use the GCC driver to create shared libraries, you may
10590 find that they are not always linked with the shared @file{libgcc}.
10591 If GCC finds, at its configuration time, that you have a non-GNU linker
10592 or a GNU linker that does not support option @option{--eh-frame-hdr},
10593 it links the shared version of @file{libgcc} into shared libraries
10594 by default. Otherwise, it takes advantage of the linker and optimizes
10595 away the linking with the shared version of @file{libgcc}, linking with
10596 the static version of libgcc by default. This allows exceptions to
10597 propagate through such shared libraries, without incurring relocation
10598 costs at library load time.
10600 However, if a library or main executable is supposed to throw or catch
10601 exceptions, you must link it using the G++ or GCJ driver, as appropriate
10602 for the languages used in the program, or using the option
10603 @option{-shared-libgcc}, such that it is linked with the shared
10606 @item -static-libasan
10607 @opindex static-libasan
10608 When the @option{-fsanitize=address} option is used to link a program,
10609 the GCC driver automatically links against @option{libasan}. If
10610 @file{libasan} is available as a shared library, and the @option{-static}
10611 option is not used, then this links against the shared version of
10612 @file{libasan}. The @option{-static-libasan} option directs the GCC
10613 driver to link @file{libasan} statically, without necessarily linking
10614 other libraries statically.
10616 @item -static-libtsan
10617 @opindex static-libtsan
10618 When the @option{-fsanitize=thread} option is used to link a program,
10619 the GCC driver automatically links against @option{libtsan}. If
10620 @file{libtsan} is available as a shared library, and the @option{-static}
10621 option is not used, then this links against the shared version of
10622 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
10623 driver to link @file{libtsan} statically, without necessarily linking
10624 other libraries statically.
10626 @item -static-liblsan
10627 @opindex static-liblsan
10628 When the @option{-fsanitize=leak} option is used to link a program,
10629 the GCC driver automatically links against @option{liblsan}. If
10630 @file{liblsan} is available as a shared library, and the @option{-static}
10631 option is not used, then this links against the shared version of
10632 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
10633 driver to link @file{liblsan} statically, without necessarily linking
10634 other libraries statically.
10636 @item -static-libubsan
10637 @opindex static-libubsan
10638 When the @option{-fsanitize=undefined} option is used to link a program,
10639 the GCC driver automatically links against @option{libubsan}. If
10640 @file{libubsan} is available as a shared library, and the @option{-static}
10641 option is not used, then this links against the shared version of
10642 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
10643 driver to link @file{libubsan} statically, without necessarily linking
10644 other libraries statically.
10646 @item -static-libstdc++
10647 @opindex static-libstdc++
10648 When the @command{g++} program is used to link a C++ program, it
10649 normally automatically links against @option{libstdc++}. If
10650 @file{libstdc++} is available as a shared library, and the
10651 @option{-static} option is not used, then this links against the
10652 shared version of @file{libstdc++}. That is normally fine. However, it
10653 is sometimes useful to freeze the version of @file{libstdc++} used by
10654 the program without going all the way to a fully static link. The
10655 @option{-static-libstdc++} option directs the @command{g++} driver to
10656 link @file{libstdc++} statically, without necessarily linking other
10657 libraries statically.
10661 Bind references to global symbols when building a shared object. Warn
10662 about any unresolved references (unless overridden by the link editor
10663 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
10666 @item -T @var{script}
10668 @cindex linker script
10669 Use @var{script} as the linker script. This option is supported by most
10670 systems using the GNU linker. On some targets, such as bare-board
10671 targets without an operating system, the @option{-T} option may be required
10672 when linking to avoid references to undefined symbols.
10674 @item -Xlinker @var{option}
10676 Pass @var{option} as an option to the linker. You can use this to
10677 supply system-specific linker options that GCC does not recognize.
10679 If you want to pass an option that takes a separate argument, you must use
10680 @option{-Xlinker} twice, once for the option and once for the argument.
10681 For example, to pass @option{-assert definitions}, you must write
10682 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
10683 @option{-Xlinker "-assert definitions"}, because this passes the entire
10684 string as a single argument, which is not what the linker expects.
10686 When using the GNU linker, it is usually more convenient to pass
10687 arguments to linker options using the @option{@var{option}=@var{value}}
10688 syntax than as separate arguments. For example, you can specify
10689 @option{-Xlinker -Map=output.map} rather than
10690 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
10691 this syntax for command-line options.
10693 @item -Wl,@var{option}
10695 Pass @var{option} as an option to the linker. If @var{option} contains
10696 commas, it is split into multiple options at the commas. You can use this
10697 syntax to pass an argument to the option.
10698 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
10699 linker. When using the GNU linker, you can also get the same effect with
10700 @option{-Wl,-Map=output.map}.
10702 @item -u @var{symbol}
10704 Pretend the symbol @var{symbol} is undefined, to force linking of
10705 library modules to define it. You can use @option{-u} multiple times with
10706 different symbols to force loading of additional library modules.
10708 @item -z @var{keyword}
10710 @option{-z} is passed directly on to the linker along with the keyword
10711 @var{keyword}. See the section in the documentation of your linker for
10712 permitted values and their meanings.
10715 @node Directory Options
10716 @section Options for Directory Search
10717 @cindex directory options
10718 @cindex options, directory search
10719 @cindex search path
10721 These options specify directories to search for header files, for
10722 libraries and for parts of the compiler:
10727 Add the directory @var{dir} to the head of the list of directories to be
10728 searched for header files. This can be used to override a system header
10729 file, substituting your own version, since these directories are
10730 searched before the system header file directories. However, you should
10731 not use this option to add directories that contain vendor-supplied
10732 system header files (use @option{-isystem} for that). If you use more than
10733 one @option{-I} option, the directories are scanned in left-to-right
10734 order; the standard system directories come after.
10736 If a standard system include directory, or a directory specified with
10737 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
10738 option is ignored. The directory is still searched but as a
10739 system directory at its normal position in the system include chain.
10740 This is to ensure that GCC's procedure to fix buggy system headers and
10741 the ordering for the @code{include_next} directive are not inadvertently changed.
10742 If you really need to change the search order for system directories,
10743 use the @option{-nostdinc} and/or @option{-isystem} options.
10745 @item -iplugindir=@var{dir}
10746 @opindex iplugindir=
10747 Set the directory to search for plugins that are passed
10748 by @option{-fplugin=@var{name}} instead of
10749 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
10750 to be used by the user, but only passed by the driver.
10752 @item -iquote@var{dir}
10754 Add the directory @var{dir} to the head of the list of directories to
10755 be searched for header files only for the case of @samp{#include
10756 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
10757 otherwise just like @option{-I}.
10761 Add directory @var{dir} to the list of directories to be searched
10764 @item -B@var{prefix}
10766 This option specifies where to find the executables, libraries,
10767 include files, and data files of the compiler itself.
10769 The compiler driver program runs one or more of the subprograms
10770 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
10771 @var{prefix} as a prefix for each program it tries to run, both with and
10772 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
10774 For each subprogram to be run, the compiler driver first tries the
10775 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
10776 is not specified, the driver tries two standard prefixes,
10777 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
10778 those results in a file name that is found, the unmodified program
10779 name is searched for using the directories specified in your
10780 @env{PATH} environment variable.
10782 The compiler checks to see if the path provided by the @option{-B}
10783 refers to a directory, and if necessary it adds a directory
10784 separator character at the end of the path.
10786 @option{-B} prefixes that effectively specify directory names also apply
10787 to libraries in the linker, because the compiler translates these
10788 options into @option{-L} options for the linker. They also apply to
10789 include files in the preprocessor, because the compiler translates these
10790 options into @option{-isystem} options for the preprocessor. In this case,
10791 the compiler appends @samp{include} to the prefix.
10793 The runtime support file @file{libgcc.a} can also be searched for using
10794 the @option{-B} prefix, if needed. If it is not found there, the two
10795 standard prefixes above are tried, and that is all. The file is left
10796 out of the link if it is not found by those means.
10798 Another way to specify a prefix much like the @option{-B} prefix is to use
10799 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
10802 As a special kludge, if the path provided by @option{-B} is
10803 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
10804 9, then it is replaced by @file{[dir/]include}. This is to help
10805 with boot-strapping the compiler.
10807 @item -specs=@var{file}
10809 Process @var{file} after the compiler reads in the standard @file{specs}
10810 file, in order to override the defaults which the @command{gcc} driver
10811 program uses when determining what switches to pass to @command{cc1},
10812 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
10813 @option{-specs=@var{file}} can be specified on the command line, and they
10814 are processed in order, from left to right.
10816 @item --sysroot=@var{dir}
10818 Use @var{dir} as the logical root directory for headers and libraries.
10819 For example, if the compiler normally searches for headers in
10820 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
10821 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
10823 If you use both this option and the @option{-isysroot} option, then
10824 the @option{--sysroot} option applies to libraries, but the
10825 @option{-isysroot} option applies to header files.
10827 The GNU linker (beginning with version 2.16) has the necessary support
10828 for this option. If your linker does not support this option, the
10829 header file aspect of @option{--sysroot} still works, but the
10830 library aspect does not.
10832 @item --no-sysroot-suffix
10833 @opindex no-sysroot-suffix
10834 For some targets, a suffix is added to the root directory specified
10835 with @option{--sysroot}, depending on the other options used, so that
10836 headers may for example be found in
10837 @file{@var{dir}/@var{suffix}/usr/include} instead of
10838 @file{@var{dir}/usr/include}. This option disables the addition of
10843 This option has been deprecated. Please use @option{-iquote} instead for
10844 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
10845 Any directories you specify with @option{-I} options before the @option{-I-}
10846 option are searched only for the case of @samp{#include "@var{file}"};
10847 they are not searched for @samp{#include <@var{file}>}.
10849 If additional directories are specified with @option{-I} options after
10850 the @option{-I-}, these directories are searched for all @samp{#include}
10851 directives. (Ordinarily @emph{all} @option{-I} directories are used
10854 In addition, the @option{-I-} option inhibits the use of the current
10855 directory (where the current input file came from) as the first search
10856 directory for @samp{#include "@var{file}"}. There is no way to
10857 override this effect of @option{-I-}. With @option{-I.} you can specify
10858 searching the directory that is current when the compiler is
10859 invoked. That is not exactly the same as what the preprocessor does
10860 by default, but it is often satisfactory.
10862 @option{-I-} does not inhibit the use of the standard system directories
10863 for header files. Thus, @option{-I-} and @option{-nostdinc} are
10870 @section Specifying subprocesses and the switches to pass to them
10873 @command{gcc} is a driver program. It performs its job by invoking a
10874 sequence of other programs to do the work of compiling, assembling and
10875 linking. GCC interprets its command-line parameters and uses these to
10876 deduce which programs it should invoke, and which command-line options
10877 it ought to place on their command lines. This behavior is controlled
10878 by @dfn{spec strings}. In most cases there is one spec string for each
10879 program that GCC can invoke, but a few programs have multiple spec
10880 strings to control their behavior. The spec strings built into GCC can
10881 be overridden by using the @option{-specs=} command-line switch to specify
10884 @dfn{Spec files} are plaintext files that are used to construct spec
10885 strings. They consist of a sequence of directives separated by blank
10886 lines. The type of directive is determined by the first non-whitespace
10887 character on the line, which can be one of the following:
10890 @item %@var{command}
10891 Issues a @var{command} to the spec file processor. The commands that can
10895 @item %include <@var{file}>
10896 @cindex @code{%include}
10897 Search for @var{file} and insert its text at the current point in the
10900 @item %include_noerr <@var{file}>
10901 @cindex @code{%include_noerr}
10902 Just like @samp{%include}, but do not generate an error message if the include
10903 file cannot be found.
10905 @item %rename @var{old_name} @var{new_name}
10906 @cindex @code{%rename}
10907 Rename the spec string @var{old_name} to @var{new_name}.
10911 @item *[@var{spec_name}]:
10912 This tells the compiler to create, override or delete the named spec
10913 string. All lines after this directive up to the next directive or
10914 blank line are considered to be the text for the spec string. If this
10915 results in an empty string then the spec is deleted. (Or, if the
10916 spec did not exist, then nothing happens.) Otherwise, if the spec
10917 does not currently exist a new spec is created. If the spec does
10918 exist then its contents are overridden by the text of this
10919 directive, unless the first character of that text is the @samp{+}
10920 character, in which case the text is appended to the spec.
10922 @item [@var{suffix}]:
10923 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
10924 and up to the next directive or blank line are considered to make up the
10925 spec string for the indicated suffix. When the compiler encounters an
10926 input file with the named suffix, it processes the spec string in
10927 order to work out how to compile that file. For example:
10931 z-compile -input %i
10934 This says that any input file whose name ends in @samp{.ZZ} should be
10935 passed to the program @samp{z-compile}, which should be invoked with the
10936 command-line switch @option{-input} and with the result of performing the
10937 @samp{%i} substitution. (See below.)
10939 As an alternative to providing a spec string, the text following a
10940 suffix directive can be one of the following:
10943 @item @@@var{language}
10944 This says that the suffix is an alias for a known @var{language}. This is
10945 similar to using the @option{-x} command-line switch to GCC to specify a
10946 language explicitly. For example:
10953 Says that .ZZ files are, in fact, C++ source files.
10956 This causes an error messages saying:
10959 @var{name} compiler not installed on this system.
10963 GCC already has an extensive list of suffixes built into it.
10964 This directive adds an entry to the end of the list of suffixes, but
10965 since the list is searched from the end backwards, it is effectively
10966 possible to override earlier entries using this technique.
10970 GCC has the following spec strings built into it. Spec files can
10971 override these strings or create their own. Note that individual
10972 targets can also add their own spec strings to this list.
10975 asm Options to pass to the assembler
10976 asm_final Options to pass to the assembler post-processor
10977 cpp Options to pass to the C preprocessor
10978 cc1 Options to pass to the C compiler
10979 cc1plus Options to pass to the C++ compiler
10980 endfile Object files to include at the end of the link
10981 link Options to pass to the linker
10982 lib Libraries to include on the command line to the linker
10983 libgcc Decides which GCC support library to pass to the linker
10984 linker Sets the name of the linker
10985 predefines Defines to be passed to the C preprocessor
10986 signed_char Defines to pass to CPP to say whether @code{char} is signed
10988 startfile Object files to include at the start of the link
10991 Here is a small example of a spec file:
10994 %rename lib old_lib
10997 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11000 This example renames the spec called @samp{lib} to @samp{old_lib} and
11001 then overrides the previous definition of @samp{lib} with a new one.
11002 The new definition adds in some extra command-line options before
11003 including the text of the old definition.
11005 @dfn{Spec strings} are a list of command-line options to be passed to their
11006 corresponding program. In addition, the spec strings can contain
11007 @samp{%}-prefixed sequences to substitute variable text or to
11008 conditionally insert text into the command line. Using these constructs
11009 it is possible to generate quite complex command lines.
11011 Here is a table of all defined @samp{%}-sequences for spec
11012 strings. Note that spaces are not generated automatically around the
11013 results of expanding these sequences. Therefore you can concatenate them
11014 together or combine them with constant text in a single argument.
11018 Substitute one @samp{%} into the program name or argument.
11021 Substitute the name of the input file being processed.
11024 Substitute the basename of the input file being processed.
11025 This is the substring up to (and not including) the last period
11026 and not including the directory.
11029 This is the same as @samp{%b}, but include the file suffix (text after
11033 Marks the argument containing or following the @samp{%d} as a
11034 temporary file name, so that that file is deleted if GCC exits
11035 successfully. Unlike @samp{%g}, this contributes no text to the
11038 @item %g@var{suffix}
11039 Substitute a file name that has suffix @var{suffix} and is chosen
11040 once per compilation, and mark the argument in the same way as
11041 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11042 name is now chosen in a way that is hard to predict even when previously
11043 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11044 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11045 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11046 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11047 was simply substituted with a file name chosen once per compilation,
11048 without regard to any appended suffix (which was therefore treated
11049 just like ordinary text), making such attacks more likely to succeed.
11051 @item %u@var{suffix}
11052 Like @samp{%g}, but generates a new temporary file name
11053 each time it appears instead of once per compilation.
11055 @item %U@var{suffix}
11056 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11057 new one if there is no such last file name. In the absence of any
11058 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11059 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11060 involves the generation of two distinct file names, one
11061 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11062 simply substituted with a file name chosen for the previous @samp{%u},
11063 without regard to any appended suffix.
11065 @item %j@var{suffix}
11066 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11067 writable, and if @option{-save-temps} is not used;
11068 otherwise, substitute the name
11069 of a temporary file, just like @samp{%u}. This temporary file is not
11070 meant for communication between processes, but rather as a junk
11071 disposal mechanism.
11073 @item %|@var{suffix}
11074 @itemx %m@var{suffix}
11075 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11076 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11077 all. These are the two most common ways to instruct a program that it
11078 should read from standard input or write to standard output. If you
11079 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11080 construct: see for example @file{f/lang-specs.h}.
11082 @item %.@var{SUFFIX}
11083 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11084 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11085 terminated by the next space or %.
11088 Marks the argument containing or following the @samp{%w} as the
11089 designated output file of this compilation. This puts the argument
11090 into the sequence of arguments that @samp{%o} substitutes.
11093 Substitutes the names of all the output files, with spaces
11094 automatically placed around them. You should write spaces
11095 around the @samp{%o} as well or the results are undefined.
11096 @samp{%o} is for use in the specs for running the linker.
11097 Input files whose names have no recognized suffix are not compiled
11098 at all, but they are included among the output files, so they are
11102 Substitutes the suffix for object files. Note that this is
11103 handled specially when it immediately follows @samp{%g, %u, or %U},
11104 because of the need for those to form complete file names. The
11105 handling is such that @samp{%O} is treated exactly as if it had already
11106 been substituted, except that @samp{%g, %u, and %U} do not currently
11107 support additional @var{suffix} characters following @samp{%O} as they do
11108 following, for example, @samp{.o}.
11111 Substitutes the standard macro predefinitions for the
11112 current target machine. Use this when running @code{cpp}.
11115 Like @samp{%p}, but puts @samp{__} before and after the name of each
11116 predefined macro, except for macros that start with @samp{__} or with
11117 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11121 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11122 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11123 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11124 and @option{-imultilib} as necessary.
11127 Current argument is the name of a library or startup file of some sort.
11128 Search for that file in a standard list of directories and substitute
11129 the full name found. The current working directory is included in the
11130 list of directories scanned.
11133 Current argument is the name of a linker script. Search for that file
11134 in the current list of directories to scan for libraries. If the file
11135 is located insert a @option{--script} option into the command line
11136 followed by the full path name found. If the file is not found then
11137 generate an error message. Note: the current working directory is not
11141 Print @var{str} as an error message. @var{str} is terminated by a newline.
11142 Use this when inconsistent options are detected.
11144 @item %(@var{name})
11145 Substitute the contents of spec string @var{name} at this point.
11147 @item %x@{@var{option}@}
11148 Accumulate an option for @samp{%X}.
11151 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11155 Output the accumulated assembler options specified by @option{-Wa}.
11158 Output the accumulated preprocessor options specified by @option{-Wp}.
11161 Process the @code{asm} spec. This is used to compute the
11162 switches to be passed to the assembler.
11165 Process the @code{asm_final} spec. This is a spec string for
11166 passing switches to an assembler post-processor, if such a program is
11170 Process the @code{link} spec. This is the spec for computing the
11171 command line passed to the linker. Typically it makes use of the
11172 @samp{%L %G %S %D and %E} sequences.
11175 Dump out a @option{-L} option for each directory that GCC believes might
11176 contain startup files. If the target supports multilibs then the
11177 current multilib directory is prepended to each of these paths.
11180 Process the @code{lib} spec. This is a spec string for deciding which
11181 libraries are included on the command line to the linker.
11184 Process the @code{libgcc} spec. This is a spec string for deciding
11185 which GCC support library is included on the command line to the linker.
11188 Process the @code{startfile} spec. This is a spec for deciding which
11189 object files are the first ones passed to the linker. Typically
11190 this might be a file named @file{crt0.o}.
11193 Process the @code{endfile} spec. This is a spec string that specifies
11194 the last object files that are passed to the linker.
11197 Process the @code{cpp} spec. This is used to construct the arguments
11198 to be passed to the C preprocessor.
11201 Process the @code{cc1} spec. This is used to construct the options to be
11202 passed to the actual C compiler (@samp{cc1}).
11205 Process the @code{cc1plus} spec. This is used to construct the options to be
11206 passed to the actual C++ compiler (@samp{cc1plus}).
11209 Substitute the variable part of a matched option. See below.
11210 Note that each comma in the substituted string is replaced by
11214 Remove all occurrences of @code{-S} from the command line. Note---this
11215 command is position dependent. @samp{%} commands in the spec string
11216 before this one see @code{-S}, @samp{%} commands in the spec string
11217 after this one do not.
11219 @item %:@var{function}(@var{args})
11220 Call the named function @var{function}, passing it @var{args}.
11221 @var{args} is first processed as a nested spec string, then split
11222 into an argument vector in the usual fashion. The function returns
11223 a string which is processed as if it had appeared literally as part
11224 of the current spec.
11226 The following built-in spec functions are provided:
11229 @item @code{getenv}
11230 The @code{getenv} spec function takes two arguments: an environment
11231 variable name and a string. If the environment variable is not
11232 defined, a fatal error is issued. Otherwise, the return value is the
11233 value of the environment variable concatenated with the string. For
11234 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11237 %:getenv(TOPDIR /include)
11240 expands to @file{/path/to/top/include}.
11242 @item @code{if-exists}
11243 The @code{if-exists} spec function takes one argument, an absolute
11244 pathname to a file. If the file exists, @code{if-exists} returns the
11245 pathname. Here is a small example of its usage:
11249 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11252 @item @code{if-exists-else}
11253 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11254 spec function, except that it takes two arguments. The first argument is
11255 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11256 returns the pathname. If it does not exist, it returns the second argument.
11257 This way, @code{if-exists-else} can be used to select one file or another,
11258 based on the existence of the first. Here is a small example of its usage:
11262 crt0%O%s %:if-exists(crti%O%s) \
11263 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11266 @item @code{replace-outfile}
11267 The @code{replace-outfile} spec function takes two arguments. It looks for the
11268 first argument in the outfiles array and replaces it with the second argument. Here
11269 is a small example of its usage:
11272 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11275 @item @code{remove-outfile}
11276 The @code{remove-outfile} spec function takes one argument. It looks for the
11277 first argument in the outfiles array and removes it. Here is a small example
11281 %:remove-outfile(-lm)
11284 @item @code{pass-through-libs}
11285 The @code{pass-through-libs} spec function takes any number of arguments. It
11286 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11287 assumes are the names of linker input library archive files) and returns a
11288 result containing all the found arguments each prepended by
11289 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11290 intended to be passed to the LTO linker plugin.
11293 %:pass-through-libs(%G %L %G)
11296 @item @code{print-asm-header}
11297 The @code{print-asm-header} function takes no arguments and simply
11298 prints a banner like:
11304 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11307 It is used to separate compiler options from assembler options
11308 in the @option{--target-help} output.
11311 @item %@{@code{S}@}
11312 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11313 If that switch is not specified, this substitutes nothing. Note that
11314 the leading dash is omitted when specifying this option, and it is
11315 automatically inserted if the substitution is performed. Thus the spec
11316 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11317 and outputs the command-line option @option{-foo}.
11319 @item %W@{@code{S}@}
11320 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11321 deleted on failure.
11323 @item %@{@code{S}*@}
11324 Substitutes all the switches specified to GCC whose names start
11325 with @code{-S}, but which also take an argument. This is used for
11326 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11327 GCC considers @option{-o foo} as being
11328 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11329 text, including the space. Thus two arguments are generated.
11331 @item %@{@code{S}*&@code{T}*@}
11332 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11333 (the order of @code{S} and @code{T} in the spec is not significant).
11334 There can be any number of ampersand-separated variables; for each the
11335 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11337 @item %@{@code{S}:@code{X}@}
11338 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11340 @item %@{!@code{S}:@code{X}@}
11341 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11343 @item %@{@code{S}*:@code{X}@}
11344 Substitutes @code{X} if one or more switches whose names start with
11345 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11346 once, no matter how many such switches appeared. However, if @code{%*}
11347 appears somewhere in @code{X}, then @code{X} is substituted once
11348 for each matching switch, with the @code{%*} replaced by the part of
11349 that switch matching the @code{*}.
11351 If @code{%*} appears as the last part of a spec sequence then a space
11352 will be added after the end of the last substitution. If there is more
11353 text in the sequence however then a space will not be generated. This
11354 allows the @code{%*} substitution to be used as part of a larger
11355 string. For example, a spec string like this:
11358 %@{mcu=*:--script=%*/memory.ld@}
11361 when matching an option like @code{-mcu=newchip} will produce:
11364 --script=newchip/memory.ld
11367 @item %@{.@code{S}:@code{X}@}
11368 Substitutes @code{X}, if processing a file with suffix @code{S}.
11370 @item %@{!.@code{S}:@code{X}@}
11371 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
11373 @item %@{,@code{S}:@code{X}@}
11374 Substitutes @code{X}, if processing a file for language @code{S}.
11376 @item %@{!,@code{S}:@code{X}@}
11377 Substitutes @code{X}, if not processing a file for language @code{S}.
11379 @item %@{@code{S}|@code{P}:@code{X}@}
11380 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
11381 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
11382 @code{*} sequences as well, although they have a stronger binding than
11383 the @samp{|}. If @code{%*} appears in @code{X}, all of the
11384 alternatives must be starred, and only the first matching alternative
11387 For example, a spec string like this:
11390 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
11394 outputs the following command-line options from the following input
11395 command-line options:
11400 -d fred.c -foo -baz -boggle
11401 -d jim.d -bar -baz -boggle
11404 @item %@{S:X; T:Y; :D@}
11406 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
11407 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
11408 be as many clauses as you need. This may be combined with @code{.},
11409 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
11414 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
11415 construct may contain other nested @samp{%} constructs or spaces, or
11416 even newlines. They are processed as usual, as described above.
11417 Trailing white space in @code{X} is ignored. White space may also
11418 appear anywhere on the left side of the colon in these constructs,
11419 except between @code{.} or @code{*} and the corresponding word.
11421 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
11422 handled specifically in these constructs. If another value of
11423 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
11424 @option{-W} switch is found later in the command line, the earlier
11425 switch value is ignored, except with @{@code{S}*@} where @code{S} is
11426 just one letter, which passes all matching options.
11428 The character @samp{|} at the beginning of the predicate text is used to
11429 indicate that a command should be piped to the following command, but
11430 only if @option{-pipe} is specified.
11432 It is built into GCC which switches take arguments and which do not.
11433 (You might think it would be useful to generalize this to allow each
11434 compiler's spec to say which switches take arguments. But this cannot
11435 be done in a consistent fashion. GCC cannot even decide which input
11436 files have been specified without knowing which switches take arguments,
11437 and it must know which input files to compile in order to tell which
11440 GCC also knows implicitly that arguments starting in @option{-l} are to be
11441 treated as compiler output files, and passed to the linker in their
11442 proper position among the other output files.
11444 @c man begin OPTIONS
11446 @node Target Options
11447 @section Specifying Target Machine and Compiler Version
11448 @cindex target options
11449 @cindex cross compiling
11450 @cindex specifying machine version
11451 @cindex specifying compiler version and target machine
11452 @cindex compiler version, specifying
11453 @cindex target machine, specifying
11455 The usual way to run GCC is to run the executable called @command{gcc}, or
11456 @command{@var{machine}-gcc} when cross-compiling, or
11457 @command{@var{machine}-gcc-@var{version}} to run a version other than the
11458 one that was installed last.
11460 @node Submodel Options
11461 @section Hardware Models and Configurations
11462 @cindex submodel options
11463 @cindex specifying hardware config
11464 @cindex hardware models and configurations, specifying
11465 @cindex machine dependent options
11467 Each target machine types can have its own
11468 special options, starting with @samp{-m}, to choose among various
11469 hardware models or configurations---for example, 68010 vs 68020,
11470 floating coprocessor or none. A single installed version of the
11471 compiler can compile for any model or configuration, according to the
11474 Some configurations of the compiler also support additional special
11475 options, usually for compatibility with other compilers on the same
11478 @c This list is ordered alphanumerically by subsection name.
11479 @c It should be the same order and spelling as these options are listed
11480 @c in Machine Dependent Options
11483 * AArch64 Options::
11484 * Adapteva Epiphany Options::
11488 * Blackfin Options::
11493 * DEC Alpha Options::
11496 * GNU/Linux Options::
11499 * i386 and x86-64 Options::
11500 * i386 and x86-64 Windows Options::
11508 * MicroBlaze Options::
11511 * MN10300 Options::
11515 * Nios II Options::
11517 * picoChip Options::
11518 * PowerPC Options::
11520 * RS/6000 and PowerPC Options::
11522 * S/390 and zSeries Options::
11525 * Solaris 2 Options::
11528 * System V Options::
11529 * TILE-Gx Options::
11530 * TILEPro Options::
11534 * VxWorks Options::
11536 * Xstormy16 Options::
11538 * zSeries Options::
11541 @node AArch64 Options
11542 @subsection AArch64 Options
11543 @cindex AArch64 Options
11545 These options are defined for AArch64 implementations:
11549 @item -mabi=@var{name}
11551 Generate code for the specified data model. Permissible values
11552 are @samp{ilp32} for SysV-like data model where int, long int and pointer
11553 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
11554 but long int and pointer are 64-bit.
11556 The default depends on the specific target configuration. Note that
11557 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
11558 entire program with the same ABI, and link with a compatible set of libraries.
11561 @opindex mbig-endian
11562 Generate big-endian code. This is the default when GCC is configured for an
11563 @samp{aarch64_be-*-*} target.
11565 @item -mgeneral-regs-only
11566 @opindex mgeneral-regs-only
11567 Generate code which uses only the general registers.
11569 @item -mlittle-endian
11570 @opindex mlittle-endian
11571 Generate little-endian code. This is the default when GCC is configured for an
11572 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
11574 @item -mcmodel=tiny
11575 @opindex mcmodel=tiny
11576 Generate code for the tiny code model. The program and its statically defined
11577 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
11578 be statically or dynamically linked. This model is not fully implemented and
11579 mostly treated as @samp{small}.
11581 @item -mcmodel=small
11582 @opindex mcmodel=small
11583 Generate code for the small code model. The program and its statically defined
11584 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
11585 be statically or dynamically linked. This is the default code model.
11587 @item -mcmodel=large
11588 @opindex mcmodel=large
11589 Generate code for the large code model. This makes no assumptions about
11590 addresses and sizes of sections. Pointers are 64 bits. Programs can be
11591 statically linked only.
11593 @item -mstrict-align
11594 @opindex mstrict-align
11595 Do not assume that unaligned memory references will be handled by the system.
11597 @item -momit-leaf-frame-pointer
11598 @itemx -mno-omit-leaf-frame-pointer
11599 @opindex momit-leaf-frame-pointer
11600 @opindex mno-omit-leaf-frame-pointer
11601 Omit or keep the frame pointer in leaf functions. The former behaviour is the
11604 @item -mtls-dialect=desc
11605 @opindex mtls-dialect=desc
11606 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
11607 of TLS variables. This is the default.
11609 @item -mtls-dialect=traditional
11610 @opindex mtls-dialect=traditional
11611 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
11614 @item -march=@var{name}
11616 Specify the name of the target architecture, optionally suffixed by one or
11617 more feature modifiers. This option has the form
11618 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11619 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
11620 values for @var{feature} are documented in the sub-section below.
11622 Where conflicting feature modifiers are specified, the right-most feature is
11625 GCC uses this name to determine what kind of instructions it can emit when
11626 generating assembly code.
11628 Where @option{-march} is specified without either of @option{-mtune}
11629 or @option{-mcpu} also being specified, the code will be tuned to perform
11630 well across a range of target processors implementing the target
11633 @item -mtune=@var{name}
11635 Specify the name of the target processor for which GCC should tune the
11636 performance of the code. Permissible values for this option are:
11637 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}.
11639 Additionally, this option can specify that GCC should tune the performance
11640 of the code for a big.LITTLE system. The only permissible value is
11641 @samp{cortex-a57.cortex-a53}.
11643 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
11644 are specified, the code will be tuned to perform well across a range
11645 of target processors.
11647 This option cannot be suffixed by feature modifiers.
11649 @item -mcpu=@var{name}
11651 Specify the name of the target processor, optionally suffixed by one or more
11652 feature modifiers. This option has the form
11653 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
11654 permissible values for @var{cpu} are the same as those available for
11657 The permissible values for @var{feature} are documented in the sub-section
11660 Where conflicting feature modifiers are specified, the right-most feature is
11663 GCC uses this name to determine what kind of instructions it can emit when
11664 generating assembly code (as if by @option{-march}) and to determine
11665 the target processor for which to tune for performance (as if
11666 by @option{-mtune}). Where this option is used in conjunction
11667 with @option{-march} or @option{-mtune}, those options take precedence
11668 over the appropriate part of this option.
11671 @subsubsection @option{-march} and @option{-mcpu} feature modifiers
11672 @cindex @option{-march} feature modifiers
11673 @cindex @option{-mcpu} feature modifiers
11674 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
11679 Enable CRC extension.
11681 Enable Crypto extension. This implies Advanced SIMD is enabled.
11683 Enable floating-point instructions.
11685 Enable Advanced SIMD instructions. This implies floating-point instructions
11686 are enabled. This is the default for all current possible values for options
11687 @option{-march} and @option{-mcpu=}.
11690 @node Adapteva Epiphany Options
11691 @subsection Adapteva Epiphany Options
11693 These @samp{-m} options are defined for Adapteva Epiphany:
11696 @item -mhalf-reg-file
11697 @opindex mhalf-reg-file
11698 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
11699 That allows code to run on hardware variants that lack these registers.
11701 @item -mprefer-short-insn-regs
11702 @opindex mprefer-short-insn-regs
11703 Preferrentially allocate registers that allow short instruction generation.
11704 This can result in increased instruction count, so this may either reduce or
11705 increase overall code size.
11707 @item -mbranch-cost=@var{num}
11708 @opindex mbranch-cost
11709 Set the cost of branches to roughly @var{num} ``simple'' instructions.
11710 This cost is only a heuristic and is not guaranteed to produce
11711 consistent results across releases.
11715 Enable the generation of conditional moves.
11717 @item -mnops=@var{num}
11719 Emit @var{num} NOPs before every other generated instruction.
11721 @item -mno-soft-cmpsf
11722 @opindex mno-soft-cmpsf
11723 For single-precision floating-point comparisons, emit an @code{fsub} instruction
11724 and test the flags. This is faster than a software comparison, but can
11725 get incorrect results in the presence of NaNs, or when two different small
11726 numbers are compared such that their difference is calculated as zero.
11727 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
11728 software comparisons.
11730 @item -mstack-offset=@var{num}
11731 @opindex mstack-offset
11732 Set the offset between the top of the stack and the stack pointer.
11733 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
11734 can be used by leaf functions without stack allocation.
11735 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
11736 Note also that this option changes the ABI; compiling a program with a
11737 different stack offset than the libraries have been compiled with
11738 generally does not work.
11739 This option can be useful if you want to evaluate if a different stack
11740 offset would give you better code, but to actually use a different stack
11741 offset to build working programs, it is recommended to configure the
11742 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
11744 @item -mno-round-nearest
11745 @opindex mno-round-nearest
11746 Make the scheduler assume that the rounding mode has been set to
11747 truncating. The default is @option{-mround-nearest}.
11750 @opindex mlong-calls
11751 If not otherwise specified by an attribute, assume all calls might be beyond
11752 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
11753 function address into a register before performing a (otherwise direct) call.
11754 This is the default.
11756 @item -mshort-calls
11757 @opindex short-calls
11758 If not otherwise specified by an attribute, assume all direct calls are
11759 in the range of the @code{b} / @code{bl} instructions, so use these instructions
11760 for direct calls. The default is @option{-mlong-calls}.
11764 Assume addresses can be loaded as 16-bit unsigned values. This does not
11765 apply to function addresses for which @option{-mlong-calls} semantics
11768 @item -mfp-mode=@var{mode}
11770 Set the prevailing mode of the floating-point unit.
11771 This determines the floating-point mode that is provided and expected
11772 at function call and return time. Making this mode match the mode you
11773 predominantly need at function start can make your programs smaller and
11774 faster by avoiding unnecessary mode switches.
11776 @var{mode} can be set to one the following values:
11780 Any mode at function entry is valid, and retained or restored when
11781 the function returns, and when it calls other functions.
11782 This mode is useful for compiling libraries or other compilation units
11783 you might want to incorporate into different programs with different
11784 prevailing FPU modes, and the convenience of being able to use a single
11785 object file outweighs the size and speed overhead for any extra
11786 mode switching that might be needed, compared with what would be needed
11787 with a more specific choice of prevailing FPU mode.
11790 This is the mode used for floating-point calculations with
11791 truncating (i.e.@: round towards zero) rounding mode. That includes
11792 conversion from floating point to integer.
11794 @item round-nearest
11795 This is the mode used for floating-point calculations with
11796 round-to-nearest-or-even rounding mode.
11799 This is the mode used to perform integer calculations in the FPU, e.g.@:
11800 integer multiply, or integer multiply-and-accumulate.
11803 The default is @option{-mfp-mode=caller}
11805 @item -mnosplit-lohi
11806 @itemx -mno-postinc
11807 @itemx -mno-postmodify
11808 @opindex mnosplit-lohi
11809 @opindex mno-postinc
11810 @opindex mno-postmodify
11811 Code generation tweaks that disable, respectively, splitting of 32-bit
11812 loads, generation of post-increment addresses, and generation of
11813 post-modify addresses. The defaults are @option{msplit-lohi},
11814 @option{-mpost-inc}, and @option{-mpost-modify}.
11816 @item -mnovect-double
11817 @opindex mno-vect-double
11818 Change the preferred SIMD mode to SImode. The default is
11819 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
11821 @item -max-vect-align=@var{num}
11822 @opindex max-vect-align
11823 The maximum alignment for SIMD vector mode types.
11824 @var{num} may be 4 or 8. The default is 8.
11825 Note that this is an ABI change, even though many library function
11826 interfaces are unaffected if they don't use SIMD vector modes
11827 in places that affect size and/or alignment of relevant types.
11829 @item -msplit-vecmove-early
11830 @opindex msplit-vecmove-early
11831 Split vector moves into single word moves before reload. In theory this
11832 can give better register allocation, but so far the reverse seems to be
11833 generally the case.
11835 @item -m1reg-@var{reg}
11837 Specify a register to hold the constant @minus{}1, which makes loading small negative
11838 constants and certain bitmasks faster.
11839 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
11840 which specify use of that register as a fixed register,
11841 and @samp{none}, which means that no register is used for this
11842 purpose. The default is @option{-m1reg-none}.
11847 @subsection ARC Options
11848 @cindex ARC options
11850 The following options control the architecture variant for which code
11853 @c architecture variants
11856 @item -mbarrel-shifter
11857 @opindex mbarrel-shifter
11858 Generate instructions supported by barrel shifter. This is the default
11859 unless @samp{-mcpu=ARC601} is in effect.
11861 @item -mcpu=@var{cpu}
11863 Set architecture type, register usage, and instruction scheduling
11864 parameters for @var{cpu}. There are also shortcut alias options
11865 available for backward compatibility and convenience. Supported
11866 values for @var{cpu} are
11872 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
11876 Compile for ARC601. Alias: @option{-mARC601}.
11881 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
11882 This is the default when configured with @samp{--with-cpu=arc700}@.
11887 @itemx -mdpfp-compact
11888 @opindex mdpfp-compact
11889 FPX: Generate Double Precision FPX instructions, tuned for the compact
11893 @opindex mdpfp-fast
11894 FPX: Generate Double Precision FPX instructions, tuned for the fast
11897 @item -mno-dpfp-lrsr
11898 @opindex mno-dpfp-lrsr
11899 Disable LR and SR instructions from using FPX extension aux registers.
11903 Generate Extended arithmetic instructions. Currently only
11904 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
11905 supported. This is always enabled for @samp{-mcpu=ARC700}.
11909 Do not generate mpy instructions for ARC700.
11913 Generate 32x16 bit multiply and mac instructions.
11917 Generate mul64 and mulu64 instructions. Only valid for @samp{-mcpu=ARC600}.
11921 Generate norm instruction. This is the default if @samp{-mcpu=ARC700}
11926 @itemx -mspfp-compact
11927 @opindex mspfp-compact
11928 FPX: Generate Single Precision FPX instructions, tuned for the compact
11932 @opindex mspfp-fast
11933 FPX: Generate Single Precision FPX instructions, tuned for the fast
11938 Enable generation of ARC SIMD instructions via target-specific
11939 builtins. Only valid for @samp{-mcpu=ARC700}.
11942 @opindex msoft-float
11943 This option ignored; it is provided for compatibility purposes only.
11944 Software floating point code is emitted by default, and this default
11945 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
11946 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
11947 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
11951 Generate swap instructions.
11955 The following options are passed through to the assembler, and also
11956 define preprocessor macro symbols.
11958 @c Flags used by the assembler, but for which we define preprocessor
11959 @c macro symbols as well.
11962 @opindex mdsp-packa
11963 Passed down to the assembler to enable the DSP Pack A extensions.
11964 Also sets the preprocessor symbol @code{__Xdsp_packa}.
11968 Passed down to the assembler to enable the dual viterbi butterfly
11969 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
11971 @c ARC700 4.10 extension instruction
11974 Passed down to the assembler to enable the Locked Load/Store
11975 Conditional extension. Also sets the preprocessor symbol
11980 Passed down to the assembler. Also sets the preprocessor symbol
11981 @code{__Xxmac_d16}.
11985 Passed down to the assembler. Also sets the preprocessor symbol
11988 @c ARC700 4.10 extension instruction
11991 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
11992 extension instruction. Also sets the preprocessor symbol
11995 @c ARC700 4.10 extension instruction
11998 Passed down to the assembler to enable the swap byte ordering
11999 extension instruction. Also sets the preprocessor symbol
12003 @opindex mtelephony
12004 Passed down to the assembler to enable dual and single operand
12005 instructions for telephony. Also sets the preprocessor symbol
12006 @code{__Xtelephony}.
12010 Passed down to the assembler to enable the XY Memory extension. Also
12011 sets the preprocessor symbol @code{__Xxy}.
12015 The following options control how the assembly code is annotated:
12017 @c Assembly annotation options
12021 Annotate assembler instructions with estimated addresses.
12023 @item -mannotate-align
12024 @opindex mannotate-align
12025 Explain what alignment considerations lead to the decision to make an
12026 instruction short or long.
12030 The following options are passed through to the linker:
12032 @c options passed through to the linker
12036 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12037 This option is enabled by default in tool chains built for
12038 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12039 when profiling is not requested.
12041 @item -marclinux_prof
12042 @opindex marclinux_prof
12043 Passed through to the linker, to specify use of the
12044 @code{arclinux_prof} emulation. This option is enabled by default in
12045 tool chains built for @w{@code{arc-linux-uclibc}} and
12046 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12050 The following options control the semantics of generated code:
12052 @c semantically relevant code generation options
12054 @item -mepilogue-cfi
12055 @opindex mepilogue-cfi
12056 Enable generation of call frame information for epilogues.
12058 @item -mno-epilogue-cfi
12059 @opindex mno-epilogue-cfi
12060 Disable generation of call frame information for epilogues.
12063 @opindex mlong-calls
12064 Generate call insns as register indirect calls, thus providing access
12065 to the full 32-bit address range.
12067 @item -mmedium-calls
12068 @opindex mmedium-calls
12069 Don't use less than 25 bit addressing range for calls, which is the
12070 offset available for an unconditional branch-and-link
12071 instruction. Conditional execution of function calls is suppressed, to
12072 allow use of the 25-bit range, rather than the 21-bit range with
12073 conditional branch-and-link. This is the default for tool chains built
12074 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12078 Do not generate sdata references. This is the default for tool chains
12079 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12083 @opindex mucb-mcount
12084 Instrument with mcount calls as used in UCB code. I.e. do the
12085 counting in the callee, not the caller. By default ARC instrumentation
12086 counts in the caller.
12088 @item -mvolatile-cache
12089 @opindex mvolatile-cache
12090 Use ordinarily cached memory accesses for volatile references. This is the
12093 @item -mno-volatile-cache
12094 @opindex mno-volatile-cache
12095 Enable cache bypass for volatile references.
12099 The following options fine tune code generation:
12100 @c code generation tuning options
12103 @opindex malign-call
12104 Do alignment optimizations for call instructions.
12106 @item -mauto-modify-reg
12107 @opindex mauto-modify-reg
12108 Enable the use of pre/post modify with register displacement.
12110 @item -mbbit-peephole
12111 @opindex mbbit-peephole
12112 Enable bbit peephole2.
12116 This option disables a target-specific pass in @file{arc_reorg} to
12117 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12118 generation driven by the combiner pass.
12120 @item -mcase-vector-pcrel
12121 @opindex mcase-vector-pcrel
12122 Use pc-relative switch case tables - this enables case table shortening.
12123 This is the default for @option{-Os}.
12125 @item -mcompact-casesi
12126 @opindex mcompact-casesi
12127 Enable compact casesi pattern.
12128 This is the default for @option{-Os}.
12130 @item -mno-cond-exec
12131 @opindex mno-cond-exec
12132 Disable ARCompact specific pass to generate conditional execution instructions.
12133 Due to delay slot scheduling and interactions between operand numbers,
12134 literal sizes, instruction lengths, and the support for conditional execution,
12135 the target-independent pass to generate conditional execution is often lacking,
12136 so the ARC port has kept a special pass around that tries to find more
12137 conditional execution generating opportunities after register allocation,
12138 branch shortening, and delay slot scheduling have been done. This pass
12139 generally, but not always, improves performance and code size, at the cost of
12140 extra compilation time, which is why there is an option to switch it off.
12141 If you have a problem with call instructions exceeding their allowable
12142 offset range because they are conditionalized, you should consider using
12143 @option{-mmedium-calls} instead.
12145 @item -mearly-cbranchsi
12146 @opindex mearly-cbranchsi
12147 Enable pre-reload use of the cbranchsi pattern.
12149 @item -mexpand-adddi
12150 @opindex mexpand-adddi
12151 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12152 @code{add.f}, @code{adc} etc.
12154 @item -mindexed-loads
12155 @opindex mindexed-loads
12156 Enable the use of indexed loads. This can be problematic because some
12157 optimizers will then assume the that indexed stores exist, which is not
12162 Enable Local Register Allocation. This is still experimental for ARC,
12163 so by default the compiler uses standard reload
12164 (i.e. @samp{-mno-lra}).
12166 @item -mlra-priority-none
12167 @opindex mlra-priority-none
12168 Don't indicate any priority for target registers.
12170 @item -mlra-priority-compact
12171 @opindex mlra-priority-compact
12172 Indicate target register priority for r0..r3 / r12..r15.
12174 @item -mlra-priority-noncompact
12175 @opindex mlra-priority-noncompact
12176 Reduce target regsiter priority for r0..r3 / r12..r15.
12178 @item -mno-millicode
12179 @opindex mno-millicode
12180 When optimizing for size (using @option{-Os}), prologues and epilogues
12181 that have to save or restore a large number of registers are often
12182 shortened by using call to a special function in libgcc; this is
12183 referred to as a @emph{millicode} call. As these calls can pose
12184 performance issues, and/or cause linking issues when linking in a
12185 nonstandard way, this option is provided to turn off millicode call
12189 @opindex mmixed-code
12190 Tweak register allocation to help 16-bit instruction generation.
12191 This generally has the effect of decreasing the average instruction size
12192 while increasing the instruction count.
12196 Enable 'q' instruction alternatives.
12197 This is the default for @option{-Os}.
12201 Enable Rcq constraint handling - most short code generation depends on this.
12202 This is the default.
12206 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12207 This is the default.
12209 @item -msize-level=@var{level}
12210 @opindex msize-level
12211 Fine-tune size optimization with regards to instruction lengths and alignment.
12212 The recognized values for @var{level} are:
12215 No size optimization. This level is deprecated and treated like @samp{1}.
12218 Short instructions are used opportunistically.
12221 In addition, alignment of loops and of code after barriers are dropped.
12224 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12228 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12229 the behavior when this is not set is equivalent to level @samp{1}.
12231 @item -mtune=@var{cpu}
12233 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12234 by @option{-mcpu=}.
12236 Supported values for @var{cpu} are
12240 Tune for ARC600 cpu.
12243 Tune for ARC601 cpu.
12246 Tune for ARC700 cpu with standard multiplier block.
12249 Tune for ARC700 cpu with XMAC block.
12252 Tune for ARC725D cpu.
12255 Tune for ARC750D cpu.
12259 @item -mmultcost=@var{num}
12261 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12262 normal instruction.
12264 @item -munalign-prob-threshold=@var{probability}
12265 @opindex munalign-prob-threshold
12266 Set probability threshold for unaligning branches.
12267 When tuning for @samp{ARC700} and optimizing for speed, branches without
12268 filled delay slot are preferably emitted unaligned and long, unless
12269 profiling indicates that the probability for the branch to be taken
12270 is below @var{probability}. @xref{Cross-profiling}.
12271 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12275 The following options are maintained for backward compatibility, but
12276 are now deprecated and will be removed in a future release:
12278 @c Deprecated options
12286 @opindex mbig-endian
12289 Compile code for big endian targets. Use of these options is now
12290 deprecated. Users wanting big-endian code, should use the
12291 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12292 building the tool chain, for which big-endian is the default.
12294 @item -mlittle-endian
12295 @opindex mlittle-endian
12298 Compile code for little endian targets. Use of these options is now
12299 deprecated. Users wanting little-endian code should use the
12300 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12301 building the tool chain, for which little-endian is the default.
12303 @item -mbarrel_shifter
12304 @opindex mbarrel_shifter
12305 Replaced by @samp{-mbarrel-shifter}
12307 @item -mdpfp_compact
12308 @opindex mdpfp_compact
12309 Replaced by @samp{-mdpfp-compact}
12312 @opindex mdpfp_fast
12313 Replaced by @samp{-mdpfp-fast}
12316 @opindex mdsp_packa
12317 Replaced by @samp{-mdsp-packa}
12321 Replaced by @samp{-mea}
12325 Replaced by @samp{-mmac-24}
12329 Replaced by @samp{-mmac-d16}
12331 @item -mspfp_compact
12332 @opindex mspfp_compact
12333 Replaced by @samp{-mspfp-compact}
12336 @opindex mspfp_fast
12337 Replaced by @samp{-mspfp-fast}
12339 @item -mtune=@var{cpu}
12341 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
12342 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
12343 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
12345 @item -multcost=@var{num}
12347 Replaced by @samp{-mmultcost}.
12352 @subsection ARM Options
12353 @cindex ARM options
12355 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
12359 @item -mabi=@var{name}
12361 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
12362 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
12365 @opindex mapcs-frame
12366 Generate a stack frame that is compliant with the ARM Procedure Call
12367 Standard for all functions, even if this is not strictly necessary for
12368 correct execution of the code. Specifying @option{-fomit-frame-pointer}
12369 with this option causes the stack frames not to be generated for
12370 leaf functions. The default is @option{-mno-apcs-frame}.
12374 This is a synonym for @option{-mapcs-frame}.
12377 @c not currently implemented
12378 @item -mapcs-stack-check
12379 @opindex mapcs-stack-check
12380 Generate code to check the amount of stack space available upon entry to
12381 every function (that actually uses some stack space). If there is
12382 insufficient space available then either the function
12383 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} is
12384 called, depending upon the amount of stack space required. The runtime
12385 system is required to provide these functions. The default is
12386 @option{-mno-apcs-stack-check}, since this produces smaller code.
12388 @c not currently implemented
12390 @opindex mapcs-float
12391 Pass floating-point arguments using the floating-point registers. This is
12392 one of the variants of the APCS@. This option is recommended if the
12393 target hardware has a floating-point unit or if a lot of floating-point
12394 arithmetic is going to be performed by the code. The default is
12395 @option{-mno-apcs-float}, since the size of integer-only code is
12396 slightly increased if @option{-mapcs-float} is used.
12398 @c not currently implemented
12399 @item -mapcs-reentrant
12400 @opindex mapcs-reentrant
12401 Generate reentrant, position-independent code. The default is
12402 @option{-mno-apcs-reentrant}.
12405 @item -mthumb-interwork
12406 @opindex mthumb-interwork
12407 Generate code that supports calling between the ARM and Thumb
12408 instruction sets. Without this option, on pre-v5 architectures, the
12409 two instruction sets cannot be reliably used inside one program. The
12410 default is @option{-mno-thumb-interwork}, since slightly larger code
12411 is generated when @option{-mthumb-interwork} is specified. In AAPCS
12412 configurations this option is meaningless.
12414 @item -mno-sched-prolog
12415 @opindex mno-sched-prolog
12416 Prevent the reordering of instructions in the function prologue, or the
12417 merging of those instruction with the instructions in the function's
12418 body. This means that all functions start with a recognizable set
12419 of instructions (or in fact one of a choice from a small set of
12420 different function prologues), and this information can be used to
12421 locate the start of functions inside an executable piece of code. The
12422 default is @option{-msched-prolog}.
12424 @item -mfloat-abi=@var{name}
12425 @opindex mfloat-abi
12426 Specifies which floating-point ABI to use. Permissible values
12427 are: @samp{soft}, @samp{softfp} and @samp{hard}.
12429 Specifying @samp{soft} causes GCC to generate output containing
12430 library calls for floating-point operations.
12431 @samp{softfp} allows the generation of code using hardware floating-point
12432 instructions, but still uses the soft-float calling conventions.
12433 @samp{hard} allows generation of floating-point instructions
12434 and uses FPU-specific calling conventions.
12436 The default depends on the specific target configuration. Note that
12437 the hard-float and soft-float ABIs are not link-compatible; you must
12438 compile your entire program with the same ABI, and link with a
12439 compatible set of libraries.
12441 @item -mlittle-endian
12442 @opindex mlittle-endian
12443 Generate code for a processor running in little-endian mode. This is
12444 the default for all standard configurations.
12447 @opindex mbig-endian
12448 Generate code for a processor running in big-endian mode; the default is
12449 to compile code for a little-endian processor.
12451 @item -march=@var{name}
12453 This specifies the name of the target ARM architecture. GCC uses this
12454 name to determine what kind of instructions it can emit when generating
12455 assembly code. This option can be used in conjunction with or instead
12456 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
12457 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
12458 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
12459 @samp{armv6}, @samp{armv6j},
12460 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
12461 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
12462 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
12463 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
12465 @option{-march=armv7ve} is the armv7-a architecture with virtualization
12468 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
12469 architecture together with the optional CRC32 extensions.
12471 @option{-march=native} causes the compiler to auto-detect the architecture
12472 of the build computer. At present, this feature is only supported on
12473 Linux, and not all architectures are recognized. If the auto-detect is
12474 unsuccessful the option has no effect.
12476 @item -mtune=@var{name}
12478 This option specifies the name of the target ARM processor for
12479 which GCC should tune the performance of the code.
12480 For some ARM implementations better performance can be obtained by using
12482 Permissible names are: @samp{arm2}, @samp{arm250},
12483 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
12484 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
12485 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
12486 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
12488 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
12489 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
12490 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
12491 @samp{strongarm1110},
12492 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
12493 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
12494 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
12495 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
12496 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
12497 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
12498 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
12499 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
12500 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53}, @samp{cortex-a57},
12502 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m4},
12506 @samp{cortex-m0plus},
12507 @samp{marvell-pj4},
12508 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
12509 @samp{fa526}, @samp{fa626},
12510 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
12512 Additionally, this option can specify that GCC should tune the performance
12513 of the code for a big.LITTLE system. Permissible names are:
12514 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53}.
12516 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
12517 performance for a blend of processors within architecture @var{arch}.
12518 The aim is to generate code that run well on the current most popular
12519 processors, balancing between optimizations that benefit some CPUs in the
12520 range, and avoiding performance pitfalls of other CPUs. The effects of
12521 this option may change in future GCC versions as CPU models come and go.
12523 @option{-mtune=native} causes the compiler to auto-detect the CPU
12524 of the build computer. At present, this feature is only supported on
12525 Linux, and not all architectures are recognized. If the auto-detect is
12526 unsuccessful the option has no effect.
12528 @item -mcpu=@var{name}
12530 This specifies the name of the target ARM processor. GCC uses this name
12531 to derive the name of the target ARM architecture (as if specified
12532 by @option{-march}) and the ARM processor type for which to tune for
12533 performance (as if specified by @option{-mtune}). Where this option
12534 is used in conjunction with @option{-march} or @option{-mtune},
12535 those options take precedence over the appropriate part of this option.
12537 Permissible names for this option are the same as those for
12540 @option{-mcpu=generic-@var{arch}} is also permissible, and is
12541 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
12542 See @option{-mtune} for more information.
12544 @option{-mcpu=native} causes the compiler to auto-detect the CPU
12545 of the build computer. At present, this feature is only supported on
12546 Linux, and not all architectures are recognized. If the auto-detect is
12547 unsuccessful the option has no effect.
12549 @item -mfpu=@var{name}
12551 This specifies what floating-point hardware (or hardware emulation) is
12552 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
12553 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
12554 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
12555 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
12556 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
12558 If @option{-msoft-float} is specified this specifies the format of
12559 floating-point values.
12561 If the selected floating-point hardware includes the NEON extension
12562 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
12563 operations are not generated by GCC's auto-vectorization pass unless
12564 @option{-funsafe-math-optimizations} is also specified. This is
12565 because NEON hardware does not fully implement the IEEE 754 standard for
12566 floating-point arithmetic (in particular denormal values are treated as
12567 zero), so the use of NEON instructions may lead to a loss of precision.
12569 @item -mfp16-format=@var{name}
12570 @opindex mfp16-format
12571 Specify the format of the @code{__fp16} half-precision floating-point type.
12572 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
12573 the default is @samp{none}, in which case the @code{__fp16} type is not
12574 defined. @xref{Half-Precision}, for more information.
12576 @item -mstructure-size-boundary=@var{n}
12577 @opindex mstructure-size-boundary
12578 The sizes of all structures and unions are rounded up to a multiple
12579 of the number of bits set by this option. Permissible values are 8, 32
12580 and 64. The default value varies for different toolchains. For the COFF
12581 targeted toolchain the default value is 8. A value of 64 is only allowed
12582 if the underlying ABI supports it.
12584 Specifying a larger number can produce faster, more efficient code, but
12585 can also increase the size of the program. Different values are potentially
12586 incompatible. Code compiled with one value cannot necessarily expect to
12587 work with code or libraries compiled with another value, if they exchange
12588 information using structures or unions.
12590 @item -mabort-on-noreturn
12591 @opindex mabort-on-noreturn
12592 Generate a call to the function @code{abort} at the end of a
12593 @code{noreturn} function. It is executed if the function tries to
12597 @itemx -mno-long-calls
12598 @opindex mlong-calls
12599 @opindex mno-long-calls
12600 Tells the compiler to perform function calls by first loading the
12601 address of the function into a register and then performing a subroutine
12602 call on this register. This switch is needed if the target function
12603 lies outside of the 64-megabyte addressing range of the offset-based
12604 version of subroutine call instruction.
12606 Even if this switch is enabled, not all function calls are turned
12607 into long calls. The heuristic is that static functions, functions
12608 that have the @samp{short-call} attribute, functions that are inside
12609 the scope of a @samp{#pragma no_long_calls} directive, and functions whose
12610 definitions have already been compiled within the current compilation
12611 unit are not turned into long calls. The exceptions to this rule are
12612 that weak function definitions, functions with the @samp{long-call}
12613 attribute or the @samp{section} attribute, and functions that are within
12614 the scope of a @samp{#pragma long_calls} directive are always
12615 turned into long calls.
12617 This feature is not enabled by default. Specifying
12618 @option{-mno-long-calls} restores the default behavior, as does
12619 placing the function calls within the scope of a @samp{#pragma
12620 long_calls_off} directive. Note these switches have no effect on how
12621 the compiler generates code to handle function calls via function
12624 @item -msingle-pic-base
12625 @opindex msingle-pic-base
12626 Treat the register used for PIC addressing as read-only, rather than
12627 loading it in the prologue for each function. The runtime system is
12628 responsible for initializing this register with an appropriate value
12629 before execution begins.
12631 @item -mpic-register=@var{reg}
12632 @opindex mpic-register
12633 Specify the register to be used for PIC addressing.
12634 For standard PIC base case, the default will be any suitable register
12635 determined by compiler. For single PIC base case, the default is
12636 @samp{R9} if target is EABI based or stack-checking is enabled,
12637 otherwise the default is @samp{R10}.
12639 @item -mpic-data-is-text-relative
12640 @opindex mpic-data-is-text-relative
12641 Assume that each data segments are relative to text segment at load time.
12642 Therefore, it permits addressing data using PC-relative operations.
12643 This option is on by default for targets other than VxWorks RTP.
12645 @item -mpoke-function-name
12646 @opindex mpoke-function-name
12647 Write the name of each function into the text section, directly
12648 preceding the function prologue. The generated code is similar to this:
12652 .ascii "arm_poke_function_name", 0
12655 .word 0xff000000 + (t1 - t0)
12656 arm_poke_function_name
12658 stmfd sp!, @{fp, ip, lr, pc@}
12662 When performing a stack backtrace, code can inspect the value of
12663 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
12664 location @code{pc - 12} and the top 8 bits are set, then we know that
12665 there is a function name embedded immediately preceding this location
12666 and has length @code{((pc[-3]) & 0xff000000)}.
12673 Select between generating code that executes in ARM and Thumb
12674 states. The default for most configurations is to generate code
12675 that executes in ARM state, but the default can be changed by
12676 configuring GCC with the @option{--with-mode=}@var{state}
12680 @opindex mtpcs-frame
12681 Generate a stack frame that is compliant with the Thumb Procedure Call
12682 Standard for all non-leaf functions. (A leaf function is one that does
12683 not call any other functions.) The default is @option{-mno-tpcs-frame}.
12685 @item -mtpcs-leaf-frame
12686 @opindex mtpcs-leaf-frame
12687 Generate a stack frame that is compliant with the Thumb Procedure Call
12688 Standard for all leaf functions. (A leaf function is one that does
12689 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
12691 @item -mcallee-super-interworking
12692 @opindex mcallee-super-interworking
12693 Gives all externally visible functions in the file being compiled an ARM
12694 instruction set header which switches to Thumb mode before executing the
12695 rest of the function. This allows these functions to be called from
12696 non-interworking code. This option is not valid in AAPCS configurations
12697 because interworking is enabled by default.
12699 @item -mcaller-super-interworking
12700 @opindex mcaller-super-interworking
12701 Allows calls via function pointers (including virtual functions) to
12702 execute correctly regardless of whether the target code has been
12703 compiled for interworking or not. There is a small overhead in the cost
12704 of executing a function pointer if this option is enabled. This option
12705 is not valid in AAPCS configurations because interworking is enabled
12708 @item -mtp=@var{name}
12710 Specify the access model for the thread local storage pointer. The valid
12711 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
12712 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
12713 (supported in the arm6k architecture), and @option{auto}, which uses the
12714 best available method for the selected processor. The default setting is
12717 @item -mtls-dialect=@var{dialect}
12718 @opindex mtls-dialect
12719 Specify the dialect to use for accessing thread local storage. Two
12720 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
12721 @samp{gnu} dialect selects the original GNU scheme for supporting
12722 local and global dynamic TLS models. The @samp{gnu2} dialect
12723 selects the GNU descriptor scheme, which provides better performance
12724 for shared libraries. The GNU descriptor scheme is compatible with
12725 the original scheme, but does require new assembler, linker and
12726 library support. Initial and local exec TLS models are unaffected by
12727 this option and always use the original scheme.
12729 @item -mword-relocations
12730 @opindex mword-relocations
12731 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
12732 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
12733 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
12736 @item -mfix-cortex-m3-ldrd
12737 @opindex mfix-cortex-m3-ldrd
12738 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
12739 with overlapping destination and base registers are used. This option avoids
12740 generating these instructions. This option is enabled by default when
12741 @option{-mcpu=cortex-m3} is specified.
12743 @item -munaligned-access
12744 @itemx -mno-unaligned-access
12745 @opindex munaligned-access
12746 @opindex mno-unaligned-access
12747 Enables (or disables) reading and writing of 16- and 32- bit values
12748 from addresses that are not 16- or 32- bit aligned. By default
12749 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
12750 architectures, and enabled for all other architectures. If unaligned
12751 access is not enabled then words in packed data structures will be
12752 accessed a byte at a time.
12754 The ARM attribute @code{Tag_CPU_unaligned_access} will be set in the
12755 generated object file to either true or false, depending upon the
12756 setting of this option. If unaligned access is enabled then the
12757 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} will also be
12760 @item -mneon-for-64bits
12761 @opindex mneon-for-64bits
12762 Enables using Neon to handle scalar 64-bits operations. This is
12763 disabled by default since the cost of moving data from core registers
12766 @item -mslow-flash-data
12767 @opindex mslow-flash-data
12768 Assume loading data from flash is slower than fetching instruction.
12769 Therefore literal load is minimized for better performance.
12770 This option is only supported when compiling for ARMv7 M-profile and
12773 @item -mrestrict-it
12774 @opindex mrestrict-it
12775 Restricts generation of IT blocks to conform to the rules of ARMv8.
12776 IT blocks can only contain a single 16-bit instruction from a select
12777 set of instructions. This option is on by default for ARMv8 Thumb mode.
12781 @subsection AVR Options
12782 @cindex AVR Options
12784 These options are defined for AVR implementations:
12787 @item -mmcu=@var{mcu}
12789 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
12791 The default for this option is@tie{}@code{avr2}.
12793 GCC supports the following AVR devices and ISAs:
12795 @include avr-mmcu.texi
12797 @item -maccumulate-args
12798 @opindex maccumulate-args
12799 Accumulate outgoing function arguments and acquire/release the needed
12800 stack space for outgoing function arguments once in function
12801 prologue/epilogue. Without this option, outgoing arguments are pushed
12802 before calling a function and popped afterwards.
12804 Popping the arguments after the function call can be expensive on
12805 AVR so that accumulating the stack space might lead to smaller
12806 executables because arguments need not to be removed from the
12807 stack after such a function call.
12809 This option can lead to reduced code size for functions that perform
12810 several calls to functions that get their arguments on the stack like
12811 calls to printf-like functions.
12813 @item -mbranch-cost=@var{cost}
12814 @opindex mbranch-cost
12815 Set the branch costs for conditional branch instructions to
12816 @var{cost}. Reasonable values for @var{cost} are small, non-negative
12817 integers. The default branch cost is 0.
12819 @item -mcall-prologues
12820 @opindex mcall-prologues
12821 Functions prologues/epilogues are expanded as calls to appropriate
12822 subroutines. Code size is smaller.
12826 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
12827 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
12828 and @code{long long} is 4 bytes. Please note that this option does not
12829 conform to the C standards, but it results in smaller code
12832 @item -mno-interrupts
12833 @opindex mno-interrupts
12834 Generated code is not compatible with hardware interrupts.
12835 Code size is smaller.
12839 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
12840 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
12841 Setting @code{-mrelax} just adds the @code{--relax} option to the
12842 linker command line when the linker is called.
12844 Jump relaxing is performed by the linker because jump offsets are not
12845 known before code is located. Therefore, the assembler code generated by the
12846 compiler is the same, but the instructions in the executable may
12847 differ from instructions in the assembler code.
12849 Relaxing must be turned on if linker stubs are needed, see the
12850 section on @code{EIND} and linker stubs below.
12854 Treat the stack pointer register as an 8-bit register,
12855 i.e.@: assume the high byte of the stack pointer is zero.
12856 In general, you don't need to set this option by hand.
12858 This option is used internally by the compiler to select and
12859 build multilibs for architectures @code{avr2} and @code{avr25}.
12860 These architectures mix devices with and without @code{SPH}.
12861 For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25}
12862 the compiler driver will add or remove this option from the compiler
12863 proper's command line, because the compiler then knows if the device
12864 or architecture has an 8-bit stack pointer and thus no @code{SPH}
12869 Use address register @code{X} in a way proposed by the hardware. This means
12870 that @code{X} is only used in indirect, post-increment or
12871 pre-decrement addressing.
12873 Without this option, the @code{X} register may be used in the same way
12874 as @code{Y} or @code{Z} which then is emulated by additional
12876 For example, loading a value with @code{X+const} addressing with a
12877 small non-negative @code{const < 64} to a register @var{Rn} is
12881 adiw r26, const ; X += const
12882 ld @var{Rn}, X ; @var{Rn} = *X
12883 sbiw r26, const ; X -= const
12887 @opindex mtiny-stack
12888 Only change the lower 8@tie{}bits of the stack pointer.
12890 @item -Waddr-space-convert
12891 @opindex Waddr-space-convert
12892 Warn about conversions between address spaces in the case where the
12893 resulting address space is not contained in the incoming address space.
12896 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
12897 @cindex @code{EIND}
12898 Pointers in the implementation are 16@tie{}bits wide.
12899 The address of a function or label is represented as word address so
12900 that indirect jumps and calls can target any code address in the
12901 range of 64@tie{}Ki words.
12903 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
12904 bytes of program memory space, there is a special function register called
12905 @code{EIND} that serves as most significant part of the target address
12906 when @code{EICALL} or @code{EIJMP} instructions are used.
12908 Indirect jumps and calls on these devices are handled as follows by
12909 the compiler and are subject to some limitations:
12914 The compiler never sets @code{EIND}.
12917 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
12918 instructions or might read @code{EIND} directly in order to emulate an
12919 indirect call/jump by means of a @code{RET} instruction.
12922 The compiler assumes that @code{EIND} never changes during the startup
12923 code or during the application. In particular, @code{EIND} is not
12924 saved/restored in function or interrupt service routine
12928 For indirect calls to functions and computed goto, the linker
12929 generates @emph{stubs}. Stubs are jump pads sometimes also called
12930 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
12931 The stub contains a direct jump to the desired address.
12934 Linker relaxation must be turned on so that the linker will generate
12935 the stubs correctly an all situaltion. See the compiler option
12936 @code{-mrelax} and the linler option @code{--relax}.
12937 There are corner cases where the linker is supposed to generate stubs
12938 but aborts without relaxation and without a helpful error message.
12941 The default linker script is arranged for code with @code{EIND = 0}.
12942 If code is supposed to work for a setup with @code{EIND != 0}, a custom
12943 linker script has to be used in order to place the sections whose
12944 name start with @code{.trampolines} into the segment where @code{EIND}
12948 The startup code from libgcc never sets @code{EIND}.
12949 Notice that startup code is a blend of code from libgcc and AVR-LibC.
12950 For the impact of AVR-LibC on @code{EIND}, see the
12951 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
12954 It is legitimate for user-specific startup code to set up @code{EIND}
12955 early, for example by means of initialization code located in
12956 section @code{.init3}. Such code runs prior to general startup code
12957 that initializes RAM and calls constructors, but after the bit
12958 of startup code from AVR-LibC that sets @code{EIND} to the segment
12959 where the vector table is located.
12961 #include <avr/io.h>
12964 __attribute__((section(".init3"),naked,used,no_instrument_function))
12965 init3_set_eind (void)
12967 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
12968 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
12973 The @code{__trampolines_start} symbol is defined in the linker script.
12976 Stubs are generated automatically by the linker if
12977 the following two conditions are met:
12980 @item The address of a label is taken by means of the @code{gs} modifier
12981 (short for @emph{generate stubs}) like so:
12983 LDI r24, lo8(gs(@var{func}))
12984 LDI r25, hi8(gs(@var{func}))
12986 @item The final location of that label is in a code segment
12987 @emph{outside} the segment where the stubs are located.
12991 The compiler emits such @code{gs} modifiers for code labels in the
12992 following situations:
12994 @item Taking address of a function or code label.
12995 @item Computed goto.
12996 @item If prologue-save function is used, see @option{-mcall-prologues}
12997 command-line option.
12998 @item Switch/case dispatch tables. If you do not want such dispatch
12999 tables you can specify the @option{-fno-jump-tables} command-line option.
13000 @item C and C++ constructors/destructors called during startup/shutdown.
13001 @item If the tools hit a @code{gs()} modifier explained above.
13005 Jumping to non-symbolic addresses like so is @emph{not} supported:
13010 /* Call function at word address 0x2 */
13011 return ((int(*)(void)) 0x2)();
13015 Instead, a stub has to be set up, i.e.@: the function has to be called
13016 through a symbol (@code{func_4} in the example):
13021 extern int func_4 (void);
13023 /* Call function at byte address 0x4 */
13028 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
13029 Alternatively, @code{func_4} can be defined in the linker script.
13032 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13033 @cindex @code{RAMPD}
13034 @cindex @code{RAMPX}
13035 @cindex @code{RAMPY}
13036 @cindex @code{RAMPZ}
13037 Some AVR devices support memories larger than the 64@tie{}KiB range
13038 that can be accessed with 16-bit pointers. To access memory locations
13039 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13040 register is used as high part of the address:
13041 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13042 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13043 register, respectively, to get a wide address. Similarly,
13044 @code{RAMPD} is used together with direct addressing.
13048 The startup code initializes the @code{RAMP} special function
13049 registers with zero.
13052 If a @ref{AVR Named Address Spaces,named address space} other than
13053 generic or @code{__flash} is used, then @code{RAMPZ} is set
13054 as needed before the operation.
13057 If the device supports RAM larger than 64@tie{}KiB and the compiler
13058 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13059 is reset to zero after the operation.
13062 If the device comes with a specific @code{RAMP} register, the ISR
13063 prologue/epilogue saves/restores that SFR and initializes it with
13064 zero in case the ISR code might (implicitly) use it.
13067 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13068 If you use inline assembler to read from locations outside the
13069 16-bit address range and change one of the @code{RAMP} registers,
13070 you must reset it to zero after the access.
13074 @subsubsection AVR Built-in Macros
13076 GCC defines several built-in macros so that the user code can test
13077 for the presence or absence of features. Almost any of the following
13078 built-in macros are deduced from device capabilities and thus
13079 triggered by the @code{-mmcu=} command-line option.
13081 For even more AVR-specific built-in macros see
13082 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13087 Build-in macro that resolves to a decimal number that identifies the
13088 architecture and depends on the @code{-mmcu=@var{mcu}} option.
13089 Possible values are:
13091 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13092 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
13093 @code{105}, @code{106}, @code{107}
13095 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
13096 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
13097 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
13098 @code{avrxmega6}, @code{avrxmega7}, respectively.
13099 If @var{mcu} specifies a device, this built-in macro is set
13100 accordingly. For example, with @code{-mmcu=atmega8} the macro will be
13101 defined to @code{4}.
13103 @item __AVR_@var{Device}__
13104 Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects
13105 the device's name. For example, @code{-mmcu=atmega8} defines the
13106 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
13107 @code{__AVR_ATtiny261A__}, etc.
13109 The built-in macros' names follow
13110 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13111 the device name as from the AVR user manual. The difference between
13112 @var{Device} in the built-in macro and @var{device} in
13113 @code{-mmcu=@var{device}} is that the latter is always lowercase.
13115 If @var{device} is not a device but only a core architecture like
13116 @code{avr51}, this macro will not be defined.
13118 @item __AVR_DEVICE_NAME__
13119 Setting @code{-mmcu=@var{device}} defines this built-in macro to
13120 the device's name. For example, with @code{-mmcu=atmega8} the macro
13121 will be defined to @code{atmega8}.
13123 If @var{device} is not a device but only a core architecture like
13124 @code{avr51}, this macro will not be defined.
13126 @item __AVR_XMEGA__
13127 The device / architecture belongs to the XMEGA family of devices.
13129 @item __AVR_HAVE_ELPM__
13130 The device has the the @code{ELPM} instruction.
13132 @item __AVR_HAVE_ELPMX__
13133 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13134 R@var{n},Z+} instructions.
13136 @item __AVR_HAVE_MOVW__
13137 The device has the @code{MOVW} instruction to perform 16-bit
13138 register-register moves.
13140 @item __AVR_HAVE_LPMX__
13141 The device has the @code{LPM R@var{n},Z} and
13142 @code{LPM R@var{n},Z+} instructions.
13144 @item __AVR_HAVE_MUL__
13145 The device has a hardware multiplier.
13147 @item __AVR_HAVE_JMP_CALL__
13148 The device has the @code{JMP} and @code{CALL} instructions.
13149 This is the case for devices with at least 16@tie{}KiB of program
13152 @item __AVR_HAVE_EIJMP_EICALL__
13153 @itemx __AVR_3_BYTE_PC__
13154 The device has the @code{EIJMP} and @code{EICALL} instructions.
13155 This is the case for devices with more than 128@tie{}KiB of program memory.
13156 This also means that the program counter
13157 (PC) is 3@tie{}bytes wide.
13159 @item __AVR_2_BYTE_PC__
13160 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13161 with up to 128@tie{}KiB of program memory.
13163 @item __AVR_HAVE_8BIT_SP__
13164 @itemx __AVR_HAVE_16BIT_SP__
13165 The stack pointer (SP) register is treated as 8-bit respectively
13166 16-bit register by the compiler.
13167 The definition of these macros is affected by @code{-mtiny-stack}.
13169 @item __AVR_HAVE_SPH__
13171 The device has the SPH (high part of stack pointer) special function
13172 register or has an 8-bit stack pointer, respectively.
13173 The definition of these macros is affected by @code{-mmcu=} and
13174 in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also
13177 @item __AVR_HAVE_RAMPD__
13178 @itemx __AVR_HAVE_RAMPX__
13179 @itemx __AVR_HAVE_RAMPY__
13180 @itemx __AVR_HAVE_RAMPZ__
13181 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13182 @code{RAMPZ} special function register, respectively.
13184 @item __NO_INTERRUPTS__
13185 This macro reflects the @code{-mno-interrupts} command line option.
13187 @item __AVR_ERRATA_SKIP__
13188 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13189 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13190 instructions because of a hardware erratum. Skip instructions are
13191 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13192 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13195 @item __AVR_ISA_RMW__
13196 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13198 @item __AVR_SFR_OFFSET__=@var{offset}
13199 Instructions that can address I/O special function registers directly
13200 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13201 address as if addressed by an instruction to access RAM like @code{LD}
13202 or @code{STS}. This offset depends on the device architecture and has
13203 to be subtracted from the RAM address in order to get the
13204 respective I/O@tie{}address.
13206 @item __WITH_AVRLIBC__
13207 The compiler is configured to be used together with AVR-Libc.
13208 See the @code{--with-avrlibc} configure option.
13212 @node Blackfin Options
13213 @subsection Blackfin Options
13214 @cindex Blackfin Options
13217 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13219 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13220 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13221 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13222 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13223 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13224 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13225 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13226 @samp{bf561}, @samp{bf592}.
13228 The optional @var{sirevision} specifies the silicon revision of the target
13229 Blackfin processor. Any workarounds available for the targeted silicon revision
13230 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13231 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13232 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13233 hexadecimal digits representing the major and minor numbers in the silicon
13234 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13235 is not defined. If @var{sirevision} is @samp{any}, the
13236 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13237 If this optional @var{sirevision} is not used, GCC assumes the latest known
13238 silicon revision of the targeted Blackfin processor.
13240 GCC defines a preprocessor macro for the specified @var{cpu}.
13241 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13242 provided by libgloss to be linked in if @option{-msim} is not given.
13244 Without this option, @samp{bf532} is used as the processor by default.
13246 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13247 only the preprocessor macro is defined.
13251 Specifies that the program will be run on the simulator. This causes
13252 the simulator BSP provided by libgloss to be linked in. This option
13253 has effect only for @samp{bfin-elf} toolchain.
13254 Certain other options, such as @option{-mid-shared-library} and
13255 @option{-mfdpic}, imply @option{-msim}.
13257 @item -momit-leaf-frame-pointer
13258 @opindex momit-leaf-frame-pointer
13259 Don't keep the frame pointer in a register for leaf functions. This
13260 avoids the instructions to save, set up and restore frame pointers and
13261 makes an extra register available in leaf functions. The option
13262 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13263 which might make debugging harder.
13265 @item -mspecld-anomaly
13266 @opindex mspecld-anomaly
13267 When enabled, the compiler ensures that the generated code does not
13268 contain speculative loads after jump instructions. If this option is used,
13269 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13271 @item -mno-specld-anomaly
13272 @opindex mno-specld-anomaly
13273 Don't generate extra code to prevent speculative loads from occurring.
13275 @item -mcsync-anomaly
13276 @opindex mcsync-anomaly
13277 When enabled, the compiler ensures that the generated code does not
13278 contain CSYNC or SSYNC instructions too soon after conditional branches.
13279 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13281 @item -mno-csync-anomaly
13282 @opindex mno-csync-anomaly
13283 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13284 occurring too soon after a conditional branch.
13288 When enabled, the compiler is free to take advantage of the knowledge that
13289 the entire program fits into the low 64k of memory.
13292 @opindex mno-low-64k
13293 Assume that the program is arbitrarily large. This is the default.
13295 @item -mstack-check-l1
13296 @opindex mstack-check-l1
13297 Do stack checking using information placed into L1 scratchpad memory by the
13300 @item -mid-shared-library
13301 @opindex mid-shared-library
13302 Generate code that supports shared libraries via the library ID method.
13303 This allows for execute in place and shared libraries in an environment
13304 without virtual memory management. This option implies @option{-fPIC}.
13305 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13307 @item -mno-id-shared-library
13308 @opindex mno-id-shared-library
13309 Generate code that doesn't assume ID-based shared libraries are being used.
13310 This is the default.
13312 @item -mleaf-id-shared-library
13313 @opindex mleaf-id-shared-library
13314 Generate code that supports shared libraries via the library ID method,
13315 but assumes that this library or executable won't link against any other
13316 ID shared libraries. That allows the compiler to use faster code for jumps
13319 @item -mno-leaf-id-shared-library
13320 @opindex mno-leaf-id-shared-library
13321 Do not assume that the code being compiled won't link against any ID shared
13322 libraries. Slower code is generated for jump and call insns.
13324 @item -mshared-library-id=n
13325 @opindex mshared-library-id
13326 Specifies the identification number of the ID-based shared library being
13327 compiled. Specifying a value of 0 generates more compact code; specifying
13328 other values forces the allocation of that number to the current
13329 library but is no more space- or time-efficient than omitting this option.
13333 Generate code that allows the data segment to be located in a different
13334 area of memory from the text segment. This allows for execute in place in
13335 an environment without virtual memory management by eliminating relocations
13336 against the text section.
13338 @item -mno-sep-data
13339 @opindex mno-sep-data
13340 Generate code that assumes that the data segment follows the text segment.
13341 This is the default.
13344 @itemx -mno-long-calls
13345 @opindex mlong-calls
13346 @opindex mno-long-calls
13347 Tells the compiler to perform function calls by first loading the
13348 address of the function into a register and then performing a subroutine
13349 call on this register. This switch is needed if the target function
13350 lies outside of the 24-bit addressing range of the offset-based
13351 version of subroutine call instruction.
13353 This feature is not enabled by default. Specifying
13354 @option{-mno-long-calls} restores the default behavior. Note these
13355 switches have no effect on how the compiler generates code to handle
13356 function calls via function pointers.
13360 Link with the fast floating-point library. This library relaxes some of
13361 the IEEE floating-point standard's rules for checking inputs against
13362 Not-a-Number (NAN), in the interest of performance.
13365 @opindex minline-plt
13366 Enable inlining of PLT entries in function calls to functions that are
13367 not known to bind locally. It has no effect without @option{-mfdpic}.
13370 @opindex mmulticore
13371 Build a standalone application for multicore Blackfin processors.
13372 This option causes proper start files and link scripts supporting
13373 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
13374 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
13376 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
13377 selects the one-application-per-core programming model. Without
13378 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
13379 programming model is used. In this model, the main function of Core B
13380 should be named as @code{coreb_main}.
13382 If this option is not used, the single-core application programming
13387 Build a standalone application for Core A of BF561 when using
13388 the one-application-per-core programming model. Proper start files
13389 and link scripts are used to support Core A, and the macro
13390 @code{__BFIN_COREA} is defined.
13391 This option can only be used in conjunction with @option{-mmulticore}.
13395 Build a standalone application for Core B of BF561 when using
13396 the one-application-per-core programming model. Proper start files
13397 and link scripts are used to support Core B, and the macro
13398 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
13399 should be used instead of @code{main}.
13400 This option can only be used in conjunction with @option{-mmulticore}.
13404 Build a standalone application for SDRAM. Proper start files and
13405 link scripts are used to put the application into SDRAM, and the macro
13406 @code{__BFIN_SDRAM} is defined.
13407 The loader should initialize SDRAM before loading the application.
13411 Assume that ICPLBs are enabled at run time. This has an effect on certain
13412 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
13413 are enabled; for standalone applications the default is off.
13417 @subsection C6X Options
13418 @cindex C6X Options
13421 @item -march=@var{name}
13423 This specifies the name of the target architecture. GCC uses this
13424 name to determine what kind of instructions it can emit when generating
13425 assembly code. Permissible names are: @samp{c62x},
13426 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
13429 @opindex mbig-endian
13430 Generate code for a big-endian target.
13432 @item -mlittle-endian
13433 @opindex mlittle-endian
13434 Generate code for a little-endian target. This is the default.
13438 Choose startup files and linker script suitable for the simulator.
13440 @item -msdata=default
13441 @opindex msdata=default
13442 Put small global and static data in the @samp{.neardata} section,
13443 which is pointed to by register @code{B14}. Put small uninitialized
13444 global and static data in the @samp{.bss} section, which is adjacent
13445 to the @samp{.neardata} section. Put small read-only data into the
13446 @samp{.rodata} section. The corresponding sections used for large
13447 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
13450 @opindex msdata=all
13451 Put all data, not just small objects, into the sections reserved for
13452 small data, and use addressing relative to the @code{B14} register to
13456 @opindex msdata=none
13457 Make no use of the sections reserved for small data, and use absolute
13458 addresses to access all data. Put all initialized global and static
13459 data in the @samp{.fardata} section, and all uninitialized data in the
13460 @samp{.far} section. Put all constant data into the @samp{.const}
13465 @subsection CRIS Options
13466 @cindex CRIS Options
13468 These options are defined specifically for the CRIS ports.
13471 @item -march=@var{architecture-type}
13472 @itemx -mcpu=@var{architecture-type}
13475 Generate code for the specified architecture. The choices for
13476 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
13477 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
13478 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
13481 @item -mtune=@var{architecture-type}
13483 Tune to @var{architecture-type} everything applicable about the generated
13484 code, except for the ABI and the set of available instructions. The
13485 choices for @var{architecture-type} are the same as for
13486 @option{-march=@var{architecture-type}}.
13488 @item -mmax-stack-frame=@var{n}
13489 @opindex mmax-stack-frame
13490 Warn when the stack frame of a function exceeds @var{n} bytes.
13496 The options @option{-metrax4} and @option{-metrax100} are synonyms for
13497 @option{-march=v3} and @option{-march=v8} respectively.
13499 @item -mmul-bug-workaround
13500 @itemx -mno-mul-bug-workaround
13501 @opindex mmul-bug-workaround
13502 @opindex mno-mul-bug-workaround
13503 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
13504 models where it applies. This option is active by default.
13508 Enable CRIS-specific verbose debug-related information in the assembly
13509 code. This option also has the effect of turning off the @samp{#NO_APP}
13510 formatted-code indicator to the assembler at the beginning of the
13515 Do not use condition-code results from previous instruction; always emit
13516 compare and test instructions before use of condition codes.
13518 @item -mno-side-effects
13519 @opindex mno-side-effects
13520 Do not emit instructions with side effects in addressing modes other than
13523 @item -mstack-align
13524 @itemx -mno-stack-align
13525 @itemx -mdata-align
13526 @itemx -mno-data-align
13527 @itemx -mconst-align
13528 @itemx -mno-const-align
13529 @opindex mstack-align
13530 @opindex mno-stack-align
13531 @opindex mdata-align
13532 @opindex mno-data-align
13533 @opindex mconst-align
13534 @opindex mno-const-align
13535 These options (@samp{no-} options) arrange (eliminate arrangements) for the
13536 stack frame, individual data and constants to be aligned for the maximum
13537 single data access size for the chosen CPU model. The default is to
13538 arrange for 32-bit alignment. ABI details such as structure layout are
13539 not affected by these options.
13547 Similar to the stack- data- and const-align options above, these options
13548 arrange for stack frame, writable data and constants to all be 32-bit,
13549 16-bit or 8-bit aligned. The default is 32-bit alignment.
13551 @item -mno-prologue-epilogue
13552 @itemx -mprologue-epilogue
13553 @opindex mno-prologue-epilogue
13554 @opindex mprologue-epilogue
13555 With @option{-mno-prologue-epilogue}, the normal function prologue and
13556 epilogue which set up the stack frame are omitted and no return
13557 instructions or return sequences are generated in the code. Use this
13558 option only together with visual inspection of the compiled code: no
13559 warnings or errors are generated when call-saved registers must be saved,
13560 or storage for local variables needs to be allocated.
13564 @opindex mno-gotplt
13566 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
13567 instruction sequences that load addresses for functions from the PLT part
13568 of the GOT rather than (traditional on other architectures) calls to the
13569 PLT@. The default is @option{-mgotplt}.
13573 Legacy no-op option only recognized with the cris-axis-elf and
13574 cris-axis-linux-gnu targets.
13578 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
13582 This option, recognized for the cris-axis-elf, arranges
13583 to link with input-output functions from a simulator library. Code,
13584 initialized data and zero-initialized data are allocated consecutively.
13588 Like @option{-sim}, but pass linker options to locate initialized data at
13589 0x40000000 and zero-initialized data at 0x80000000.
13593 @subsection CR16 Options
13594 @cindex CR16 Options
13596 These options are defined specifically for the CR16 ports.
13602 Enable the use of multiply-accumulate instructions. Disabled by default.
13606 @opindex mcr16cplus
13608 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
13613 Links the library libsim.a which is in compatible with simulator. Applicable
13614 to ELF compiler only.
13618 Choose integer type as 32-bit wide.
13622 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
13624 @item -mdata-model=@var{model}
13625 @opindex mdata-model
13626 Choose a data model. The choices for @var{model} are @samp{near},
13627 @samp{far} or @samp{medium}. @samp{medium} is default.
13628 However, @samp{far} is not valid with @option{-mcr16c}, as the
13629 CR16C architecture does not support the far data model.
13632 @node Darwin Options
13633 @subsection Darwin Options
13634 @cindex Darwin options
13636 These options are defined for all architectures running the Darwin operating
13639 FSF GCC on Darwin does not create ``fat'' object files; it creates
13640 an object file for the single architecture that GCC was built to
13641 target. Apple's GCC on Darwin does create ``fat'' files if multiple
13642 @option{-arch} options are used; it does so by running the compiler or
13643 linker multiple times and joining the results together with
13646 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
13647 @samp{i686}) is determined by the flags that specify the ISA
13648 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
13649 @option{-force_cpusubtype_ALL} option can be used to override this.
13651 The Darwin tools vary in their behavior when presented with an ISA
13652 mismatch. The assembler, @file{as}, only permits instructions to
13653 be used that are valid for the subtype of the file it is generating,
13654 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
13655 The linker for shared libraries, @file{/usr/bin/libtool}, fails
13656 and prints an error if asked to create a shared library with a less
13657 restrictive subtype than its input files (for instance, trying to put
13658 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
13659 for executables, @command{ld}, quietly gives the executable the most
13660 restrictive subtype of any of its input files.
13665 Add the framework directory @var{dir} to the head of the list of
13666 directories to be searched for header files. These directories are
13667 interleaved with those specified by @option{-I} options and are
13668 scanned in a left-to-right order.
13670 A framework directory is a directory with frameworks in it. A
13671 framework is a directory with a @file{Headers} and/or
13672 @file{PrivateHeaders} directory contained directly in it that ends
13673 in @file{.framework}. The name of a framework is the name of this
13674 directory excluding the @file{.framework}. Headers associated with
13675 the framework are found in one of those two directories, with
13676 @file{Headers} being searched first. A subframework is a framework
13677 directory that is in a framework's @file{Frameworks} directory.
13678 Includes of subframework headers can only appear in a header of a
13679 framework that contains the subframework, or in a sibling subframework
13680 header. Two subframeworks are siblings if they occur in the same
13681 framework. A subframework should not have the same name as a
13682 framework; a warning is issued if this is violated. Currently a
13683 subframework cannot have subframeworks; in the future, the mechanism
13684 may be extended to support this. The standard frameworks can be found
13685 in @file{/System/Library/Frameworks} and
13686 @file{/Library/Frameworks}. An example include looks like
13687 @code{#include <Framework/header.h>}, where @file{Framework} denotes
13688 the name of the framework and @file{header.h} is found in the
13689 @file{PrivateHeaders} or @file{Headers} directory.
13691 @item -iframework@var{dir}
13692 @opindex iframework
13693 Like @option{-F} except the directory is a treated as a system
13694 directory. The main difference between this @option{-iframework} and
13695 @option{-F} is that with @option{-iframework} the compiler does not
13696 warn about constructs contained within header files found via
13697 @var{dir}. This option is valid only for the C family of languages.
13701 Emit debugging information for symbols that are used. For stabs
13702 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
13703 This is by default ON@.
13707 Emit debugging information for all symbols and types.
13709 @item -mmacosx-version-min=@var{version}
13710 The earliest version of MacOS X that this executable will run on
13711 is @var{version}. Typical values of @var{version} include @code{10.1},
13712 @code{10.2}, and @code{10.3.9}.
13714 If the compiler was built to use the system's headers by default,
13715 then the default for this option is the system version on which the
13716 compiler is running, otherwise the default is to make choices that
13717 are compatible with as many systems and code bases as possible.
13721 Enable kernel development mode. The @option{-mkernel} option sets
13722 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
13723 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
13724 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
13725 applicable. This mode also sets @option{-mno-altivec},
13726 @option{-msoft-float}, @option{-fno-builtin} and
13727 @option{-mlong-branch} for PowerPC targets.
13729 @item -mone-byte-bool
13730 @opindex mone-byte-bool
13731 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
13732 By default @samp{sizeof(bool)} is @samp{4} when compiling for
13733 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
13734 option has no effect on x86.
13736 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
13737 to generate code that is not binary compatible with code generated
13738 without that switch. Using this switch may require recompiling all
13739 other modules in a program, including system libraries. Use this
13740 switch to conform to a non-default data model.
13742 @item -mfix-and-continue
13743 @itemx -ffix-and-continue
13744 @itemx -findirect-data
13745 @opindex mfix-and-continue
13746 @opindex ffix-and-continue
13747 @opindex findirect-data
13748 Generate code suitable for fast turnaround development, such as to
13749 allow GDB to dynamically load @code{.o} files into already-running
13750 programs. @option{-findirect-data} and @option{-ffix-and-continue}
13751 are provided for backwards compatibility.
13755 Loads all members of static archive libraries.
13756 See man ld(1) for more information.
13758 @item -arch_errors_fatal
13759 @opindex arch_errors_fatal
13760 Cause the errors having to do with files that have the wrong architecture
13763 @item -bind_at_load
13764 @opindex bind_at_load
13765 Causes the output file to be marked such that the dynamic linker will
13766 bind all undefined references when the file is loaded or launched.
13770 Produce a Mach-o bundle format file.
13771 See man ld(1) for more information.
13773 @item -bundle_loader @var{executable}
13774 @opindex bundle_loader
13775 This option specifies the @var{executable} that will load the build
13776 output file being linked. See man ld(1) for more information.
13779 @opindex dynamiclib
13780 When passed this option, GCC produces a dynamic library instead of
13781 an executable when linking, using the Darwin @file{libtool} command.
13783 @item -force_cpusubtype_ALL
13784 @opindex force_cpusubtype_ALL
13785 This causes GCC's output file to have the @var{ALL} subtype, instead of
13786 one controlled by the @option{-mcpu} or @option{-march} option.
13788 @item -allowable_client @var{client_name}
13789 @itemx -client_name
13790 @itemx -compatibility_version
13791 @itemx -current_version
13793 @itemx -dependency-file
13795 @itemx -dylinker_install_name
13797 @itemx -exported_symbols_list
13800 @itemx -flat_namespace
13801 @itemx -force_flat_namespace
13802 @itemx -headerpad_max_install_names
13805 @itemx -install_name
13806 @itemx -keep_private_externs
13807 @itemx -multi_module
13808 @itemx -multiply_defined
13809 @itemx -multiply_defined_unused
13812 @itemx -no_dead_strip_inits_and_terms
13813 @itemx -nofixprebinding
13814 @itemx -nomultidefs
13816 @itemx -noseglinkedit
13817 @itemx -pagezero_size
13819 @itemx -prebind_all_twolevel_modules
13820 @itemx -private_bundle
13822 @itemx -read_only_relocs
13824 @itemx -sectobjectsymbols
13828 @itemx -sectobjectsymbols
13831 @itemx -segs_read_only_addr
13833 @itemx -segs_read_write_addr
13834 @itemx -seg_addr_table
13835 @itemx -seg_addr_table_filename
13836 @itemx -seglinkedit
13838 @itemx -segs_read_only_addr
13839 @itemx -segs_read_write_addr
13840 @itemx -single_module
13842 @itemx -sub_library
13844 @itemx -sub_umbrella
13845 @itemx -twolevel_namespace
13848 @itemx -unexported_symbols_list
13849 @itemx -weak_reference_mismatches
13850 @itemx -whatsloaded
13851 @opindex allowable_client
13852 @opindex client_name
13853 @opindex compatibility_version
13854 @opindex current_version
13855 @opindex dead_strip
13856 @opindex dependency-file
13857 @opindex dylib_file
13858 @opindex dylinker_install_name
13860 @opindex exported_symbols_list
13862 @opindex flat_namespace
13863 @opindex force_flat_namespace
13864 @opindex headerpad_max_install_names
13865 @opindex image_base
13867 @opindex install_name
13868 @opindex keep_private_externs
13869 @opindex multi_module
13870 @opindex multiply_defined
13871 @opindex multiply_defined_unused
13872 @opindex noall_load
13873 @opindex no_dead_strip_inits_and_terms
13874 @opindex nofixprebinding
13875 @opindex nomultidefs
13877 @opindex noseglinkedit
13878 @opindex pagezero_size
13880 @opindex prebind_all_twolevel_modules
13881 @opindex private_bundle
13882 @opindex read_only_relocs
13884 @opindex sectobjectsymbols
13887 @opindex sectcreate
13888 @opindex sectobjectsymbols
13891 @opindex segs_read_only_addr
13892 @opindex segs_read_write_addr
13893 @opindex seg_addr_table
13894 @opindex seg_addr_table_filename
13895 @opindex seglinkedit
13897 @opindex segs_read_only_addr
13898 @opindex segs_read_write_addr
13899 @opindex single_module
13901 @opindex sub_library
13902 @opindex sub_umbrella
13903 @opindex twolevel_namespace
13906 @opindex unexported_symbols_list
13907 @opindex weak_reference_mismatches
13908 @opindex whatsloaded
13909 These options are passed to the Darwin linker. The Darwin linker man page
13910 describes them in detail.
13913 @node DEC Alpha Options
13914 @subsection DEC Alpha Options
13916 These @samp{-m} options are defined for the DEC Alpha implementations:
13919 @item -mno-soft-float
13920 @itemx -msoft-float
13921 @opindex mno-soft-float
13922 @opindex msoft-float
13923 Use (do not use) the hardware floating-point instructions for
13924 floating-point operations. When @option{-msoft-float} is specified,
13925 functions in @file{libgcc.a} are used to perform floating-point
13926 operations. Unless they are replaced by routines that emulate the
13927 floating-point operations, or compiled in such a way as to call such
13928 emulations routines, these routines issue floating-point
13929 operations. If you are compiling for an Alpha without floating-point
13930 operations, you must ensure that the library is built so as not to call
13933 Note that Alpha implementations without floating-point operations are
13934 required to have floating-point registers.
13937 @itemx -mno-fp-regs
13939 @opindex mno-fp-regs
13940 Generate code that uses (does not use) the floating-point register set.
13941 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
13942 register set is not used, floating-point operands are passed in integer
13943 registers as if they were integers and floating-point results are passed
13944 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
13945 so any function with a floating-point argument or return value called by code
13946 compiled with @option{-mno-fp-regs} must also be compiled with that
13949 A typical use of this option is building a kernel that does not use,
13950 and hence need not save and restore, any floating-point registers.
13954 The Alpha architecture implements floating-point hardware optimized for
13955 maximum performance. It is mostly compliant with the IEEE floating-point
13956 standard. However, for full compliance, software assistance is
13957 required. This option generates code fully IEEE-compliant code
13958 @emph{except} that the @var{inexact-flag} is not maintained (see below).
13959 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
13960 defined during compilation. The resulting code is less efficient but is
13961 able to correctly support denormalized numbers and exceptional IEEE
13962 values such as not-a-number and plus/minus infinity. Other Alpha
13963 compilers call this option @option{-ieee_with_no_inexact}.
13965 @item -mieee-with-inexact
13966 @opindex mieee-with-inexact
13967 This is like @option{-mieee} except the generated code also maintains
13968 the IEEE @var{inexact-flag}. Turning on this option causes the
13969 generated code to implement fully-compliant IEEE math. In addition to
13970 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
13971 macro. On some Alpha implementations the resulting code may execute
13972 significantly slower than the code generated by default. Since there is
13973 very little code that depends on the @var{inexact-flag}, you should
13974 normally not specify this option. Other Alpha compilers call this
13975 option @option{-ieee_with_inexact}.
13977 @item -mfp-trap-mode=@var{trap-mode}
13978 @opindex mfp-trap-mode
13979 This option controls what floating-point related traps are enabled.
13980 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
13981 The trap mode can be set to one of four values:
13985 This is the default (normal) setting. The only traps that are enabled
13986 are the ones that cannot be disabled in software (e.g., division by zero
13990 In addition to the traps enabled by @samp{n}, underflow traps are enabled
13994 Like @samp{u}, but the instructions are marked to be safe for software
13995 completion (see Alpha architecture manual for details).
13998 Like @samp{su}, but inexact traps are enabled as well.
14001 @item -mfp-rounding-mode=@var{rounding-mode}
14002 @opindex mfp-rounding-mode
14003 Selects the IEEE rounding mode. Other Alpha compilers call this option
14004 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14009 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14010 the nearest machine number or towards the even machine number in case
14014 Round towards minus infinity.
14017 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14020 Dynamic rounding mode. A field in the floating-point control register
14021 (@var{fpcr}, see Alpha architecture reference manual) controls the
14022 rounding mode in effect. The C library initializes this register for
14023 rounding towards plus infinity. Thus, unless your program modifies the
14024 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14027 @item -mtrap-precision=@var{trap-precision}
14028 @opindex mtrap-precision
14029 In the Alpha architecture, floating-point traps are imprecise. This
14030 means without software assistance it is impossible to recover from a
14031 floating trap and program execution normally needs to be terminated.
14032 GCC can generate code that can assist operating system trap handlers
14033 in determining the exact location that caused a floating-point trap.
14034 Depending on the requirements of an application, different levels of
14035 precisions can be selected:
14039 Program precision. This option is the default and means a trap handler
14040 can only identify which program caused a floating-point exception.
14043 Function precision. The trap handler can determine the function that
14044 caused a floating-point exception.
14047 Instruction precision. The trap handler can determine the exact
14048 instruction that caused a floating-point exception.
14051 Other Alpha compilers provide the equivalent options called
14052 @option{-scope_safe} and @option{-resumption_safe}.
14054 @item -mieee-conformant
14055 @opindex mieee-conformant
14056 This option marks the generated code as IEEE conformant. You must not
14057 use this option unless you also specify @option{-mtrap-precision=i} and either
14058 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14059 is to emit the line @samp{.eflag 48} in the function prologue of the
14060 generated assembly file.
14062 @item -mbuild-constants
14063 @opindex mbuild-constants
14064 Normally GCC examines a 32- or 64-bit integer constant to
14065 see if it can construct it from smaller constants in two or three
14066 instructions. If it cannot, it outputs the constant as a literal and
14067 generates code to load it from the data segment at run time.
14069 Use this option to require GCC to construct @emph{all} integer constants
14070 using code, even if it takes more instructions (the maximum is six).
14072 You typically use this option to build a shared library dynamic
14073 loader. Itself a shared library, it must relocate itself in memory
14074 before it can find the variables and constants in its own data segment.
14092 Indicate whether GCC should generate code to use the optional BWX,
14093 CIX, FIX and MAX instruction sets. The default is to use the instruction
14094 sets supported by the CPU type specified via @option{-mcpu=} option or that
14095 of the CPU on which GCC was built if none is specified.
14098 @itemx -mfloat-ieee
14099 @opindex mfloat-vax
14100 @opindex mfloat-ieee
14101 Generate code that uses (does not use) VAX F and G floating-point
14102 arithmetic instead of IEEE single and double precision.
14104 @item -mexplicit-relocs
14105 @itemx -mno-explicit-relocs
14106 @opindex mexplicit-relocs
14107 @opindex mno-explicit-relocs
14108 Older Alpha assemblers provided no way to generate symbol relocations
14109 except via assembler macros. Use of these macros does not allow
14110 optimal instruction scheduling. GNU binutils as of version 2.12
14111 supports a new syntax that allows the compiler to explicitly mark
14112 which relocations should apply to which instructions. This option
14113 is mostly useful for debugging, as GCC detects the capabilities of
14114 the assembler when it is built and sets the default accordingly.
14117 @itemx -mlarge-data
14118 @opindex msmall-data
14119 @opindex mlarge-data
14120 When @option{-mexplicit-relocs} is in effect, static data is
14121 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14122 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14123 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14124 16-bit relocations off of the @code{$gp} register. This limits the
14125 size of the small data area to 64KB, but allows the variables to be
14126 directly accessed via a single instruction.
14128 The default is @option{-mlarge-data}. With this option the data area
14129 is limited to just below 2GB@. Programs that require more than 2GB of
14130 data must use @code{malloc} or @code{mmap} to allocate the data in the
14131 heap instead of in the program's data segment.
14133 When generating code for shared libraries, @option{-fpic} implies
14134 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14137 @itemx -mlarge-text
14138 @opindex msmall-text
14139 @opindex mlarge-text
14140 When @option{-msmall-text} is used, the compiler assumes that the
14141 code of the entire program (or shared library) fits in 4MB, and is
14142 thus reachable with a branch instruction. When @option{-msmall-data}
14143 is used, the compiler can assume that all local symbols share the
14144 same @code{$gp} value, and thus reduce the number of instructions
14145 required for a function call from 4 to 1.
14147 The default is @option{-mlarge-text}.
14149 @item -mcpu=@var{cpu_type}
14151 Set the instruction set and instruction scheduling parameters for
14152 machine type @var{cpu_type}. You can specify either the @samp{EV}
14153 style name or the corresponding chip number. GCC supports scheduling
14154 parameters for the EV4, EV5 and EV6 family of processors and
14155 chooses the default values for the instruction set from the processor
14156 you specify. If you do not specify a processor type, GCC defaults
14157 to the processor on which the compiler was built.
14159 Supported values for @var{cpu_type} are
14165 Schedules as an EV4 and has no instruction set extensions.
14169 Schedules as an EV5 and has no instruction set extensions.
14173 Schedules as an EV5 and supports the BWX extension.
14178 Schedules as an EV5 and supports the BWX and MAX extensions.
14182 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14186 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14189 Native toolchains also support the value @samp{native},
14190 which selects the best architecture option for the host processor.
14191 @option{-mcpu=native} has no effect if GCC does not recognize
14194 @item -mtune=@var{cpu_type}
14196 Set only the instruction scheduling parameters for machine type
14197 @var{cpu_type}. The instruction set is not changed.
14199 Native toolchains also support the value @samp{native},
14200 which selects the best architecture option for the host processor.
14201 @option{-mtune=native} has no effect if GCC does not recognize
14204 @item -mmemory-latency=@var{time}
14205 @opindex mmemory-latency
14206 Sets the latency the scheduler should assume for typical memory
14207 references as seen by the application. This number is highly
14208 dependent on the memory access patterns used by the application
14209 and the size of the external cache on the machine.
14211 Valid options for @var{time} are
14215 A decimal number representing clock cycles.
14221 The compiler contains estimates of the number of clock cycles for
14222 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14223 (also called Dcache, Scache, and Bcache), as well as to main memory.
14224 Note that L3 is only valid for EV5.
14230 @subsection FR30 Options
14231 @cindex FR30 Options
14233 These options are defined specifically for the FR30 port.
14237 @item -msmall-model
14238 @opindex msmall-model
14239 Use the small address space model. This can produce smaller code, but
14240 it does assume that all symbolic values and addresses fit into a
14245 Assume that runtime support has been provided and so there is no need
14246 to include the simulator library (@file{libsim.a}) on the linker
14252 @subsection FRV Options
14253 @cindex FRV Options
14259 Only use the first 32 general-purpose registers.
14264 Use all 64 general-purpose registers.
14269 Use only the first 32 floating-point registers.
14274 Use all 64 floating-point registers.
14277 @opindex mhard-float
14279 Use hardware instructions for floating-point operations.
14282 @opindex msoft-float
14284 Use library routines for floating-point operations.
14289 Dynamically allocate condition code registers.
14294 Do not try to dynamically allocate condition code registers, only
14295 use @code{icc0} and @code{fcc0}.
14300 Change ABI to use double word insns.
14305 Do not use double word instructions.
14310 Use floating-point double instructions.
14313 @opindex mno-double
14315 Do not use floating-point double instructions.
14320 Use media instructions.
14325 Do not use media instructions.
14330 Use multiply and add/subtract instructions.
14333 @opindex mno-muladd
14335 Do not use multiply and add/subtract instructions.
14340 Select the FDPIC ABI, which uses function descriptors to represent
14341 pointers to functions. Without any PIC/PIE-related options, it
14342 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
14343 assumes GOT entries and small data are within a 12-bit range from the
14344 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
14345 are computed with 32 bits.
14346 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14349 @opindex minline-plt
14351 Enable inlining of PLT entries in function calls to functions that are
14352 not known to bind locally. It has no effect without @option{-mfdpic}.
14353 It's enabled by default if optimizing for speed and compiling for
14354 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
14355 optimization option such as @option{-O3} or above is present in the
14361 Assume a large TLS segment when generating thread-local code.
14366 Do not assume a large TLS segment when generating thread-local code.
14371 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
14372 that is known to be in read-only sections. It's enabled by default,
14373 except for @option{-fpic} or @option{-fpie}: even though it may help
14374 make the global offset table smaller, it trades 1 instruction for 4.
14375 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
14376 one of which may be shared by multiple symbols, and it avoids the need
14377 for a GOT entry for the referenced symbol, so it's more likely to be a
14378 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
14380 @item -multilib-library-pic
14381 @opindex multilib-library-pic
14383 Link with the (library, not FD) pic libraries. It's implied by
14384 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
14385 @option{-fpic} without @option{-mfdpic}. You should never have to use
14389 @opindex mlinked-fp
14391 Follow the EABI requirement of always creating a frame pointer whenever
14392 a stack frame is allocated. This option is enabled by default and can
14393 be disabled with @option{-mno-linked-fp}.
14396 @opindex mlong-calls
14398 Use indirect addressing to call functions outside the current
14399 compilation unit. This allows the functions to be placed anywhere
14400 within the 32-bit address space.
14402 @item -malign-labels
14403 @opindex malign-labels
14405 Try to align labels to an 8-byte boundary by inserting NOPs into the
14406 previous packet. This option only has an effect when VLIW packing
14407 is enabled. It doesn't create new packets; it merely adds NOPs to
14410 @item -mlibrary-pic
14411 @opindex mlibrary-pic
14413 Generate position-independent EABI code.
14418 Use only the first four media accumulator registers.
14423 Use all eight media accumulator registers.
14428 Pack VLIW instructions.
14433 Do not pack VLIW instructions.
14436 @opindex mno-eflags
14438 Do not mark ABI switches in e_flags.
14441 @opindex mcond-move
14443 Enable the use of conditional-move instructions (default).
14445 This switch is mainly for debugging the compiler and will likely be removed
14446 in a future version.
14448 @item -mno-cond-move
14449 @opindex mno-cond-move
14451 Disable the use of conditional-move instructions.
14453 This switch is mainly for debugging the compiler and will likely be removed
14454 in a future version.
14459 Enable the use of conditional set instructions (default).
14461 This switch is mainly for debugging the compiler and will likely be removed
14462 in a future version.
14467 Disable the use of conditional set instructions.
14469 This switch is mainly for debugging the compiler and will likely be removed
14470 in a future version.
14473 @opindex mcond-exec
14475 Enable the use of conditional execution (default).
14477 This switch is mainly for debugging the compiler and will likely be removed
14478 in a future version.
14480 @item -mno-cond-exec
14481 @opindex mno-cond-exec
14483 Disable the use of conditional execution.
14485 This switch is mainly for debugging the compiler and will likely be removed
14486 in a future version.
14488 @item -mvliw-branch
14489 @opindex mvliw-branch
14491 Run a pass to pack branches into VLIW instructions (default).
14493 This switch is mainly for debugging the compiler and will likely be removed
14494 in a future version.
14496 @item -mno-vliw-branch
14497 @opindex mno-vliw-branch
14499 Do not run a pass to pack branches into VLIW instructions.
14501 This switch is mainly for debugging the compiler and will likely be removed
14502 in a future version.
14504 @item -mmulti-cond-exec
14505 @opindex mmulti-cond-exec
14507 Enable optimization of @code{&&} and @code{||} in conditional execution
14510 This switch is mainly for debugging the compiler and will likely be removed
14511 in a future version.
14513 @item -mno-multi-cond-exec
14514 @opindex mno-multi-cond-exec
14516 Disable optimization of @code{&&} and @code{||} in conditional execution.
14518 This switch is mainly for debugging the compiler and will likely be removed
14519 in a future version.
14521 @item -mnested-cond-exec
14522 @opindex mnested-cond-exec
14524 Enable nested conditional execution optimizations (default).
14526 This switch is mainly for debugging the compiler and will likely be removed
14527 in a future version.
14529 @item -mno-nested-cond-exec
14530 @opindex mno-nested-cond-exec
14532 Disable nested conditional execution optimizations.
14534 This switch is mainly for debugging the compiler and will likely be removed
14535 in a future version.
14537 @item -moptimize-membar
14538 @opindex moptimize-membar
14540 This switch removes redundant @code{membar} instructions from the
14541 compiler-generated code. It is enabled by default.
14543 @item -mno-optimize-membar
14544 @opindex mno-optimize-membar
14546 This switch disables the automatic removal of redundant @code{membar}
14547 instructions from the generated code.
14549 @item -mtomcat-stats
14550 @opindex mtomcat-stats
14552 Cause gas to print out tomcat statistics.
14554 @item -mcpu=@var{cpu}
14557 Select the processor type for which to generate code. Possible values are
14558 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
14559 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
14563 @node GNU/Linux Options
14564 @subsection GNU/Linux Options
14566 These @samp{-m} options are defined for GNU/Linux targets:
14571 Use the GNU C library. This is the default except
14572 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
14576 Use uClibc C library. This is the default on
14577 @samp{*-*-linux-*uclibc*} targets.
14581 Use Bionic C library. This is the default on
14582 @samp{*-*-linux-*android*} targets.
14586 Compile code compatible with Android platform. This is the default on
14587 @samp{*-*-linux-*android*} targets.
14589 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
14590 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
14591 this option makes the GCC driver pass Android-specific options to the linker.
14592 Finally, this option causes the preprocessor macro @code{__ANDROID__}
14595 @item -tno-android-cc
14596 @opindex tno-android-cc
14597 Disable compilation effects of @option{-mandroid}, i.e., do not enable
14598 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
14599 @option{-fno-rtti} by default.
14601 @item -tno-android-ld
14602 @opindex tno-android-ld
14603 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
14604 linking options to the linker.
14608 @node H8/300 Options
14609 @subsection H8/300 Options
14611 These @samp{-m} options are defined for the H8/300 implementations:
14616 Shorten some address references at link time, when possible; uses the
14617 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
14618 ld, Using ld}, for a fuller description.
14622 Generate code for the H8/300H@.
14626 Generate code for the H8S@.
14630 Generate code for the H8S and H8/300H in the normal mode. This switch
14631 must be used either with @option{-mh} or @option{-ms}.
14635 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
14639 Extended registers are stored on stack before execution of function
14640 with monitor attribute. Default option is @option{-mexr}.
14641 This option is valid only for H8S targets.
14645 Extended registers are not stored on stack before execution of function
14646 with monitor attribute. Default option is @option{-mno-exr}.
14647 This option is valid only for H8S targets.
14651 Make @code{int} data 32 bits by default.
14654 @opindex malign-300
14655 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
14656 The default for the H8/300H and H8S is to align longs and floats on
14658 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
14659 This option has no effect on the H8/300.
14663 @subsection HPPA Options
14664 @cindex HPPA Options
14666 These @samp{-m} options are defined for the HPPA family of computers:
14669 @item -march=@var{architecture-type}
14671 Generate code for the specified architecture. The choices for
14672 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
14673 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
14674 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
14675 architecture option for your machine. Code compiled for lower numbered
14676 architectures runs on higher numbered architectures, but not the
14679 @item -mpa-risc-1-0
14680 @itemx -mpa-risc-1-1
14681 @itemx -mpa-risc-2-0
14682 @opindex mpa-risc-1-0
14683 @opindex mpa-risc-1-1
14684 @opindex mpa-risc-2-0
14685 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
14687 @item -mjump-in-delay
14688 @opindex mjump-in-delay
14689 Fill delay slots of function calls with unconditional jump instructions
14690 by modifying the return pointer for the function call to be the target
14691 of the conditional jump.
14693 @item -mdisable-fpregs
14694 @opindex mdisable-fpregs
14695 Prevent floating-point registers from being used in any manner. This is
14696 necessary for compiling kernels that perform lazy context switching of
14697 floating-point registers. If you use this option and attempt to perform
14698 floating-point operations, the compiler aborts.
14700 @item -mdisable-indexing
14701 @opindex mdisable-indexing
14702 Prevent the compiler from using indexing address modes. This avoids some
14703 rather obscure problems when compiling MIG generated code under MACH@.
14705 @item -mno-space-regs
14706 @opindex mno-space-regs
14707 Generate code that assumes the target has no space registers. This allows
14708 GCC to generate faster indirect calls and use unscaled index address modes.
14710 Such code is suitable for level 0 PA systems and kernels.
14712 @item -mfast-indirect-calls
14713 @opindex mfast-indirect-calls
14714 Generate code that assumes calls never cross space boundaries. This
14715 allows GCC to emit code that performs faster indirect calls.
14717 This option does not work in the presence of shared libraries or nested
14720 @item -mfixed-range=@var{register-range}
14721 @opindex mfixed-range
14722 Generate code treating the given register range as fixed registers.
14723 A fixed register is one that the register allocator cannot use. This is
14724 useful when compiling kernel code. A register range is specified as
14725 two registers separated by a dash. Multiple register ranges can be
14726 specified separated by a comma.
14728 @item -mlong-load-store
14729 @opindex mlong-load-store
14730 Generate 3-instruction load and store sequences as sometimes required by
14731 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
14734 @item -mportable-runtime
14735 @opindex mportable-runtime
14736 Use the portable calling conventions proposed by HP for ELF systems.
14740 Enable the use of assembler directives only GAS understands.
14742 @item -mschedule=@var{cpu-type}
14744 Schedule code according to the constraints for the machine type
14745 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
14746 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
14747 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
14748 proper scheduling option for your machine. The default scheduling is
14752 @opindex mlinker-opt
14753 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
14754 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
14755 linkers in which they give bogus error messages when linking some programs.
14758 @opindex msoft-float
14759 Generate output containing library calls for floating point.
14760 @strong{Warning:} the requisite libraries are not available for all HPPA
14761 targets. Normally the facilities of the machine's usual C compiler are
14762 used, but this cannot be done directly in cross-compilation. You must make
14763 your own arrangements to provide suitable library functions for
14766 @option{-msoft-float} changes the calling convention in the output file;
14767 therefore, it is only useful if you compile @emph{all} of a program with
14768 this option. In particular, you need to compile @file{libgcc.a}, the
14769 library that comes with GCC, with @option{-msoft-float} in order for
14774 Generate the predefine, @code{_SIO}, for server IO@. The default is
14775 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
14776 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
14777 options are available under HP-UX and HI-UX@.
14781 Use options specific to GNU @command{ld}.
14782 This passes @option{-shared} to @command{ld} when
14783 building a shared library. It is the default when GCC is configured,
14784 explicitly or implicitly, with the GNU linker. This option does not
14785 affect which @command{ld} is called; it only changes what parameters
14786 are passed to that @command{ld}.
14787 The @command{ld} that is called is determined by the
14788 @option{--with-ld} configure option, GCC's program search path, and
14789 finally by the user's @env{PATH}. The linker used by GCC can be printed
14790 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
14791 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
14795 Use options specific to HP @command{ld}.
14796 This passes @option{-b} to @command{ld} when building
14797 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
14798 links. It is the default when GCC is configured, explicitly or
14799 implicitly, with the HP linker. This option does not affect
14800 which @command{ld} is called; it only changes what parameters are passed to that
14802 The @command{ld} that is called is determined by the @option{--with-ld}
14803 configure option, GCC's program search path, and finally by the user's
14804 @env{PATH}. The linker used by GCC can be printed using @samp{which
14805 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
14806 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
14809 @opindex mno-long-calls
14810 Generate code that uses long call sequences. This ensures that a call
14811 is always able to reach linker generated stubs. The default is to generate
14812 long calls only when the distance from the call site to the beginning
14813 of the function or translation unit, as the case may be, exceeds a
14814 predefined limit set by the branch type being used. The limits for
14815 normal calls are 7,600,000 and 240,000 bytes, respectively for the
14816 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
14819 Distances are measured from the beginning of functions when using the
14820 @option{-ffunction-sections} option, or when using the @option{-mgas}
14821 and @option{-mno-portable-runtime} options together under HP-UX with
14824 It is normally not desirable to use this option as it degrades
14825 performance. However, it may be useful in large applications,
14826 particularly when partial linking is used to build the application.
14828 The types of long calls used depends on the capabilities of the
14829 assembler and linker, and the type of code being generated. The
14830 impact on systems that support long absolute calls, and long pic
14831 symbol-difference or pc-relative calls should be relatively small.
14832 However, an indirect call is used on 32-bit ELF systems in pic code
14833 and it is quite long.
14835 @item -munix=@var{unix-std}
14837 Generate compiler predefines and select a startfile for the specified
14838 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
14839 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
14840 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
14841 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
14842 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
14845 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
14846 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
14847 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
14848 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
14849 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
14850 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
14852 It is @emph{important} to note that this option changes the interfaces
14853 for various library routines. It also affects the operational behavior
14854 of the C library. Thus, @emph{extreme} care is needed in using this
14857 Library code that is intended to operate with more than one UNIX
14858 standard must test, set and restore the variable @var{__xpg4_extended_mask}
14859 as appropriate. Most GNU software doesn't provide this capability.
14863 Suppress the generation of link options to search libdld.sl when the
14864 @option{-static} option is specified on HP-UX 10 and later.
14868 The HP-UX implementation of setlocale in libc has a dependency on
14869 libdld.sl. There isn't an archive version of libdld.sl. Thus,
14870 when the @option{-static} option is specified, special link options
14871 are needed to resolve this dependency.
14873 On HP-UX 10 and later, the GCC driver adds the necessary options to
14874 link with libdld.sl when the @option{-static} option is specified.
14875 This causes the resulting binary to be dynamic. On the 64-bit port,
14876 the linkers generate dynamic binaries by default in any case. The
14877 @option{-nolibdld} option can be used to prevent the GCC driver from
14878 adding these link options.
14882 Add support for multithreading with the @dfn{dce thread} library
14883 under HP-UX@. This option sets flags for both the preprocessor and
14887 @node i386 and x86-64 Options
14888 @subsection Intel 386 and AMD x86-64 Options
14889 @cindex i386 Options
14890 @cindex x86-64 Options
14891 @cindex Intel 386 Options
14892 @cindex AMD x86-64 Options
14894 These @samp{-m} options are defined for the i386 and x86-64 family of
14899 @item -march=@var{cpu-type}
14901 Generate instructions for the machine type @var{cpu-type}. In contrast to
14902 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
14903 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
14904 to generate code that may not run at all on processors other than the one
14905 indicated. Specifying @option{-march=@var{cpu-type}} implies
14906 @option{-mtune=@var{cpu-type}}.
14908 The choices for @var{cpu-type} are:
14912 This selects the CPU to generate code for at compilation time by determining
14913 the processor type of the compiling machine. Using @option{-march=native}
14914 enables all instruction subsets supported by the local machine (hence
14915 the result might not run on different machines). Using @option{-mtune=native}
14916 produces code optimized for the local machine under the constraints
14917 of the selected instruction set.
14920 Original Intel i386 CPU@.
14923 Intel i486 CPU@. (No scheduling is implemented for this chip.)
14927 Intel Pentium CPU with no MMX support.
14930 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
14933 Intel Pentium Pro CPU@.
14936 When used with @option{-march}, the Pentium Pro
14937 instruction set is used, so the code runs on all i686 family chips.
14938 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
14941 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
14946 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
14950 Intel Pentium M; low-power version of Intel Pentium III CPU
14951 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
14955 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
14958 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
14962 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
14963 SSE2 and SSE3 instruction set support.
14966 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
14967 instruction set support.
14970 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14971 SSE4.1, SSE4.2 and POPCNT instruction set support.
14974 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14975 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
14978 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14979 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
14982 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
14983 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
14984 instruction set support.
14987 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
14988 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
14989 BMI, BMI2 and F16C instruction set support.
14992 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
14993 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
14994 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
14997 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
14998 instruction set support.
15001 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15002 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
15005 AMD K6 CPU with MMX instruction set support.
15009 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
15012 @itemx athlon-tbird
15013 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
15019 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
15020 instruction set support.
15026 Processors based on the AMD K8 core with x86-64 instruction set support,
15027 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
15028 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
15029 instruction set extensions.)
15032 @itemx opteron-sse3
15033 @itemx athlon64-sse3
15034 Improved versions of AMD K8 cores with SSE3 instruction set support.
15038 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
15039 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
15040 instruction set extensions.)
15043 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
15044 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
15045 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
15047 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15048 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
15049 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
15052 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15053 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
15054 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
15055 64-bit instruction set extensions.
15057 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15058 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
15059 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
15060 SSE4.2, ABM and 64-bit instruction set extensions.
15063 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
15064 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
15065 instruction set extensions.)
15068 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
15069 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
15070 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
15073 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
15077 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
15078 instruction set support.
15081 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
15082 implemented for this chip.)
15085 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
15087 implemented for this chip.)
15090 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
15093 @item -mtune=@var{cpu-type}
15095 Tune to @var{cpu-type} everything applicable about the generated code, except
15096 for the ABI and the set of available instructions.
15097 While picking a specific @var{cpu-type} schedules things appropriately
15098 for that particular chip, the compiler does not generate any code that
15099 cannot run on the default machine type unless you use a
15100 @option{-march=@var{cpu-type}} option.
15101 For example, if GCC is configured for i686-pc-linux-gnu
15102 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
15103 but still runs on i686 machines.
15105 The choices for @var{cpu-type} are the same as for @option{-march}.
15106 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
15110 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
15111 If you know the CPU on which your code will run, then you should use
15112 the corresponding @option{-mtune} or @option{-march} option instead of
15113 @option{-mtune=generic}. But, if you do not know exactly what CPU users
15114 of your application will have, then you should use this option.
15116 As new processors are deployed in the marketplace, the behavior of this
15117 option will change. Therefore, if you upgrade to a newer version of
15118 GCC, code generation controlled by this option will change to reflect
15120 that are most common at the time that version of GCC is released.
15122 There is no @option{-march=generic} option because @option{-march}
15123 indicates the instruction set the compiler can use, and there is no
15124 generic instruction set applicable to all processors. In contrast,
15125 @option{-mtune} indicates the processor (or, in this case, collection of
15126 processors) for which the code is optimized.
15129 Produce code optimized for the most current Intel processors, which are
15130 Haswell and Silvermont for this version of GCC. If you know the CPU
15131 on which your code will run, then you should use the corresponding
15132 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
15133 But, if you want your application performs better on both Haswell and
15134 Silvermont, then you should use this option.
15136 As new Intel processors are deployed in the marketplace, the behavior of
15137 this option will change. Therefore, if you upgrade to a newer version of
15138 GCC, code generation controlled by this option will change to reflect
15139 the most current Intel processors at the time that version of GCC is
15142 There is no @option{-march=intel} option because @option{-march} indicates
15143 the instruction set the compiler can use, and there is no common
15144 instruction set applicable to all processors. In contrast,
15145 @option{-mtune} indicates the processor (or, in this case, collection of
15146 processors) for which the code is optimized.
15149 @item -mcpu=@var{cpu-type}
15151 A deprecated synonym for @option{-mtune}.
15153 @item -mfpmath=@var{unit}
15155 Generate floating-point arithmetic for selected unit @var{unit}. The choices
15156 for @var{unit} are:
15160 Use the standard 387 floating-point coprocessor present on the majority of chips and
15161 emulated otherwise. Code compiled with this option runs almost everywhere.
15162 The temporary results are computed in 80-bit precision instead of the precision
15163 specified by the type, resulting in slightly different results compared to most
15164 of other chips. See @option{-ffloat-store} for more detailed description.
15166 This is the default choice for i386 compiler.
15169 Use scalar floating-point instructions present in the SSE instruction set.
15170 This instruction set is supported by Pentium III and newer chips,
15171 and in the AMD line
15172 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
15173 instruction set supports only single-precision arithmetic, thus the double and
15174 extended-precision arithmetic are still done using 387. A later version, present
15175 only in Pentium 4 and AMD x86-64 chips, supports double-precision
15178 For the i386 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
15179 or @option{-msse2} switches to enable SSE extensions and make this option
15180 effective. For the x86-64 compiler, these extensions are enabled by default.
15182 The resulting code should be considerably faster in the majority of cases and avoid
15183 the numerical instability problems of 387 code, but may break some existing
15184 code that expects temporaries to be 80 bits.
15186 This is the default choice for the x86-64 compiler.
15191 Attempt to utilize both instruction sets at once. This effectively doubles the
15192 amount of available registers, and on chips with separate execution units for
15193 387 and SSE the execution resources too. Use this option with care, as it is
15194 still experimental, because the GCC register allocator does not model separate
15195 functional units well, resulting in unstable performance.
15198 @item -masm=@var{dialect}
15199 @opindex masm=@var{dialect}
15200 Output assembly instructions using selected @var{dialect}. Supported
15201 choices are @samp{intel} or @samp{att} (the default). Darwin does
15202 not support @samp{intel}.
15205 @itemx -mno-ieee-fp
15207 @opindex mno-ieee-fp
15208 Control whether or not the compiler uses IEEE floating-point
15209 comparisons. These correctly handle the case where the result of a
15210 comparison is unordered.
15213 @opindex msoft-float
15214 Generate output containing library calls for floating point.
15216 @strong{Warning:} the requisite libraries are not part of GCC@.
15217 Normally the facilities of the machine's usual C compiler are used, but
15218 this can't be done directly in cross-compilation. You must make your
15219 own arrangements to provide suitable library functions for
15222 On machines where a function returns floating-point results in the 80387
15223 register stack, some floating-point opcodes may be emitted even if
15224 @option{-msoft-float} is used.
15226 @item -mno-fp-ret-in-387
15227 @opindex mno-fp-ret-in-387
15228 Do not use the FPU registers for return values of functions.
15230 The usual calling convention has functions return values of types
15231 @code{float} and @code{double} in an FPU register, even if there
15232 is no FPU@. The idea is that the operating system should emulate
15235 The option @option{-mno-fp-ret-in-387} causes such values to be returned
15236 in ordinary CPU registers instead.
15238 @item -mno-fancy-math-387
15239 @opindex mno-fancy-math-387
15240 Some 387 emulators do not support the @code{sin}, @code{cos} and
15241 @code{sqrt} instructions for the 387. Specify this option to avoid
15242 generating those instructions. This option is the default on FreeBSD,
15243 OpenBSD and NetBSD@. This option is overridden when @option{-march}
15244 indicates that the target CPU always has an FPU and so the
15245 instruction does not need emulation. These
15246 instructions are not generated unless you also use the
15247 @option{-funsafe-math-optimizations} switch.
15249 @item -malign-double
15250 @itemx -mno-align-double
15251 @opindex malign-double
15252 @opindex mno-align-double
15253 Control whether GCC aligns @code{double}, @code{long double}, and
15254 @code{long long} variables on a two-word boundary or a one-word
15255 boundary. Aligning @code{double} variables on a two-word boundary
15256 produces code that runs somewhat faster on a Pentium at the
15257 expense of more memory.
15259 On x86-64, @option{-malign-double} is enabled by default.
15261 @strong{Warning:} if you use the @option{-malign-double} switch,
15262 structures containing the above types are aligned differently than
15263 the published application binary interface specifications for the 386
15264 and are not binary compatible with structures in code compiled
15265 without that switch.
15267 @item -m96bit-long-double
15268 @itemx -m128bit-long-double
15269 @opindex m96bit-long-double
15270 @opindex m128bit-long-double
15271 These switches control the size of @code{long double} type. The i386
15272 application binary interface specifies the size to be 96 bits,
15273 so @option{-m96bit-long-double} is the default in 32-bit mode.
15275 Modern architectures (Pentium and newer) prefer @code{long double}
15276 to be aligned to an 8- or 16-byte boundary. In arrays or structures
15277 conforming to the ABI, this is not possible. So specifying
15278 @option{-m128bit-long-double} aligns @code{long double}
15279 to a 16-byte boundary by padding the @code{long double} with an additional
15282 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
15283 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
15285 Notice that neither of these options enable any extra precision over the x87
15286 standard of 80 bits for a @code{long double}.
15288 @strong{Warning:} if you override the default value for your target ABI, this
15289 changes the size of
15290 structures and arrays containing @code{long double} variables,
15291 as well as modifying the function calling convention for functions taking
15292 @code{long double}. Hence they are not binary-compatible
15293 with code compiled without that switch.
15295 @item -mlong-double-64
15296 @itemx -mlong-double-80
15297 @itemx -mlong-double-128
15298 @opindex mlong-double-64
15299 @opindex mlong-double-80
15300 @opindex mlong-double-128
15301 These switches control the size of @code{long double} type. A size
15302 of 64 bits makes the @code{long double} type equivalent to the @code{double}
15303 type. This is the default for 32-bit Bionic C library. A size
15304 of 128 bits makes the @code{long double} type equivalent to the
15305 @code{__float128} type. This is the default for 64-bit Bionic C library.
15307 @strong{Warning:} if you override the default value for your target ABI, this
15308 changes the size of
15309 structures and arrays containing @code{long double} variables,
15310 as well as modifying the function calling convention for functions taking
15311 @code{long double}. Hence they are not binary-compatible
15312 with code compiled without that switch.
15314 @item -mlarge-data-threshold=@var{threshold}
15315 @opindex mlarge-data-threshold
15316 When @option{-mcmodel=medium} is specified, data objects larger than
15317 @var{threshold} are placed in the large data section. This value must be the
15318 same across all objects linked into the binary, and defaults to 65535.
15322 Use a different function-calling convention, in which functions that
15323 take a fixed number of arguments return with the @code{ret @var{num}}
15324 instruction, which pops their arguments while returning. This saves one
15325 instruction in the caller since there is no need to pop the arguments
15328 You can specify that an individual function is called with this calling
15329 sequence with the function attribute @samp{stdcall}. You can also
15330 override the @option{-mrtd} option by using the function attribute
15331 @samp{cdecl}. @xref{Function Attributes}.
15333 @strong{Warning:} this calling convention is incompatible with the one
15334 normally used on Unix, so you cannot use it if you need to call
15335 libraries compiled with the Unix compiler.
15337 Also, you must provide function prototypes for all functions that
15338 take variable numbers of arguments (including @code{printf});
15339 otherwise incorrect code is generated for calls to those
15342 In addition, seriously incorrect code results if you call a
15343 function with too many arguments. (Normally, extra arguments are
15344 harmlessly ignored.)
15346 @item -mregparm=@var{num}
15348 Control how many registers are used to pass integer arguments. By
15349 default, no registers are used to pass arguments, and at most 3
15350 registers can be used. You can control this behavior for a specific
15351 function by using the function attribute @samp{regparm}.
15352 @xref{Function Attributes}.
15354 @strong{Warning:} if you use this switch, and
15355 @var{num} is nonzero, then you must build all modules with the same
15356 value, including any libraries. This includes the system libraries and
15360 @opindex msseregparm
15361 Use SSE register passing conventions for float and double arguments
15362 and return values. You can control this behavior for a specific
15363 function by using the function attribute @samp{sseregparm}.
15364 @xref{Function Attributes}.
15366 @strong{Warning:} if you use this switch then you must build all
15367 modules with the same value, including any libraries. This includes
15368 the system libraries and startup modules.
15370 @item -mvect8-ret-in-mem
15371 @opindex mvect8-ret-in-mem
15372 Return 8-byte vectors in memory instead of MMX registers. This is the
15373 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
15374 Studio compilers until version 12. Later compiler versions (starting
15375 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
15376 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
15377 you need to remain compatible with existing code produced by those
15378 previous compiler versions or older versions of GCC@.
15387 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
15388 is specified, the significands of results of floating-point operations are
15389 rounded to 24 bits (single precision); @option{-mpc64} rounds the
15390 significands of results of floating-point operations to 53 bits (double
15391 precision) and @option{-mpc80} rounds the significands of results of
15392 floating-point operations to 64 bits (extended double precision), which is
15393 the default. When this option is used, floating-point operations in higher
15394 precisions are not available to the programmer without setting the FPU
15395 control word explicitly.
15397 Setting the rounding of floating-point operations to less than the default
15398 80 bits can speed some programs by 2% or more. Note that some mathematical
15399 libraries assume that extended-precision (80-bit) floating-point operations
15400 are enabled by default; routines in such libraries could suffer significant
15401 loss of accuracy, typically through so-called ``catastrophic cancellation'',
15402 when this option is used to set the precision to less than extended precision.
15404 @item -mstackrealign
15405 @opindex mstackrealign
15406 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
15407 option generates an alternate prologue and epilogue that realigns the
15408 run-time stack if necessary. This supports mixing legacy codes that keep
15409 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
15410 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
15411 applicable to individual functions.
15413 @item -mpreferred-stack-boundary=@var{num}
15414 @opindex mpreferred-stack-boundary
15415 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
15416 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
15417 the default is 4 (16 bytes or 128 bits).
15419 @strong{Warning:} When generating code for the x86-64 architecture with
15420 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
15421 used to keep the stack boundary aligned to 8 byte boundary. Since
15422 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
15423 intended to be used in controlled environment where stack space is
15424 important limitation. This option will lead to wrong code when functions
15425 compiled with 16 byte stack alignment (such as functions from a standard
15426 library) are called with misaligned stack. In this case, SSE
15427 instructions may lead to misaligned memory access traps. In addition,
15428 variable arguments will be handled incorrectly for 16 byte aligned
15429 objects (including x87 long double and __int128), leading to wrong
15430 results. You must build all modules with
15431 @option{-mpreferred-stack-boundary=3}, including any libraries. This
15432 includes the system libraries and startup modules.
15434 @item -mincoming-stack-boundary=@var{num}
15435 @opindex mincoming-stack-boundary
15436 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
15437 boundary. If @option{-mincoming-stack-boundary} is not specified,
15438 the one specified by @option{-mpreferred-stack-boundary} is used.
15440 On Pentium and Pentium Pro, @code{double} and @code{long double} values
15441 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
15442 suffer significant run time performance penalties. On Pentium III, the
15443 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
15444 properly if it is not 16-byte aligned.
15446 To ensure proper alignment of this values on the stack, the stack boundary
15447 must be as aligned as that required by any value stored on the stack.
15448 Further, every function must be generated such that it keeps the stack
15449 aligned. Thus calling a function compiled with a higher preferred
15450 stack boundary from a function compiled with a lower preferred stack
15451 boundary most likely misaligns the stack. It is recommended that
15452 libraries that use callbacks always use the default setting.
15454 This extra alignment does consume extra stack space, and generally
15455 increases code size. Code that is sensitive to stack space usage, such
15456 as embedded systems and operating system kernels, may want to reduce the
15457 preferred alignment to @option{-mpreferred-stack-boundary=2}.
15481 @itemx -mno-avx512f
15484 @itemx -mno-avx512pf
15486 @itemx -mno-avx512er
15488 @itemx -mno-avx512cd
15496 @itemx -mno-clflsuhopt
15499 @itemx -mno-fsgsbase
15506 @itemx -mprefetchwt1
15507 @itemx -mno-prefetchwt1
15545 These switches enable or disable the use of instructions in the MMX, SSE,
15546 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
15547 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
15548 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, or 3DNow!@:
15549 extended instruction sets.
15550 These extensions are also available as built-in functions: see
15551 @ref{X86 Built-in Functions}, for details of the functions enabled and
15552 disabled by these switches.
15554 To generate SSE/SSE2 instructions automatically from floating-point
15555 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
15557 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
15558 generates new AVX instructions or AVX equivalence for all SSEx instructions
15561 These options enable GCC to use these extended instructions in
15562 generated code, even without @option{-mfpmath=sse}. Applications that
15563 perform run-time CPU detection must compile separate files for each
15564 supported architecture, using the appropriate flags. In particular,
15565 the file containing the CPU detection code should be compiled without
15568 @item -mdump-tune-features
15569 @opindex mdump-tune-features
15570 This option instructs GCC to dump the names of the x86 performance
15571 tuning features and default settings. The names can be used in
15572 @option{-mtune-ctrl=@var{feature-list}}.
15574 @item -mtune-ctrl=@var{feature-list}
15575 @opindex mtune-ctrl=@var{feature-list}
15576 This option is used to do fine grain control of x86 code generation features.
15577 @var{feature-list} is a comma separated list of @var{feature} names. See also
15578 @option{-mdump-tune-features}. When specified, the @var{feature} will be turned
15579 on if it is not preceded with @code{^}, otherwise, it will be turned off.
15580 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
15581 developers. Using it may lead to code paths not covered by testing and can
15582 potentially result in compiler ICEs or runtime errors.
15585 @opindex mno-default
15586 This option instructs GCC to turn off all tunable features. See also
15587 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
15591 This option instructs GCC to emit a @code{cld} instruction in the prologue
15592 of functions that use string instructions. String instructions depend on
15593 the DF flag to select between autoincrement or autodecrement mode. While the
15594 ABI specifies the DF flag to be cleared on function entry, some operating
15595 systems violate this specification by not clearing the DF flag in their
15596 exception dispatchers. The exception handler can be invoked with the DF flag
15597 set, which leads to wrong direction mode when string instructions are used.
15598 This option can be enabled by default on 32-bit x86 targets by configuring
15599 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
15600 instructions can be suppressed with the @option{-mno-cld} compiler option
15604 @opindex mvzeroupper
15605 This option instructs GCC to emit a @code{vzeroupper} instruction
15606 before a transfer of control flow out of the function to minimize
15607 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
15610 @item -mprefer-avx128
15611 @opindex mprefer-avx128
15612 This option instructs GCC to use 128-bit AVX instructions instead of
15613 256-bit AVX instructions in the auto-vectorizer.
15617 This option enables GCC to generate @code{CMPXCHG16B} instructions.
15618 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
15619 (or oword) data types.
15620 This is useful for high-resolution counters that can be updated
15621 by multiple processors (or cores). This instruction is generated as part of
15622 atomic built-in functions: see @ref{__sync Builtins} or
15623 @ref{__atomic Builtins} for details.
15627 This option enables generation of @code{SAHF} instructions in 64-bit code.
15628 Early Intel Pentium 4 CPUs with Intel 64 support,
15629 prior to the introduction of Pentium 4 G1 step in December 2005,
15630 lacked the @code{LAHF} and @code{SAHF} instructions
15631 which were supported by AMD64.
15632 These are load and store instructions, respectively, for certain status flags.
15633 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
15634 @code{drem}, and @code{remainder} built-in functions;
15635 see @ref{Other Builtins} for details.
15639 This option enables use of the @code{movbe} instruction to implement
15640 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
15644 This option enables built-in functions @code{__builtin_ia32_crc32qi},
15645 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
15646 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
15650 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
15651 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
15652 with an additional Newton-Raphson step
15653 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
15654 (and their vectorized
15655 variants) for single-precision floating-point arguments. These instructions
15656 are generated only when @option{-funsafe-math-optimizations} is enabled
15657 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
15658 Note that while the throughput of the sequence is higher than the throughput
15659 of the non-reciprocal instruction, the precision of the sequence can be
15660 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
15662 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
15663 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
15664 combination), and doesn't need @option{-mrecip}.
15666 Also note that GCC emits the above sequence with additional Newton-Raphson step
15667 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
15668 already with @option{-ffast-math} (or the above option combination), and
15669 doesn't need @option{-mrecip}.
15671 @item -mrecip=@var{opt}
15672 @opindex mrecip=opt
15673 This option controls which reciprocal estimate instructions
15674 may be used. @var{opt} is a comma-separated list of options, which may
15675 be preceded by a @samp{!} to invert the option:
15679 Enable all estimate instructions.
15682 Enable the default instructions, equivalent to @option{-mrecip}.
15685 Disable all estimate instructions, equivalent to @option{-mno-recip}.
15688 Enable the approximation for scalar division.
15691 Enable the approximation for vectorized division.
15694 Enable the approximation for scalar square root.
15697 Enable the approximation for vectorized square root.
15700 So, for example, @option{-mrecip=all,!sqrt} enables
15701 all of the reciprocal approximations, except for square root.
15703 @item -mveclibabi=@var{type}
15704 @opindex mveclibabi
15705 Specifies the ABI type to use for vectorizing intrinsics using an
15706 external library. Supported values for @var{type} are @samp{svml}
15707 for the Intel short
15708 vector math library and @samp{acml} for the AMD math core library.
15709 To use this option, both @option{-ftree-vectorize} and
15710 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
15711 ABI-compatible library must be specified at link time.
15713 GCC currently emits calls to @code{vmldExp2},
15714 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
15715 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
15716 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
15717 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
15718 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
15719 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
15720 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
15721 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
15722 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
15723 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
15724 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
15725 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
15726 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
15727 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
15728 when @option{-mveclibabi=acml} is used.
15730 @item -mabi=@var{name}
15732 Generate code for the specified calling convention. Permissible values
15733 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
15734 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
15735 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
15736 You can control this behavior for a specific function by
15737 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
15738 @xref{Function Attributes}.
15740 @item -mtls-dialect=@var{type}
15741 @opindex mtls-dialect
15742 Generate code to access thread-local storage using the @samp{gnu} or
15743 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
15744 @samp{gnu2} is more efficient, but it may add compile- and run-time
15745 requirements that cannot be satisfied on all systems.
15748 @itemx -mno-push-args
15749 @opindex mpush-args
15750 @opindex mno-push-args
15751 Use PUSH operations to store outgoing parameters. This method is shorter
15752 and usually equally fast as method using SUB/MOV operations and is enabled
15753 by default. In some cases disabling it may improve performance because of
15754 improved scheduling and reduced dependencies.
15756 @item -maccumulate-outgoing-args
15757 @opindex maccumulate-outgoing-args
15758 If enabled, the maximum amount of space required for outgoing arguments is
15759 computed in the function prologue. This is faster on most modern CPUs
15760 because of reduced dependencies, improved scheduling and reduced stack usage
15761 when the preferred stack boundary is not equal to 2. The drawback is a notable
15762 increase in code size. This switch implies @option{-mno-push-args}.
15766 Support thread-safe exception handling on MinGW. Programs that rely
15767 on thread-safe exception handling must compile and link all code with the
15768 @option{-mthreads} option. When compiling, @option{-mthreads} defines
15769 @code{-D_MT}; when linking, it links in a special thread helper library
15770 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
15772 @item -mno-align-stringops
15773 @opindex mno-align-stringops
15774 Do not align the destination of inlined string operations. This switch reduces
15775 code size and improves performance in case the destination is already aligned,
15776 but GCC doesn't know about it.
15778 @item -minline-all-stringops
15779 @opindex minline-all-stringops
15780 By default GCC inlines string operations only when the destination is
15781 known to be aligned to least a 4-byte boundary.
15782 This enables more inlining and increases code
15783 size, but may improve performance of code that depends on fast
15784 @code{memcpy}, @code{strlen},
15785 and @code{memset} for short lengths.
15787 @item -minline-stringops-dynamically
15788 @opindex minline-stringops-dynamically
15789 For string operations of unknown size, use run-time checks with
15790 inline code for small blocks and a library call for large blocks.
15792 @item -mstringop-strategy=@var{alg}
15793 @opindex mstringop-strategy=@var{alg}
15794 Override the internal decision heuristic for the particular algorithm to use
15795 for inlining string operations. The allowed values for @var{alg} are:
15801 Expand using i386 @code{rep} prefix of the specified size.
15805 @itemx unrolled_loop
15806 Expand into an inline loop.
15809 Always use a library call.
15812 @item -mmemcpy-strategy=@var{strategy}
15813 @opindex mmemcpy-strategy=@var{strategy}
15814 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
15815 should be inlined and what inline algorithm to use when the expected size
15816 of the copy operation is known. @var{strategy}
15817 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
15818 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
15819 the max byte size with which inline algorithm @var{alg} is allowed. For the last
15820 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
15821 in the list must be specified in increasing order. The minimal byte size for
15822 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
15825 @item -mmemset-strategy=@var{strategy}
15826 @opindex mmemset-strategy=@var{strategy}
15827 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
15828 @code{__builtin_memset} expansion.
15830 @item -momit-leaf-frame-pointer
15831 @opindex momit-leaf-frame-pointer
15832 Don't keep the frame pointer in a register for leaf functions. This
15833 avoids the instructions to save, set up, and restore frame pointers and
15834 makes an extra register available in leaf functions. The option
15835 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
15836 which might make debugging harder.
15838 @item -mtls-direct-seg-refs
15839 @itemx -mno-tls-direct-seg-refs
15840 @opindex mtls-direct-seg-refs
15841 Controls whether TLS variables may be accessed with offsets from the
15842 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
15843 or whether the thread base pointer must be added. Whether or not this
15844 is valid depends on the operating system, and whether it maps the
15845 segment to cover the entire TLS area.
15847 For systems that use the GNU C Library, the default is on.
15850 @itemx -mno-sse2avx
15852 Specify that the assembler should encode SSE instructions with VEX
15853 prefix. The option @option{-mavx} turns this on by default.
15858 If profiling is active (@option{-pg}), put the profiling
15859 counter call before the prologue.
15860 Note: On x86 architectures the attribute @code{ms_hook_prologue}
15861 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
15864 @itemx -mno-8bit-idiv
15866 On some processors, like Intel Atom, 8-bit unsigned integer divide is
15867 much faster than 32-bit/64-bit integer divide. This option generates a
15868 run-time check. If both dividend and divisor are within range of 0
15869 to 255, 8-bit unsigned integer divide is used instead of
15870 32-bit/64-bit integer divide.
15872 @item -mavx256-split-unaligned-load
15873 @itemx -mavx256-split-unaligned-store
15874 @opindex avx256-split-unaligned-load
15875 @opindex avx256-split-unaligned-store
15876 Split 32-byte AVX unaligned load and store.
15878 @item -mstack-protector-guard=@var{guard}
15879 @opindex mstack-protector-guard=@var{guard}
15880 Generate stack protection code using canary at @var{guard}. Supported
15881 locations are @samp{global} for global canary or @samp{tls} for per-thread
15882 canary in the TLS block (the default). This option has effect only when
15883 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
15887 These @samp{-m} switches are supported in addition to the above
15888 on x86-64 processors in 64-bit environments.
15899 Generate code for a 16-bit, 32-bit or 64-bit environment.
15900 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
15902 generates code that runs on any i386 system.
15904 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
15905 types to 64 bits, and generates code for the x86-64 architecture.
15906 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
15907 and @option{-mdynamic-no-pic} options.
15909 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
15911 generates code for the x86-64 architecture.
15913 The @option{-m16} option is the same as @option{-m32}, except for that
15914 it outputs the @code{.code16gcc} assembly directive at the beginning of
15915 the assembly output so that the binary can run in 16-bit mode.
15917 @item -mno-red-zone
15918 @opindex mno-red-zone
15919 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
15920 by the x86-64 ABI; it is a 128-byte area beyond the location of the
15921 stack pointer that is not modified by signal or interrupt handlers
15922 and therefore can be used for temporary data without adjusting the stack
15923 pointer. The flag @option{-mno-red-zone} disables this red zone.
15925 @item -mcmodel=small
15926 @opindex mcmodel=small
15927 Generate code for the small code model: the program and its symbols must
15928 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
15929 Programs can be statically or dynamically linked. This is the default
15932 @item -mcmodel=kernel
15933 @opindex mcmodel=kernel
15934 Generate code for the kernel code model. The kernel runs in the
15935 negative 2 GB of the address space.
15936 This model has to be used for Linux kernel code.
15938 @item -mcmodel=medium
15939 @opindex mcmodel=medium
15940 Generate code for the medium model: the program is linked in the lower 2
15941 GB of the address space. Small symbols are also placed there. Symbols
15942 with sizes larger than @option{-mlarge-data-threshold} are put into
15943 large data or BSS sections and can be located above 2GB. Programs can
15944 be statically or dynamically linked.
15946 @item -mcmodel=large
15947 @opindex mcmodel=large
15948 Generate code for the large model. This model makes no assumptions
15949 about addresses and sizes of sections.
15951 @item -maddress-mode=long
15952 @opindex maddress-mode=long
15953 Generate code for long address mode. This is only supported for 64-bit
15954 and x32 environments. It is the default address mode for 64-bit
15957 @item -maddress-mode=short
15958 @opindex maddress-mode=short
15959 Generate code for short address mode. This is only supported for 32-bit
15960 and x32 environments. It is the default address mode for 32-bit and
15964 @node i386 and x86-64 Windows Options
15965 @subsection i386 and x86-64 Windows Options
15966 @cindex i386 and x86-64 Windows Options
15968 These additional options are available for Microsoft Windows targets:
15974 specifies that a console application is to be generated, by
15975 instructing the linker to set the PE header subsystem type
15976 required for console applications.
15977 This option is available for Cygwin and MinGW targets and is
15978 enabled by default on those targets.
15982 This option is available for Cygwin and MinGW targets. It
15983 specifies that a DLL---a dynamic link library---is to be
15984 generated, enabling the selection of the required runtime
15985 startup object and entry point.
15987 @item -mnop-fun-dllimport
15988 @opindex mnop-fun-dllimport
15989 This option is available for Cygwin and MinGW targets. It
15990 specifies that the @code{dllimport} attribute should be ignored.
15994 This option is available for MinGW targets. It specifies
15995 that MinGW-specific thread support is to be used.
15999 This option is available for MinGW-w64 targets. It causes
16000 the @code{UNICODE} preprocessor macro to be predefined, and
16001 chooses Unicode-capable runtime startup code.
16005 This option is available for Cygwin and MinGW targets. It
16006 specifies that the typical Microsoft Windows predefined macros are to
16007 be set in the pre-processor, but does not influence the choice
16008 of runtime library/startup code.
16012 This option is available for Cygwin and MinGW targets. It
16013 specifies that a GUI application is to be generated by
16014 instructing the linker to set the PE header subsystem type
16017 @item -fno-set-stack-executable
16018 @opindex fno-set-stack-executable
16019 This option is available for MinGW targets. It specifies that
16020 the executable flag for the stack used by nested functions isn't
16021 set. This is necessary for binaries running in kernel mode of
16022 Microsoft Windows, as there the User32 API, which is used to set executable
16023 privileges, isn't available.
16025 @item -fwritable-relocated-rdata
16026 @opindex fno-writable-relocated-rdata
16027 This option is available for MinGW and Cygwin targets. It specifies
16028 that relocated-data in read-only section is put into .data
16029 section. This is a necessary for older runtimes not supporting
16030 modification of .rdata sections for pseudo-relocation.
16032 @item -mpe-aligned-commons
16033 @opindex mpe-aligned-commons
16034 This option is available for Cygwin and MinGW targets. It
16035 specifies that the GNU extension to the PE file format that
16036 permits the correct alignment of COMMON variables should be
16037 used when generating code. It is enabled by default if
16038 GCC detects that the target assembler found during configuration
16039 supports the feature.
16042 See also under @ref{i386 and x86-64 Options} for standard options.
16044 @node IA-64 Options
16045 @subsection IA-64 Options
16046 @cindex IA-64 Options
16048 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16052 @opindex mbig-endian
16053 Generate code for a big-endian target. This is the default for HP-UX@.
16055 @item -mlittle-endian
16056 @opindex mlittle-endian
16057 Generate code for a little-endian target. This is the default for AIX5
16063 @opindex mno-gnu-as
16064 Generate (or don't) code for the GNU assembler. This is the default.
16065 @c Also, this is the default if the configure option @option{--with-gnu-as}
16071 @opindex mno-gnu-ld
16072 Generate (or don't) code for the GNU linker. This is the default.
16073 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16078 Generate code that does not use a global pointer register. The result
16079 is not position independent code, and violates the IA-64 ABI@.
16081 @item -mvolatile-asm-stop
16082 @itemx -mno-volatile-asm-stop
16083 @opindex mvolatile-asm-stop
16084 @opindex mno-volatile-asm-stop
16085 Generate (or don't) a stop bit immediately before and after volatile asm
16088 @item -mregister-names
16089 @itemx -mno-register-names
16090 @opindex mregister-names
16091 @opindex mno-register-names
16092 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16093 the stacked registers. This may make assembler output more readable.
16099 Disable (or enable) optimizations that use the small data section. This may
16100 be useful for working around optimizer bugs.
16102 @item -mconstant-gp
16103 @opindex mconstant-gp
16104 Generate code that uses a single constant global pointer value. This is
16105 useful when compiling kernel code.
16109 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16110 This is useful when compiling firmware code.
16112 @item -minline-float-divide-min-latency
16113 @opindex minline-float-divide-min-latency
16114 Generate code for inline divides of floating-point values
16115 using the minimum latency algorithm.
16117 @item -minline-float-divide-max-throughput
16118 @opindex minline-float-divide-max-throughput
16119 Generate code for inline divides of floating-point values
16120 using the maximum throughput algorithm.
16122 @item -mno-inline-float-divide
16123 @opindex mno-inline-float-divide
16124 Do not generate inline code for divides of floating-point values.
16126 @item -minline-int-divide-min-latency
16127 @opindex minline-int-divide-min-latency
16128 Generate code for inline divides of integer values
16129 using the minimum latency algorithm.
16131 @item -minline-int-divide-max-throughput
16132 @opindex minline-int-divide-max-throughput
16133 Generate code for inline divides of integer values
16134 using the maximum throughput algorithm.
16136 @item -mno-inline-int-divide
16137 @opindex mno-inline-int-divide
16138 Do not generate inline code for divides of integer values.
16140 @item -minline-sqrt-min-latency
16141 @opindex minline-sqrt-min-latency
16142 Generate code for inline square roots
16143 using the minimum latency algorithm.
16145 @item -minline-sqrt-max-throughput
16146 @opindex minline-sqrt-max-throughput
16147 Generate code for inline square roots
16148 using the maximum throughput algorithm.
16150 @item -mno-inline-sqrt
16151 @opindex mno-inline-sqrt
16152 Do not generate inline code for @code{sqrt}.
16155 @itemx -mno-fused-madd
16156 @opindex mfused-madd
16157 @opindex mno-fused-madd
16158 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16159 instructions. The default is to use these instructions.
16161 @item -mno-dwarf2-asm
16162 @itemx -mdwarf2-asm
16163 @opindex mno-dwarf2-asm
16164 @opindex mdwarf2-asm
16165 Don't (or do) generate assembler code for the DWARF 2 line number debugging
16166 info. This may be useful when not using the GNU assembler.
16168 @item -mearly-stop-bits
16169 @itemx -mno-early-stop-bits
16170 @opindex mearly-stop-bits
16171 @opindex mno-early-stop-bits
16172 Allow stop bits to be placed earlier than immediately preceding the
16173 instruction that triggered the stop bit. This can improve instruction
16174 scheduling, but does not always do so.
16176 @item -mfixed-range=@var{register-range}
16177 @opindex mfixed-range
16178 Generate code treating the given register range as fixed registers.
16179 A fixed register is one that the register allocator cannot use. This is
16180 useful when compiling kernel code. A register range is specified as
16181 two registers separated by a dash. Multiple register ranges can be
16182 specified separated by a comma.
16184 @item -mtls-size=@var{tls-size}
16186 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16189 @item -mtune=@var{cpu-type}
16191 Tune the instruction scheduling for a particular CPU, Valid values are
16192 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16193 and @samp{mckinley}.
16199 Generate code for a 32-bit or 64-bit environment.
16200 The 32-bit environment sets int, long and pointer to 32 bits.
16201 The 64-bit environment sets int to 32 bits and long and pointer
16202 to 64 bits. These are HP-UX specific flags.
16204 @item -mno-sched-br-data-spec
16205 @itemx -msched-br-data-spec
16206 @opindex mno-sched-br-data-spec
16207 @opindex msched-br-data-spec
16208 (Dis/En)able data speculative scheduling before reload.
16209 This results in generation of @code{ld.a} instructions and
16210 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16211 The default is 'disable'.
16213 @item -msched-ar-data-spec
16214 @itemx -mno-sched-ar-data-spec
16215 @opindex msched-ar-data-spec
16216 @opindex mno-sched-ar-data-spec
16217 (En/Dis)able data speculative scheduling after reload.
16218 This results in generation of @code{ld.a} instructions and
16219 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16220 The default is 'enable'.
16222 @item -mno-sched-control-spec
16223 @itemx -msched-control-spec
16224 @opindex mno-sched-control-spec
16225 @opindex msched-control-spec
16226 (Dis/En)able control speculative scheduling. This feature is
16227 available only during region scheduling (i.e.@: before reload).
16228 This results in generation of the @code{ld.s} instructions and
16229 the corresponding check instructions @code{chk.s}.
16230 The default is 'disable'.
16232 @item -msched-br-in-data-spec
16233 @itemx -mno-sched-br-in-data-spec
16234 @opindex msched-br-in-data-spec
16235 @opindex mno-sched-br-in-data-spec
16236 (En/Dis)able speculative scheduling of the instructions that
16237 are dependent on the data speculative loads before reload.
16238 This is effective only with @option{-msched-br-data-spec} enabled.
16239 The default is 'enable'.
16241 @item -msched-ar-in-data-spec
16242 @itemx -mno-sched-ar-in-data-spec
16243 @opindex msched-ar-in-data-spec
16244 @opindex mno-sched-ar-in-data-spec
16245 (En/Dis)able speculative scheduling of the instructions that
16246 are dependent on the data speculative loads after reload.
16247 This is effective only with @option{-msched-ar-data-spec} enabled.
16248 The default is 'enable'.
16250 @item -msched-in-control-spec
16251 @itemx -mno-sched-in-control-spec
16252 @opindex msched-in-control-spec
16253 @opindex mno-sched-in-control-spec
16254 (En/Dis)able speculative scheduling of the instructions that
16255 are dependent on the control speculative loads.
16256 This is effective only with @option{-msched-control-spec} enabled.
16257 The default is 'enable'.
16259 @item -mno-sched-prefer-non-data-spec-insns
16260 @itemx -msched-prefer-non-data-spec-insns
16261 @opindex mno-sched-prefer-non-data-spec-insns
16262 @opindex msched-prefer-non-data-spec-insns
16263 If enabled, data-speculative instructions are chosen for schedule
16264 only if there are no other choices at the moment. This makes
16265 the use of the data speculation much more conservative.
16266 The default is 'disable'.
16268 @item -mno-sched-prefer-non-control-spec-insns
16269 @itemx -msched-prefer-non-control-spec-insns
16270 @opindex mno-sched-prefer-non-control-spec-insns
16271 @opindex msched-prefer-non-control-spec-insns
16272 If enabled, control-speculative instructions are chosen for schedule
16273 only if there are no other choices at the moment. This makes
16274 the use of the control speculation much more conservative.
16275 The default is 'disable'.
16277 @item -mno-sched-count-spec-in-critical-path
16278 @itemx -msched-count-spec-in-critical-path
16279 @opindex mno-sched-count-spec-in-critical-path
16280 @opindex msched-count-spec-in-critical-path
16281 If enabled, speculative dependencies are considered during
16282 computation of the instructions priorities. This makes the use of the
16283 speculation a bit more conservative.
16284 The default is 'disable'.
16286 @item -msched-spec-ldc
16287 @opindex msched-spec-ldc
16288 Use a simple data speculation check. This option is on by default.
16290 @item -msched-control-spec-ldc
16291 @opindex msched-spec-ldc
16292 Use a simple check for control speculation. This option is on by default.
16294 @item -msched-stop-bits-after-every-cycle
16295 @opindex msched-stop-bits-after-every-cycle
16296 Place a stop bit after every cycle when scheduling. This option is on
16299 @item -msched-fp-mem-deps-zero-cost
16300 @opindex msched-fp-mem-deps-zero-cost
16301 Assume that floating-point stores and loads are not likely to cause a conflict
16302 when placed into the same instruction group. This option is disabled by
16305 @item -msel-sched-dont-check-control-spec
16306 @opindex msel-sched-dont-check-control-spec
16307 Generate checks for control speculation in selective scheduling.
16308 This flag is disabled by default.
16310 @item -msched-max-memory-insns=@var{max-insns}
16311 @opindex msched-max-memory-insns
16312 Limit on the number of memory insns per instruction group, giving lower
16313 priority to subsequent memory insns attempting to schedule in the same
16314 instruction group. Frequently useful to prevent cache bank conflicts.
16315 The default value is 1.
16317 @item -msched-max-memory-insns-hard-limit
16318 @opindex msched-max-memory-insns-hard-limit
16319 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16320 disallowing more than that number in an instruction group.
16321 Otherwise, the limit is ``soft'', meaning that non-memory operations
16322 are preferred when the limit is reached, but memory operations may still
16328 @subsection LM32 Options
16329 @cindex LM32 options
16331 These @option{-m} options are defined for the LatticeMico32 architecture:
16334 @item -mbarrel-shift-enabled
16335 @opindex mbarrel-shift-enabled
16336 Enable barrel-shift instructions.
16338 @item -mdivide-enabled
16339 @opindex mdivide-enabled
16340 Enable divide and modulus instructions.
16342 @item -mmultiply-enabled
16343 @opindex multiply-enabled
16344 Enable multiply instructions.
16346 @item -msign-extend-enabled
16347 @opindex msign-extend-enabled
16348 Enable sign extend instructions.
16350 @item -muser-enabled
16351 @opindex muser-enabled
16352 Enable user-defined instructions.
16357 @subsection M32C Options
16358 @cindex M32C options
16361 @item -mcpu=@var{name}
16363 Select the CPU for which code is generated. @var{name} may be one of
16364 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16365 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16366 the M32C/80 series.
16370 Specifies that the program will be run on the simulator. This causes
16371 an alternate runtime library to be linked in which supports, for
16372 example, file I/O@. You must not use this option when generating
16373 programs that will run on real hardware; you must provide your own
16374 runtime library for whatever I/O functions are needed.
16376 @item -memregs=@var{number}
16378 Specifies the number of memory-based pseudo-registers GCC uses
16379 during code generation. These pseudo-registers are used like real
16380 registers, so there is a tradeoff between GCC's ability to fit the
16381 code into available registers, and the performance penalty of using
16382 memory instead of registers. Note that all modules in a program must
16383 be compiled with the same value for this option. Because of that, you
16384 must not use this option with GCC's default runtime libraries.
16388 @node M32R/D Options
16389 @subsection M32R/D Options
16390 @cindex M32R/D options
16392 These @option{-m} options are defined for Renesas M32R/D architectures:
16397 Generate code for the M32R/2@.
16401 Generate code for the M32R/X@.
16405 Generate code for the M32R@. This is the default.
16407 @item -mmodel=small
16408 @opindex mmodel=small
16409 Assume all objects live in the lower 16MB of memory (so that their addresses
16410 can be loaded with the @code{ld24} instruction), and assume all subroutines
16411 are reachable with the @code{bl} instruction.
16412 This is the default.
16414 The addressability of a particular object can be set with the
16415 @code{model} attribute.
16417 @item -mmodel=medium
16418 @opindex mmodel=medium
16419 Assume objects may be anywhere in the 32-bit address space (the compiler
16420 generates @code{seth/add3} instructions to load their addresses), and
16421 assume all subroutines are reachable with the @code{bl} instruction.
16423 @item -mmodel=large
16424 @opindex mmodel=large
16425 Assume objects may be anywhere in the 32-bit address space (the compiler
16426 generates @code{seth/add3} instructions to load their addresses), and
16427 assume subroutines may not be reachable with the @code{bl} instruction
16428 (the compiler generates the much slower @code{seth/add3/jl}
16429 instruction sequence).
16432 @opindex msdata=none
16433 Disable use of the small data area. Variables are put into
16434 one of @samp{.data}, @samp{.bss}, or @samp{.rodata} (unless the
16435 @code{section} attribute has been specified).
16436 This is the default.
16438 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
16439 Objects may be explicitly put in the small data area with the
16440 @code{section} attribute using one of these sections.
16442 @item -msdata=sdata
16443 @opindex msdata=sdata
16444 Put small global and static data in the small data area, but do not
16445 generate special code to reference them.
16448 @opindex msdata=use
16449 Put small global and static data in the small data area, and generate
16450 special instructions to reference them.
16454 @cindex smaller data references
16455 Put global and static objects less than or equal to @var{num} bytes
16456 into the small data or BSS sections instead of the normal data or BSS
16457 sections. The default value of @var{num} is 8.
16458 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16459 for this option to have any effect.
16461 All modules should be compiled with the same @option{-G @var{num}} value.
16462 Compiling with different values of @var{num} may or may not work; if it
16463 doesn't the linker gives an error message---incorrect code is not
16468 Makes the M32R-specific code in the compiler display some statistics
16469 that might help in debugging programs.
16471 @item -malign-loops
16472 @opindex malign-loops
16473 Align all loops to a 32-byte boundary.
16475 @item -mno-align-loops
16476 @opindex mno-align-loops
16477 Do not enforce a 32-byte alignment for loops. This is the default.
16479 @item -missue-rate=@var{number}
16480 @opindex missue-rate=@var{number}
16481 Issue @var{number} instructions per cycle. @var{number} can only be 1
16484 @item -mbranch-cost=@var{number}
16485 @opindex mbranch-cost=@var{number}
16486 @var{number} can only be 1 or 2. If it is 1 then branches are
16487 preferred over conditional code, if it is 2, then the opposite applies.
16489 @item -mflush-trap=@var{number}
16490 @opindex mflush-trap=@var{number}
16491 Specifies the trap number to use to flush the cache. The default is
16492 12. Valid numbers are between 0 and 15 inclusive.
16494 @item -mno-flush-trap
16495 @opindex mno-flush-trap
16496 Specifies that the cache cannot be flushed by using a trap.
16498 @item -mflush-func=@var{name}
16499 @opindex mflush-func=@var{name}
16500 Specifies the name of the operating system function to call to flush
16501 the cache. The default is @emph{_flush_cache}, but a function call
16502 is only used if a trap is not available.
16504 @item -mno-flush-func
16505 @opindex mno-flush-func
16506 Indicates that there is no OS function for flushing the cache.
16510 @node M680x0 Options
16511 @subsection M680x0 Options
16512 @cindex M680x0 options
16514 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16515 The default settings depend on which architecture was selected when
16516 the compiler was configured; the defaults for the most common choices
16520 @item -march=@var{arch}
16522 Generate code for a specific M680x0 or ColdFire instruction set
16523 architecture. Permissible values of @var{arch} for M680x0
16524 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16525 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16526 architectures are selected according to Freescale's ISA classification
16527 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16528 @samp{isab} and @samp{isac}.
16530 GCC defines a macro @samp{__mcf@var{arch}__} whenever it is generating
16531 code for a ColdFire target. The @var{arch} in this macro is one of the
16532 @option{-march} arguments given above.
16534 When used together, @option{-march} and @option{-mtune} select code
16535 that runs on a family of similar processors but that is optimized
16536 for a particular microarchitecture.
16538 @item -mcpu=@var{cpu}
16540 Generate code for a specific M680x0 or ColdFire processor.
16541 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16542 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16543 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16544 below, which also classifies the CPUs into families:
16546 @multitable @columnfractions 0.20 0.80
16547 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16548 @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}
16549 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16550 @item @samp{5206e} @tab @samp{5206e}
16551 @item @samp{5208} @tab @samp{5207} @samp{5208}
16552 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16553 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16554 @item @samp{5216} @tab @samp{5214} @samp{5216}
16555 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16556 @item @samp{5225} @tab @samp{5224} @samp{5225}
16557 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16558 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16559 @item @samp{5249} @tab @samp{5249}
16560 @item @samp{5250} @tab @samp{5250}
16561 @item @samp{5271} @tab @samp{5270} @samp{5271}
16562 @item @samp{5272} @tab @samp{5272}
16563 @item @samp{5275} @tab @samp{5274} @samp{5275}
16564 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16565 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16566 @item @samp{5307} @tab @samp{5307}
16567 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16568 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16569 @item @samp{5407} @tab @samp{5407}
16570 @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}
16573 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16574 @var{arch} is compatible with @var{cpu}. Other combinations of
16575 @option{-mcpu} and @option{-march} are rejected.
16577 GCC defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
16578 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
16579 where the value of @var{family} is given by the table above.
16581 @item -mtune=@var{tune}
16583 Tune the code for a particular microarchitecture within the
16584 constraints set by @option{-march} and @option{-mcpu}.
16585 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16586 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16587 and @samp{cpu32}. The ColdFire microarchitectures
16588 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16590 You can also use @option{-mtune=68020-40} for code that needs
16591 to run relatively well on 68020, 68030 and 68040 targets.
16592 @option{-mtune=68020-60} is similar but includes 68060 targets
16593 as well. These two options select the same tuning decisions as
16594 @option{-m68020-40} and @option{-m68020-60} respectively.
16596 GCC defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
16597 when tuning for 680x0 architecture @var{arch}. It also defines
16598 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16599 option is used. If GCC is tuning for a range of architectures,
16600 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16601 it defines the macros for every architecture in the range.
16603 GCC also defines the macro @samp{__m@var{uarch}__} when tuning for
16604 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16605 of the arguments given above.
16611 Generate output for a 68000. This is the default
16612 when the compiler is configured for 68000-based systems.
16613 It is equivalent to @option{-march=68000}.
16615 Use this option for microcontrollers with a 68000 or EC000 core,
16616 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16620 Generate output for a 68010. This is the default
16621 when the compiler is configured for 68010-based systems.
16622 It is equivalent to @option{-march=68010}.
16628 Generate output for a 68020. This is the default
16629 when the compiler is configured for 68020-based systems.
16630 It is equivalent to @option{-march=68020}.
16634 Generate output for a 68030. This is the default when the compiler is
16635 configured for 68030-based systems. It is equivalent to
16636 @option{-march=68030}.
16640 Generate output for a 68040. This is the default when the compiler is
16641 configured for 68040-based systems. It is equivalent to
16642 @option{-march=68040}.
16644 This option inhibits the use of 68881/68882 instructions that have to be
16645 emulated by software on the 68040. Use this option if your 68040 does not
16646 have code to emulate those instructions.
16650 Generate output for a 68060. This is the default when the compiler is
16651 configured for 68060-based systems. It is equivalent to
16652 @option{-march=68060}.
16654 This option inhibits the use of 68020 and 68881/68882 instructions that
16655 have to be emulated by software on the 68060. Use this option if your 68060
16656 does not have code to emulate those instructions.
16660 Generate output for a CPU32. This is the default
16661 when the compiler is configured for CPU32-based systems.
16662 It is equivalent to @option{-march=cpu32}.
16664 Use this option for microcontrollers with a
16665 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16666 68336, 68340, 68341, 68349 and 68360.
16670 Generate output for a 520X ColdFire CPU@. This is the default
16671 when the compiler is configured for 520X-based systems.
16672 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16673 in favor of that option.
16675 Use this option for microcontroller with a 5200 core, including
16676 the MCF5202, MCF5203, MCF5204 and MCF5206.
16680 Generate output for a 5206e ColdFire CPU@. The option is now
16681 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16685 Generate output for a member of the ColdFire 528X family.
16686 The option is now deprecated in favor of the equivalent
16687 @option{-mcpu=528x}.
16691 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16692 in favor of the equivalent @option{-mcpu=5307}.
16696 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16697 in favor of the equivalent @option{-mcpu=5407}.
16701 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16702 This includes use of hardware floating-point instructions.
16703 The option is equivalent to @option{-mcpu=547x}, and is now
16704 deprecated in favor of that option.
16708 Generate output for a 68040, without using any of the new instructions.
16709 This results in code that can run relatively efficiently on either a
16710 68020/68881 or a 68030 or a 68040. The generated code does use the
16711 68881 instructions that are emulated on the 68040.
16713 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16717 Generate output for a 68060, without using any of the new instructions.
16718 This results in code that can run relatively efficiently on either a
16719 68020/68881 or a 68030 or a 68040. The generated code does use the
16720 68881 instructions that are emulated on the 68060.
16722 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16726 @opindex mhard-float
16728 Generate floating-point instructions. This is the default for 68020
16729 and above, and for ColdFire devices that have an FPU@. It defines the
16730 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
16731 on ColdFire targets.
16734 @opindex msoft-float
16735 Do not generate floating-point instructions; use library calls instead.
16736 This is the default for 68000, 68010, and 68832 targets. It is also
16737 the default for ColdFire devices that have no FPU.
16743 Generate (do not generate) ColdFire hardware divide and remainder
16744 instructions. If @option{-march} is used without @option{-mcpu},
16745 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16746 architectures. Otherwise, the default is taken from the target CPU
16747 (either the default CPU, or the one specified by @option{-mcpu}). For
16748 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16749 @option{-mcpu=5206e}.
16751 GCC defines the macro @samp{__mcfhwdiv__} when this option is enabled.
16755 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16756 Additionally, parameters passed on the stack are also aligned to a
16757 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16761 Do not consider type @code{int} to be 16 bits wide. This is the default.
16764 @itemx -mno-bitfield
16765 @opindex mnobitfield
16766 @opindex mno-bitfield
16767 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16768 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16772 Do use the bit-field instructions. The @option{-m68020} option implies
16773 @option{-mbitfield}. This is the default if you use a configuration
16774 designed for a 68020.
16778 Use a different function-calling convention, in which functions
16779 that take a fixed number of arguments return with the @code{rtd}
16780 instruction, which pops their arguments while returning. This
16781 saves one instruction in the caller since there is no need to pop
16782 the arguments there.
16784 This calling convention is incompatible with the one normally
16785 used on Unix, so you cannot use it if you need to call libraries
16786 compiled with the Unix compiler.
16788 Also, you must provide function prototypes for all functions that
16789 take variable numbers of arguments (including @code{printf});
16790 otherwise incorrect code is generated for calls to those
16793 In addition, seriously incorrect code results if you call a
16794 function with too many arguments. (Normally, extra arguments are
16795 harmlessly ignored.)
16797 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16798 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16802 Do not use the calling conventions selected by @option{-mrtd}.
16803 This is the default.
16806 @itemx -mno-align-int
16807 @opindex malign-int
16808 @opindex mno-align-int
16809 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16810 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16811 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16812 Aligning variables on 32-bit boundaries produces code that runs somewhat
16813 faster on processors with 32-bit busses at the expense of more memory.
16815 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16816 aligns structures containing the above types differently than
16817 most published application binary interface specifications for the m68k.
16821 Use the pc-relative addressing mode of the 68000 directly, instead of
16822 using a global offset table. At present, this option implies @option{-fpic},
16823 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16824 not presently supported with @option{-mpcrel}, though this could be supported for
16825 68020 and higher processors.
16827 @item -mno-strict-align
16828 @itemx -mstrict-align
16829 @opindex mno-strict-align
16830 @opindex mstrict-align
16831 Do not (do) assume that unaligned memory references are handled by
16835 Generate code that allows the data segment to be located in a different
16836 area of memory from the text segment. This allows for execute-in-place in
16837 an environment without virtual memory management. This option implies
16840 @item -mno-sep-data
16841 Generate code that assumes that the data segment follows the text segment.
16842 This is the default.
16844 @item -mid-shared-library
16845 Generate code that supports shared libraries via the library ID method.
16846 This allows for execute-in-place and shared libraries in an environment
16847 without virtual memory management. This option implies @option{-fPIC}.
16849 @item -mno-id-shared-library
16850 Generate code that doesn't assume ID-based shared libraries are being used.
16851 This is the default.
16853 @item -mshared-library-id=n
16854 Specifies the identification number of the ID-based shared library being
16855 compiled. Specifying a value of 0 generates more compact code; specifying
16856 other values forces the allocation of that number to the current
16857 library, but is no more space- or time-efficient than omitting this option.
16863 When generating position-independent code for ColdFire, generate code
16864 that works if the GOT has more than 8192 entries. This code is
16865 larger and slower than code generated without this option. On M680x0
16866 processors, this option is not needed; @option{-fPIC} suffices.
16868 GCC normally uses a single instruction to load values from the GOT@.
16869 While this is relatively efficient, it only works if the GOT
16870 is smaller than about 64k. Anything larger causes the linker
16871 to report an error such as:
16873 @cindex relocation truncated to fit (ColdFire)
16875 relocation truncated to fit: R_68K_GOT16O foobar
16878 If this happens, you should recompile your code with @option{-mxgot}.
16879 It should then work with very large GOTs. However, code generated with
16880 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16881 the value of a global symbol.
16883 Note that some linkers, including newer versions of the GNU linker,
16884 can create multiple GOTs and sort GOT entries. If you have such a linker,
16885 you should only need to use @option{-mxgot} when compiling a single
16886 object file that accesses more than 8192 GOT entries. Very few do.
16888 These options have no effect unless GCC is generating
16889 position-independent code.
16893 @node MCore Options
16894 @subsection MCore Options
16895 @cindex MCore options
16897 These are the @samp{-m} options defined for the Motorola M*Core
16903 @itemx -mno-hardlit
16905 @opindex mno-hardlit
16906 Inline constants into the code stream if it can be done in two
16907 instructions or less.
16913 Use the divide instruction. (Enabled by default).
16915 @item -mrelax-immediate
16916 @itemx -mno-relax-immediate
16917 @opindex mrelax-immediate
16918 @opindex mno-relax-immediate
16919 Allow arbitrary-sized immediates in bit operations.
16921 @item -mwide-bitfields
16922 @itemx -mno-wide-bitfields
16923 @opindex mwide-bitfields
16924 @opindex mno-wide-bitfields
16925 Always treat bit-fields as @code{int}-sized.
16927 @item -m4byte-functions
16928 @itemx -mno-4byte-functions
16929 @opindex m4byte-functions
16930 @opindex mno-4byte-functions
16931 Force all functions to be aligned to a 4-byte boundary.
16933 @item -mcallgraph-data
16934 @itemx -mno-callgraph-data
16935 @opindex mcallgraph-data
16936 @opindex mno-callgraph-data
16937 Emit callgraph information.
16940 @itemx -mno-slow-bytes
16941 @opindex mslow-bytes
16942 @opindex mno-slow-bytes
16943 Prefer word access when reading byte quantities.
16945 @item -mlittle-endian
16946 @itemx -mbig-endian
16947 @opindex mlittle-endian
16948 @opindex mbig-endian
16949 Generate code for a little-endian target.
16955 Generate code for the 210 processor.
16959 Assume that runtime support has been provided and so omit the
16960 simulator library (@file{libsim.a)} from the linker command line.
16962 @item -mstack-increment=@var{size}
16963 @opindex mstack-increment
16964 Set the maximum amount for a single stack increment operation. Large
16965 values can increase the speed of programs that contain functions
16966 that need a large amount of stack space, but they can also trigger a
16967 segmentation fault if the stack is extended too much. The default
16973 @subsection MeP Options
16974 @cindex MeP options
16980 Enables the @code{abs} instruction, which is the absolute difference
16981 between two registers.
16985 Enables all the optional instructions---average, multiply, divide, bit
16986 operations, leading zero, absolute difference, min/max, clip, and
16992 Enables the @code{ave} instruction, which computes the average of two
16995 @item -mbased=@var{n}
16997 Variables of size @var{n} bytes or smaller are placed in the
16998 @code{.based} section by default. Based variables use the @code{$tp}
16999 register as a base register, and there is a 128-byte limit to the
17000 @code{.based} section.
17004 Enables the bit operation instructions---bit test (@code{btstm}), set
17005 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17006 test-and-set (@code{tas}).
17008 @item -mc=@var{name}
17010 Selects which section constant data is placed in. @var{name} may
17011 be @code{tiny}, @code{near}, or @code{far}.
17015 Enables the @code{clip} instruction. Note that @code{-mclip} is not
17016 useful unless you also provide @code{-mminmax}.
17018 @item -mconfig=@var{name}
17020 Selects one of the built-in core configurations. Each MeP chip has
17021 one or more modules in it; each module has a core CPU and a variety of
17022 coprocessors, optional instructions, and peripherals. The
17023 @code{MeP-Integrator} tool, not part of GCC, provides these
17024 configurations through this option; using this option is the same as
17025 using all the corresponding command-line options. The default
17026 configuration is @code{default}.
17030 Enables the coprocessor instructions. By default, this is a 32-bit
17031 coprocessor. Note that the coprocessor is normally enabled via the
17032 @code{-mconfig=} option.
17036 Enables the 32-bit coprocessor's instructions.
17040 Enables the 64-bit coprocessor's instructions.
17044 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17048 Causes constant variables to be placed in the @code{.near} section.
17052 Enables the @code{div} and @code{divu} instructions.
17056 Generate big-endian code.
17060 Generate little-endian code.
17062 @item -mio-volatile
17063 @opindex mio-volatile
17064 Tells the compiler that any variable marked with the @code{io}
17065 attribute is to be considered volatile.
17069 Causes variables to be assigned to the @code{.far} section by default.
17073 Enables the @code{leadz} (leading zero) instruction.
17077 Causes variables to be assigned to the @code{.near} section by default.
17081 Enables the @code{min} and @code{max} instructions.
17085 Enables the multiplication and multiply-accumulate instructions.
17089 Disables all the optional instructions enabled by @code{-mall-opts}.
17093 Enables the @code{repeat} and @code{erepeat} instructions, used for
17094 low-overhead looping.
17098 Causes all variables to default to the @code{.tiny} section. Note
17099 that there is a 65536-byte limit to this section. Accesses to these
17100 variables use the @code{%gp} base register.
17104 Enables the saturation instructions. Note that the compiler does not
17105 currently generate these itself, but this option is included for
17106 compatibility with other tools, like @code{as}.
17110 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17114 Link the simulator run-time libraries.
17118 Link the simulator runtime libraries, excluding built-in support
17119 for reset and exception vectors and tables.
17123 Causes all functions to default to the @code{.far} section. Without
17124 this option, functions default to the @code{.near} section.
17126 @item -mtiny=@var{n}
17128 Variables that are @var{n} bytes or smaller are allocated to the
17129 @code{.tiny} section. These variables use the @code{$gp} base
17130 register. The default for this option is 4, but note that there's a
17131 65536-byte limit to the @code{.tiny} section.
17135 @node MicroBlaze Options
17136 @subsection MicroBlaze Options
17137 @cindex MicroBlaze Options
17142 @opindex msoft-float
17143 Use software emulation for floating point (default).
17146 @opindex mhard-float
17147 Use hardware floating-point instructions.
17151 Do not optimize block moves, use @code{memcpy}.
17153 @item -mno-clearbss
17154 @opindex mno-clearbss
17155 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17157 @item -mcpu=@var{cpu-type}
17159 Use features of, and schedule code for, the given CPU.
17160 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17161 where @var{X} is a major version, @var{YY} is the minor version, and
17162 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17163 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17165 @item -mxl-soft-mul
17166 @opindex mxl-soft-mul
17167 Use software multiply emulation (default).
17169 @item -mxl-soft-div
17170 @opindex mxl-soft-div
17171 Use software emulation for divides (default).
17173 @item -mxl-barrel-shift
17174 @opindex mxl-barrel-shift
17175 Use the hardware barrel shifter.
17177 @item -mxl-pattern-compare
17178 @opindex mxl-pattern-compare
17179 Use pattern compare instructions.
17181 @item -msmall-divides
17182 @opindex msmall-divides
17183 Use table lookup optimization for small signed integer divisions.
17185 @item -mxl-stack-check
17186 @opindex mxl-stack-check
17187 This option is deprecated. Use @option{-fstack-check} instead.
17190 @opindex mxl-gp-opt
17191 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17193 @item -mxl-multiply-high
17194 @opindex mxl-multiply-high
17195 Use multiply high instructions for high part of 32x32 multiply.
17197 @item -mxl-float-convert
17198 @opindex mxl-float-convert
17199 Use hardware floating-point conversion instructions.
17201 @item -mxl-float-sqrt
17202 @opindex mxl-float-sqrt
17203 Use hardware floating-point square root instruction.
17206 @opindex mbig-endian
17207 Generate code for a big-endian target.
17209 @item -mlittle-endian
17210 @opindex mlittle-endian
17211 Generate code for a little-endian target.
17214 @opindex mxl-reorder
17215 Use reorder instructions (swap and byte reversed load/store).
17217 @item -mxl-mode-@var{app-model}
17218 Select application model @var{app-model}. Valid models are
17221 normal executable (default), uses startup code @file{crt0.o}.
17224 for use with Xilinx Microprocessor Debugger (XMD) based
17225 software intrusive debug agent called xmdstub. This uses startup file
17226 @file{crt1.o} and sets the start address of the program to 0x800.
17229 for applications that are loaded using a bootloader.
17230 This model uses startup file @file{crt2.o} which does not contain a processor
17231 reset vector handler. This is suitable for transferring control on a
17232 processor reset to the bootloader rather than the application.
17235 for applications that do not require any of the
17236 MicroBlaze vectors. This option may be useful for applications running
17237 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17240 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17241 @option{-mxl-mode-@var{app-model}}.
17246 @subsection MIPS Options
17247 @cindex MIPS options
17253 Generate big-endian code.
17257 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17260 @item -march=@var{arch}
17262 Generate code that runs on @var{arch}, which can be the name of a
17263 generic MIPS ISA, or the name of a particular processor.
17265 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17266 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17267 @samp{mips64}, @samp{mips64r2}, @samp{mips64r3} and @samp{mips64r5}.
17268 The processor names are:
17269 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17270 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17271 @samp{5kc}, @samp{5kf},
17273 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17274 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17275 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17276 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17277 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17278 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17280 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17281 @samp{octeon}, @samp{octeon+}, @samp{octeon2},
17284 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17285 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17286 @samp{rm7000}, @samp{rm9000},
17287 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17290 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17291 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17292 @samp{xlr} and @samp{xlp}.
17293 The special value @samp{from-abi} selects the
17294 most compatible architecture for the selected ABI (that is,
17295 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17297 The native Linux/GNU toolchain also supports the value @samp{native},
17298 which selects the best architecture option for the host processor.
17299 @option{-march=native} has no effect if GCC does not recognize
17302 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17303 (for example, @option{-march=r2k}). Prefixes are optional, and
17304 @samp{vr} may be written @samp{r}.
17306 Names of the form @samp{@var{n}f2_1} refer to processors with
17307 FPUs clocked at half the rate of the core, names of the form
17308 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17309 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17310 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17311 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17312 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17313 accepted as synonyms for @samp{@var{n}f1_1}.
17315 GCC defines two macros based on the value of this option. The first
17316 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
17317 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
17318 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
17319 For example, @option{-march=r2000} sets @samp{_MIPS_ARCH}
17320 to @samp{"r2000"} and defines the macro @samp{_MIPS_ARCH_R2000}.
17322 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
17323 above. In other words, it has the full prefix and does not
17324 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17325 the macro names the resolved architecture (either @samp{"mips1"} or
17326 @samp{"mips3"}). It names the default architecture when no
17327 @option{-march} option is given.
17329 @item -mtune=@var{arch}
17331 Optimize for @var{arch}. Among other things, this option controls
17332 the way instructions are scheduled, and the perceived cost of arithmetic
17333 operations. The list of @var{arch} values is the same as for
17336 When this option is not used, GCC optimizes for the processor
17337 specified by @option{-march}. By using @option{-march} and
17338 @option{-mtune} together, it is possible to generate code that
17339 runs on a family of processors, but optimize the code for one
17340 particular member of that family.
17342 @option{-mtune} defines the macros @samp{_MIPS_TUNE} and
17343 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17344 @option{-march} ones described above.
17348 Equivalent to @option{-march=mips1}.
17352 Equivalent to @option{-march=mips2}.
17356 Equivalent to @option{-march=mips3}.
17360 Equivalent to @option{-march=mips4}.
17364 Equivalent to @option{-march=mips32}.
17368 Equivalent to @option{-march=mips32r3}.
17372 Equivalent to @option{-march=mips32r5}.
17376 Equivalent to @option{-march=mips64}.
17380 Equivalent to @option{-march=mips64r2}.
17384 Equivalent to @option{-march=mips64r3}.
17388 Equivalent to @option{-march=mips64r5}.
17393 @opindex mno-mips16
17394 Generate (do not generate) MIPS16 code. If GCC is targeting a
17395 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17397 MIPS16 code generation can also be controlled on a per-function basis
17398 by means of @code{mips16} and @code{nomips16} attributes.
17399 @xref{Function Attributes}, for more information.
17401 @item -mflip-mips16
17402 @opindex mflip-mips16
17403 Generate MIPS16 code on alternating functions. This option is provided
17404 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17405 not intended for ordinary use in compiling user code.
17407 @item -minterlink-compressed
17408 @item -mno-interlink-compressed
17409 @opindex minterlink-compressed
17410 @opindex mno-interlink-compressed
17411 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17412 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17414 For example, code using the standard ISA encoding cannot jump directly
17415 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17416 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17417 knows that the target of the jump is not compressed.
17419 @item -minterlink-mips16
17420 @itemx -mno-interlink-mips16
17421 @opindex minterlink-mips16
17422 @opindex mno-interlink-mips16
17423 Aliases of @option{-minterlink-compressed} and
17424 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17425 and are retained for backwards compatibility.
17437 Generate code for the given ABI@.
17439 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17440 generates 64-bit code when you select a 64-bit architecture, but you
17441 can use @option{-mgp32} to get 32-bit code instead.
17443 For information about the O64 ABI, see
17444 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17446 GCC supports a variant of the o32 ABI in which floating-point registers
17447 are 64 rather than 32 bits wide. You can select this combination with
17448 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17449 and @code{mfhc1} instructions and is therefore only supported for
17450 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17452 The register assignments for arguments and return values remain the
17453 same, but each scalar value is passed in a single 64-bit register
17454 rather than a pair of 32-bit registers. For example, scalar
17455 floating-point values are returned in @samp{$f0} only, not a
17456 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17457 remains the same, but all 64 bits are saved.
17460 @itemx -mno-abicalls
17462 @opindex mno-abicalls
17463 Generate (do not generate) code that is suitable for SVR4-style
17464 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17469 Generate (do not generate) code that is fully position-independent,
17470 and that can therefore be linked into shared libraries. This option
17471 only affects @option{-mabicalls}.
17473 All @option{-mabicalls} code has traditionally been position-independent,
17474 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17475 as an extension, the GNU toolchain allows executables to use absolute
17476 accesses for locally-binding symbols. It can also use shorter GP
17477 initialization sequences and generate direct calls to locally-defined
17478 functions. This mode is selected by @option{-mno-shared}.
17480 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17481 objects that can only be linked by the GNU linker. However, the option
17482 does not affect the ABI of the final executable; it only affects the ABI
17483 of relocatable objects. Using @option{-mno-shared} generally makes
17484 executables both smaller and quicker.
17486 @option{-mshared} is the default.
17492 Assume (do not assume) that the static and dynamic linkers
17493 support PLTs and copy relocations. This option only affects
17494 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17495 has no effect without @option{-msym32}.
17497 You can make @option{-mplt} the default by configuring
17498 GCC with @option{--with-mips-plt}. The default is
17499 @option{-mno-plt} otherwise.
17505 Lift (do not lift) the usual restrictions on the size of the global
17508 GCC normally uses a single instruction to load values from the GOT@.
17509 While this is relatively efficient, it only works if the GOT
17510 is smaller than about 64k. Anything larger causes the linker
17511 to report an error such as:
17513 @cindex relocation truncated to fit (MIPS)
17515 relocation truncated to fit: R_MIPS_GOT16 foobar
17518 If this happens, you should recompile your code with @option{-mxgot}.
17519 This works with very large GOTs, although the code is also
17520 less efficient, since it takes three instructions to fetch the
17521 value of a global symbol.
17523 Note that some linkers can create multiple GOTs. If you have such a
17524 linker, you should only need to use @option{-mxgot} when a single object
17525 file accesses more than 64k's worth of GOT entries. Very few do.
17527 These options have no effect unless GCC is generating position
17532 Assume that general-purpose registers are 32 bits wide.
17536 Assume that general-purpose registers are 64 bits wide.
17540 Assume that floating-point registers are 32 bits wide.
17544 Assume that floating-point registers are 64 bits wide.
17547 @opindex mhard-float
17548 Use floating-point coprocessor instructions.
17551 @opindex msoft-float
17552 Do not use floating-point coprocessor instructions. Implement
17553 floating-point calculations using library calls instead.
17557 Equivalent to @option{-msoft-float}, but additionally asserts that the
17558 program being compiled does not perform any floating-point operations.
17559 This option is presently supported only by some bare-metal MIPS
17560 configurations, where it may select a special set of libraries
17561 that lack all floating-point support (including, for example, the
17562 floating-point @code{printf} formats).
17563 If code compiled with @code{-mno-float} accidentally contains
17564 floating-point operations, it is likely to suffer a link-time
17565 or run-time failure.
17567 @item -msingle-float
17568 @opindex msingle-float
17569 Assume that the floating-point coprocessor only supports single-precision
17572 @item -mdouble-float
17573 @opindex mdouble-float
17574 Assume that the floating-point coprocessor supports double-precision
17575 operations. This is the default.
17578 @itemx -mabs=legacy
17580 @opindex mabs=legacy
17581 These options control the treatment of the special not-a-number (NaN)
17582 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17583 @code{neg.@i{fmt}} machine instructions.
17585 By default or when the @option{-mabs=legacy} is used the legacy
17586 treatment is selected. In this case these instructions are considered
17587 arithmetic and avoided where correct operation is required and the
17588 input operand might be a NaN. A longer sequence of instructions that
17589 manipulate the sign bit of floating-point datum manually is used
17590 instead unless the @option{-ffinite-math-only} option has also been
17593 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17594 this case these instructions are considered non-arithmetic and therefore
17595 operating correctly in all cases, including in particular where the
17596 input operand is a NaN. These instructions are therefore always used
17597 for the respective operations.
17600 @itemx -mnan=legacy
17602 @opindex mnan=legacy
17603 These options control the encoding of the special not-a-number (NaN)
17604 IEEE 754 floating-point data.
17606 The @option{-mnan=legacy} option selects the legacy encoding. In this
17607 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17608 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17609 by the first bit of their trailing significand field being 1.
17611 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17612 this case qNaNs are denoted by the first bit of their trailing
17613 significand field being 1, whereas sNaNs are denoted by the first bit of
17614 their trailing significand field being 0.
17616 The default is @option{-mnan=legacy} unless GCC has been configured with
17617 @option{--with-nan=2008}.
17623 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17624 implement atomic memory built-in functions. When neither option is
17625 specified, GCC uses the instructions if the target architecture
17628 @option{-mllsc} is useful if the runtime environment can emulate the
17629 instructions and @option{-mno-llsc} can be useful when compiling for
17630 nonstandard ISAs. You can make either option the default by
17631 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17632 respectively. @option{--with-llsc} is the default for some
17633 configurations; see the installation documentation for details.
17639 Use (do not use) revision 1 of the MIPS DSP ASE@.
17640 @xref{MIPS DSP Built-in Functions}. This option defines the
17641 preprocessor macro @samp{__mips_dsp}. It also defines
17642 @samp{__mips_dsp_rev} to 1.
17648 Use (do not use) revision 2 of the MIPS DSP ASE@.
17649 @xref{MIPS DSP Built-in Functions}. This option defines the
17650 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
17651 It also defines @samp{__mips_dsp_rev} to 2.
17654 @itemx -mno-smartmips
17655 @opindex msmartmips
17656 @opindex mno-smartmips
17657 Use (do not use) the MIPS SmartMIPS ASE.
17659 @item -mpaired-single
17660 @itemx -mno-paired-single
17661 @opindex mpaired-single
17662 @opindex mno-paired-single
17663 Use (do not use) paired-single floating-point instructions.
17664 @xref{MIPS Paired-Single Support}. This option requires
17665 hardware floating-point support to be enabled.
17671 Use (do not use) MIPS Digital Media Extension instructions.
17672 This option can only be used when generating 64-bit code and requires
17673 hardware floating-point support to be enabled.
17678 @opindex mno-mips3d
17679 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17680 The option @option{-mips3d} implies @option{-mpaired-single}.
17683 @itemx -mno-micromips
17684 @opindex mmicromips
17685 @opindex mno-mmicromips
17686 Generate (do not generate) microMIPS code.
17688 MicroMIPS code generation can also be controlled on a per-function basis
17689 by means of @code{micromips} and @code{nomicromips} attributes.
17690 @xref{Function Attributes}, for more information.
17696 Use (do not use) MT Multithreading instructions.
17702 Use (do not use) the MIPS MCU ASE instructions.
17708 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17714 Use (do not use) the MIPS Virtualization Application Specific instructions.
17720 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17724 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17725 an explanation of the default and the way that the pointer size is
17730 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17732 The default size of @code{int}s, @code{long}s and pointers depends on
17733 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17734 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17735 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17736 or the same size as integer registers, whichever is smaller.
17742 Assume (do not assume) that all symbols have 32-bit values, regardless
17743 of the selected ABI@. This option is useful in combination with
17744 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17745 to generate shorter and faster references to symbolic addresses.
17749 Put definitions of externally-visible data in a small data section
17750 if that data is no bigger than @var{num} bytes. GCC can then generate
17751 more efficient accesses to the data; see @option{-mgpopt} for details.
17753 The default @option{-G} option depends on the configuration.
17755 @item -mlocal-sdata
17756 @itemx -mno-local-sdata
17757 @opindex mlocal-sdata
17758 @opindex mno-local-sdata
17759 Extend (do not extend) the @option{-G} behavior to local data too,
17760 such as to static variables in C@. @option{-mlocal-sdata} is the
17761 default for all configurations.
17763 If the linker complains that an application is using too much small data,
17764 you might want to try rebuilding the less performance-critical parts with
17765 @option{-mno-local-sdata}. You might also want to build large
17766 libraries with @option{-mno-local-sdata}, so that the libraries leave
17767 more room for the main program.
17769 @item -mextern-sdata
17770 @itemx -mno-extern-sdata
17771 @opindex mextern-sdata
17772 @opindex mno-extern-sdata
17773 Assume (do not assume) that externally-defined data is in
17774 a small data section if the size of that data is within the @option{-G} limit.
17775 @option{-mextern-sdata} is the default for all configurations.
17777 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17778 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17779 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17780 is placed in a small data section. If @var{Var} is defined by another
17781 module, you must either compile that module with a high-enough
17782 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17783 definition. If @var{Var} is common, you must link the application
17784 with a high-enough @option{-G} setting.
17786 The easiest way of satisfying these restrictions is to compile
17787 and link every module with the same @option{-G} option. However,
17788 you may wish to build a library that supports several different
17789 small data limits. You can do this by compiling the library with
17790 the highest supported @option{-G} setting and additionally using
17791 @option{-mno-extern-sdata} to stop the library from making assumptions
17792 about externally-defined data.
17798 Use (do not use) GP-relative accesses for symbols that are known to be
17799 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17800 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17803 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17804 might not hold the value of @code{_gp}. For example, if the code is
17805 part of a library that might be used in a boot monitor, programs that
17806 call boot monitor routines pass an unknown value in @code{$gp}.
17807 (In such situations, the boot monitor itself is usually compiled
17808 with @option{-G0}.)
17810 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17811 @option{-mno-extern-sdata}.
17813 @item -membedded-data
17814 @itemx -mno-embedded-data
17815 @opindex membedded-data
17816 @opindex mno-embedded-data
17817 Allocate variables to the read-only data section first if possible, then
17818 next in the small data section if possible, otherwise in data. This gives
17819 slightly slower code than the default, but reduces the amount of RAM required
17820 when executing, and thus may be preferred for some embedded systems.
17822 @item -muninit-const-in-rodata
17823 @itemx -mno-uninit-const-in-rodata
17824 @opindex muninit-const-in-rodata
17825 @opindex mno-uninit-const-in-rodata
17826 Put uninitialized @code{const} variables in the read-only data section.
17827 This option is only meaningful in conjunction with @option{-membedded-data}.
17829 @item -mcode-readable=@var{setting}
17830 @opindex mcode-readable
17831 Specify whether GCC may generate code that reads from executable sections.
17832 There are three possible settings:
17835 @item -mcode-readable=yes
17836 Instructions may freely access executable sections. This is the
17839 @item -mcode-readable=pcrel
17840 MIPS16 PC-relative load instructions can access executable sections,
17841 but other instructions must not do so. This option is useful on 4KSc
17842 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17843 It is also useful on processors that can be configured to have a dual
17844 instruction/data SRAM interface and that, like the M4K, automatically
17845 redirect PC-relative loads to the instruction RAM.
17847 @item -mcode-readable=no
17848 Instructions must not access executable sections. This option can be
17849 useful on targets that are configured to have a dual instruction/data
17850 SRAM interface but that (unlike the M4K) do not automatically redirect
17851 PC-relative loads to the instruction RAM.
17854 @item -msplit-addresses
17855 @itemx -mno-split-addresses
17856 @opindex msplit-addresses
17857 @opindex mno-split-addresses
17858 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17859 relocation operators. This option has been superseded by
17860 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17862 @item -mexplicit-relocs
17863 @itemx -mno-explicit-relocs
17864 @opindex mexplicit-relocs
17865 @opindex mno-explicit-relocs
17866 Use (do not use) assembler relocation operators when dealing with symbolic
17867 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17868 is to use assembler macros instead.
17870 @option{-mexplicit-relocs} is the default if GCC was configured
17871 to use an assembler that supports relocation operators.
17873 @item -mcheck-zero-division
17874 @itemx -mno-check-zero-division
17875 @opindex mcheck-zero-division
17876 @opindex mno-check-zero-division
17877 Trap (do not trap) on integer division by zero.
17879 The default is @option{-mcheck-zero-division}.
17881 @item -mdivide-traps
17882 @itemx -mdivide-breaks
17883 @opindex mdivide-traps
17884 @opindex mdivide-breaks
17885 MIPS systems check for division by zero by generating either a
17886 conditional trap or a break instruction. Using traps results in
17887 smaller code, but is only supported on MIPS II and later. Also, some
17888 versions of the Linux kernel have a bug that prevents trap from
17889 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17890 allow conditional traps on architectures that support them and
17891 @option{-mdivide-breaks} to force the use of breaks.
17893 The default is usually @option{-mdivide-traps}, but this can be
17894 overridden at configure time using @option{--with-divide=breaks}.
17895 Divide-by-zero checks can be completely disabled using
17896 @option{-mno-check-zero-division}.
17901 @opindex mno-memcpy
17902 Force (do not force) the use of @code{memcpy()} for non-trivial block
17903 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17904 most constant-sized copies.
17907 @itemx -mno-long-calls
17908 @opindex mlong-calls
17909 @opindex mno-long-calls
17910 Disable (do not disable) use of the @code{jal} instruction. Calling
17911 functions using @code{jal} is more efficient but requires the caller
17912 and callee to be in the same 256 megabyte segment.
17914 This option has no effect on abicalls code. The default is
17915 @option{-mno-long-calls}.
17921 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17922 instructions, as provided by the R4650 ISA@.
17928 Enable (disable) use of the @code{madd} and @code{msub} integer
17929 instructions. The default is @option{-mimadd} on architectures
17930 that support @code{madd} and @code{msub} except for the 74k
17931 architecture where it was found to generate slower code.
17934 @itemx -mno-fused-madd
17935 @opindex mfused-madd
17936 @opindex mno-fused-madd
17937 Enable (disable) use of the floating-point multiply-accumulate
17938 instructions, when they are available. The default is
17939 @option{-mfused-madd}.
17941 On the R8000 CPU when multiply-accumulate instructions are used,
17942 the intermediate product is calculated to infinite precision
17943 and is not subject to the FCSR Flush to Zero bit. This may be
17944 undesirable in some circumstances. On other processors the result
17945 is numerically identical to the equivalent computation using
17946 separate multiply, add, subtract and negate instructions.
17950 Tell the MIPS assembler to not run its preprocessor over user
17951 assembler files (with a @samp{.s} suffix) when assembling them.
17956 @opindex mno-fix-24k
17957 Work around the 24K E48 (lost data on stores during refill) errata.
17958 The workarounds are implemented by the assembler rather than by GCC@.
17961 @itemx -mno-fix-r4000
17962 @opindex mfix-r4000
17963 @opindex mno-fix-r4000
17964 Work around certain R4000 CPU errata:
17967 A double-word or a variable shift may give an incorrect result if executed
17968 immediately after starting an integer division.
17970 A double-word or a variable shift may give an incorrect result if executed
17971 while an integer multiplication is in progress.
17973 An integer division may give an incorrect result if started in a delay slot
17974 of a taken branch or a jump.
17978 @itemx -mno-fix-r4400
17979 @opindex mfix-r4400
17980 @opindex mno-fix-r4400
17981 Work around certain R4400 CPU errata:
17984 A double-word or a variable shift may give an incorrect result if executed
17985 immediately after starting an integer division.
17989 @itemx -mno-fix-r10000
17990 @opindex mfix-r10000
17991 @opindex mno-fix-r10000
17992 Work around certain R10000 errata:
17995 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17996 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17999 This option can only be used if the target architecture supports
18000 branch-likely instructions. @option{-mfix-r10000} is the default when
18001 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18005 @itemx -mno-fix-rm7000
18006 @opindex mfix-rm7000
18007 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18008 workarounds are implemented by the assembler rather than by GCC@.
18011 @itemx -mno-fix-vr4120
18012 @opindex mfix-vr4120
18013 Work around certain VR4120 errata:
18016 @code{dmultu} does not always produce the correct result.
18018 @code{div} and @code{ddiv} do not always produce the correct result if one
18019 of the operands is negative.
18021 The workarounds for the division errata rely on special functions in
18022 @file{libgcc.a}. At present, these functions are only provided by
18023 the @code{mips64vr*-elf} configurations.
18025 Other VR4120 errata require a NOP to be inserted between certain pairs of
18026 instructions. These errata are handled by the assembler, not by GCC itself.
18029 @opindex mfix-vr4130
18030 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18031 workarounds are implemented by the assembler rather than by GCC,
18032 although GCC avoids using @code{mflo} and @code{mfhi} if the
18033 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18034 instructions are available instead.
18037 @itemx -mno-fix-sb1
18039 Work around certain SB-1 CPU core errata.
18040 (This flag currently works around the SB-1 revision 2
18041 ``F1'' and ``F2'' floating-point errata.)
18043 @item -mr10k-cache-barrier=@var{setting}
18044 @opindex mr10k-cache-barrier
18045 Specify whether GCC should insert cache barriers to avoid the
18046 side-effects of speculation on R10K processors.
18048 In common with many processors, the R10K tries to predict the outcome
18049 of a conditional branch and speculatively executes instructions from
18050 the ``taken'' branch. It later aborts these instructions if the
18051 predicted outcome is wrong. However, on the R10K, even aborted
18052 instructions can have side effects.
18054 This problem only affects kernel stores and, depending on the system,
18055 kernel loads. As an example, a speculatively-executed store may load
18056 the target memory into cache and mark the cache line as dirty, even if
18057 the store itself is later aborted. If a DMA operation writes to the
18058 same area of memory before the ``dirty'' line is flushed, the cached
18059 data overwrites the DMA-ed data. See the R10K processor manual
18060 for a full description, including other potential problems.
18062 One workaround is to insert cache barrier instructions before every memory
18063 access that might be speculatively executed and that might have side
18064 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18065 controls GCC's implementation of this workaround. It assumes that
18066 aborted accesses to any byte in the following regions does not have
18071 the memory occupied by the current function's stack frame;
18074 the memory occupied by an incoming stack argument;
18077 the memory occupied by an object with a link-time-constant address.
18080 It is the kernel's responsibility to ensure that speculative
18081 accesses to these regions are indeed safe.
18083 If the input program contains a function declaration such as:
18089 then the implementation of @code{foo} must allow @code{j foo} and
18090 @code{jal foo} to be executed speculatively. GCC honors this
18091 restriction for functions it compiles itself. It expects non-GCC
18092 functions (such as hand-written assembly code) to do the same.
18094 The option has three forms:
18097 @item -mr10k-cache-barrier=load-store
18098 Insert a cache barrier before a load or store that might be
18099 speculatively executed and that might have side effects even
18102 @item -mr10k-cache-barrier=store
18103 Insert a cache barrier before a store that might be speculatively
18104 executed and that might have side effects even if aborted.
18106 @item -mr10k-cache-barrier=none
18107 Disable the insertion of cache barriers. This is the default setting.
18110 @item -mflush-func=@var{func}
18111 @itemx -mno-flush-func
18112 @opindex mflush-func
18113 Specifies the function to call to flush the I and D caches, or to not
18114 call any such function. If called, the function must take the same
18115 arguments as the common @code{_flush_func()}, that is, the address of the
18116 memory range for which the cache is being flushed, the size of the
18117 memory range, and the number 3 (to flush both caches). The default
18118 depends on the target GCC was configured for, but commonly is either
18119 @samp{_flush_func} or @samp{__cpu_flush}.
18121 @item mbranch-cost=@var{num}
18122 @opindex mbranch-cost
18123 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18124 This cost is only a heuristic and is not guaranteed to produce
18125 consistent results across releases. A zero cost redundantly selects
18126 the default, which is based on the @option{-mtune} setting.
18128 @item -mbranch-likely
18129 @itemx -mno-branch-likely
18130 @opindex mbranch-likely
18131 @opindex mno-branch-likely
18132 Enable or disable use of Branch Likely instructions, regardless of the
18133 default for the selected architecture. By default, Branch Likely
18134 instructions may be generated if they are supported by the selected
18135 architecture. An exception is for the MIPS32 and MIPS64 architectures
18136 and processors that implement those architectures; for those, Branch
18137 Likely instructions are not be generated by default because the MIPS32
18138 and MIPS64 architectures specifically deprecate their use.
18140 @item -mfp-exceptions
18141 @itemx -mno-fp-exceptions
18142 @opindex mfp-exceptions
18143 Specifies whether FP exceptions are enabled. This affects how
18144 FP instructions are scheduled for some processors.
18145 The default is that FP exceptions are
18148 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18149 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18152 @item -mvr4130-align
18153 @itemx -mno-vr4130-align
18154 @opindex mvr4130-align
18155 The VR4130 pipeline is two-way superscalar, but can only issue two
18156 instructions together if the first one is 8-byte aligned. When this
18157 option is enabled, GCC aligns pairs of instructions that it
18158 thinks should execute in parallel.
18160 This option only has an effect when optimizing for the VR4130.
18161 It normally makes code faster, but at the expense of making it bigger.
18162 It is enabled by default at optimization level @option{-O3}.
18167 Enable (disable) generation of @code{synci} instructions on
18168 architectures that support it. The @code{synci} instructions (if
18169 enabled) are generated when @code{__builtin___clear_cache()} is
18172 This option defaults to @code{-mno-synci}, but the default can be
18173 overridden by configuring with @code{--with-synci}.
18175 When compiling code for single processor systems, it is generally safe
18176 to use @code{synci}. However, on many multi-core (SMP) systems, it
18177 does not invalidate the instruction caches on all cores and may lead
18178 to undefined behavior.
18180 @item -mrelax-pic-calls
18181 @itemx -mno-relax-pic-calls
18182 @opindex mrelax-pic-calls
18183 Try to turn PIC calls that are normally dispatched via register
18184 @code{$25} into direct calls. This is only possible if the linker can
18185 resolve the destination at link-time and if the destination is within
18186 range for a direct call.
18188 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18189 an assembler and a linker that support the @code{.reloc} assembly
18190 directive and @code{-mexplicit-relocs} is in effect. With
18191 @code{-mno-explicit-relocs}, this optimization can be performed by the
18192 assembler and the linker alone without help from the compiler.
18194 @item -mmcount-ra-address
18195 @itemx -mno-mcount-ra-address
18196 @opindex mmcount-ra-address
18197 @opindex mno-mcount-ra-address
18198 Emit (do not emit) code that allows @code{_mcount} to modify the
18199 calling function's return address. When enabled, this option extends
18200 the usual @code{_mcount} interface with a new @var{ra-address}
18201 parameter, which has type @code{intptr_t *} and is passed in register
18202 @code{$12}. @code{_mcount} can then modify the return address by
18203 doing both of the following:
18206 Returning the new address in register @code{$31}.
18208 Storing the new address in @code{*@var{ra-address}},
18209 if @var{ra-address} is nonnull.
18212 The default is @option{-mno-mcount-ra-address}.
18217 @subsection MMIX Options
18218 @cindex MMIX Options
18220 These options are defined for the MMIX:
18224 @itemx -mno-libfuncs
18226 @opindex mno-libfuncs
18227 Specify that intrinsic library functions are being compiled, passing all
18228 values in registers, no matter the size.
18231 @itemx -mno-epsilon
18233 @opindex mno-epsilon
18234 Generate floating-point comparison instructions that compare with respect
18235 to the @code{rE} epsilon register.
18237 @item -mabi=mmixware
18239 @opindex mabi=mmixware
18241 Generate code that passes function parameters and return values that (in
18242 the called function) are seen as registers @code{$0} and up, as opposed to
18243 the GNU ABI which uses global registers @code{$231} and up.
18245 @item -mzero-extend
18246 @itemx -mno-zero-extend
18247 @opindex mzero-extend
18248 @opindex mno-zero-extend
18249 When reading data from memory in sizes shorter than 64 bits, use (do not
18250 use) zero-extending load instructions by default, rather than
18251 sign-extending ones.
18254 @itemx -mno-knuthdiv
18256 @opindex mno-knuthdiv
18257 Make the result of a division yielding a remainder have the same sign as
18258 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18259 remainder follows the sign of the dividend. Both methods are
18260 arithmetically valid, the latter being almost exclusively used.
18262 @item -mtoplevel-symbols
18263 @itemx -mno-toplevel-symbols
18264 @opindex mtoplevel-symbols
18265 @opindex mno-toplevel-symbols
18266 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18267 code can be used with the @code{PREFIX} assembly directive.
18271 Generate an executable in the ELF format, rather than the default
18272 @samp{mmo} format used by the @command{mmix} simulator.
18274 @item -mbranch-predict
18275 @itemx -mno-branch-predict
18276 @opindex mbranch-predict
18277 @opindex mno-branch-predict
18278 Use (do not use) the probable-branch instructions, when static branch
18279 prediction indicates a probable branch.
18281 @item -mbase-addresses
18282 @itemx -mno-base-addresses
18283 @opindex mbase-addresses
18284 @opindex mno-base-addresses
18285 Generate (do not generate) code that uses @emph{base addresses}. Using a
18286 base address automatically generates a request (handled by the assembler
18287 and the linker) for a constant to be set up in a global register. The
18288 register is used for one or more base address requests within the range 0
18289 to 255 from the value held in the register. The generally leads to short
18290 and fast code, but the number of different data items that can be
18291 addressed is limited. This means that a program that uses lots of static
18292 data may require @option{-mno-base-addresses}.
18294 @item -msingle-exit
18295 @itemx -mno-single-exit
18296 @opindex msingle-exit
18297 @opindex mno-single-exit
18298 Force (do not force) generated code to have a single exit point in each
18302 @node MN10300 Options
18303 @subsection MN10300 Options
18304 @cindex MN10300 options
18306 These @option{-m} options are defined for Matsushita MN10300 architectures:
18311 Generate code to avoid bugs in the multiply instructions for the MN10300
18312 processors. This is the default.
18314 @item -mno-mult-bug
18315 @opindex mno-mult-bug
18316 Do not generate code to avoid bugs in the multiply instructions for the
18317 MN10300 processors.
18321 Generate code using features specific to the AM33 processor.
18325 Do not generate code using features specific to the AM33 processor. This
18330 Generate code using features specific to the AM33/2.0 processor.
18334 Generate code using features specific to the AM34 processor.
18336 @item -mtune=@var{cpu-type}
18338 Use the timing characteristics of the indicated CPU type when
18339 scheduling instructions. This does not change the targeted processor
18340 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18341 @samp{am33-2} or @samp{am34}.
18343 @item -mreturn-pointer-on-d0
18344 @opindex mreturn-pointer-on-d0
18345 When generating a function that returns a pointer, return the pointer
18346 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18347 only in @code{a0}, and attempts to call such functions without a prototype
18348 result in errors. Note that this option is on by default; use
18349 @option{-mno-return-pointer-on-d0} to disable it.
18353 Do not link in the C run-time initialization object file.
18357 Indicate to the linker that it should perform a relaxation optimization pass
18358 to shorten branches, calls and absolute memory addresses. This option only
18359 has an effect when used on the command line for the final link step.
18361 This option makes symbolic debugging impossible.
18365 Allow the compiler to generate @emph{Long Instruction Word}
18366 instructions if the target is the @samp{AM33} or later. This is the
18367 default. This option defines the preprocessor macro @samp{__LIW__}.
18371 Do not allow the compiler to generate @emph{Long Instruction Word}
18372 instructions. This option defines the preprocessor macro
18377 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18378 instructions if the target is the @samp{AM33} or later. This is the
18379 default. This option defines the preprocessor macro @samp{__SETLB__}.
18383 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18384 instructions. This option defines the preprocessor macro
18385 @samp{__NO_SETLB__}.
18389 @node Moxie Options
18390 @subsection Moxie Options
18391 @cindex Moxie Options
18397 Generate big-endian code. This is the default for @samp{moxie-*-*}
18402 Generate little-endian code.
18406 Do not link in the C run-time initialization object file.
18410 @node MSP430 Options
18411 @subsection MSP430 Options
18412 @cindex MSP430 Options
18414 These options are defined for the MSP430:
18420 Force assembly output to always use hex constants. Normally such
18421 constants are signed decimals, but this option is available for
18422 testsuite and/or aesthetic purposes.
18426 Select the MCU to target. This is used to create a C preprocessor
18427 symbol based upon the MCU name, converted to upper case and pre- and
18428 post- fixed with @code{__}. This in turn will be used by the
18429 @code{msp430.h} header file to select an MCU specific supplimentary
18432 The option also sets the ISA to use. If the MCU name is one that is
18433 known to only support the 430 ISA then that is selected, otherwise the
18434 430X ISA is selected. A generic MCU name of @code{msp430} can also be
18435 used to select the 430 ISA. Similarly the generic @code{msp430x} MCU
18436 name will select the 430X ISA.
18438 In addition an MCU specific linker script will be added to the linker
18439 command line. The script's name is the name of the MCU with
18440 @code{.ld} appended. Thus specifying @option{-mmcu=xxx} on the gcc
18441 command line will define the C preprocessor symbol @code{__XXX__} and
18442 cause the linker to search for a script called @file{xxx.ld}.
18444 This option is also passed on to the assembler.
18448 Specifies the ISA to use. Accepted values are @code{msp430},
18449 @code{msp430x} and @code{msp430xv2}. This option is deprecated. The
18450 @option{-mmcu=} option should be used to select the ISA.
18454 Link to the simulator runtime libraries and linker script. Overrides
18455 any scripts that would be selected by the @option{-mmcu=} option.
18459 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18463 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18467 This option is passed to the assembler and linker, and allows the
18468 linker to perform certain optimizations that cannot be done until
18473 Describes the type of hardware multiply supported by the target.
18474 Accepted values are @code{none} for no hardware multiply, @code{16bit}
18475 for the original 16-bit-only multiply supported by early MCUs.
18476 @code{32bit} for the 16/32-bit multiply supported by later MCUs and
18477 @code{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18478 A value of @code{auto} can also be given. This tells GCC to deduce
18479 the hardware multiply support based upon the MCU name provided by the
18480 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18481 @code{32bit} hardware multiply support is assumed. @code{auto} is the
18484 Hardware multiplies are normally performed by calling a library
18485 routine. This saves space in the generated code. When compiling at
18486 @code{-O3} or higher however the hardware multiplier is invoked
18487 inline. This makes for bigger, but faster code.
18489 The hardware multiply routines disable interrupts whilst running and
18490 restore the previous interrupt state when they finish. This makes
18491 them safe to use inside interrupt handlers as well as in normal code.
18495 @node NDS32 Options
18496 @subsection NDS32 Options
18497 @cindex NDS32 Options
18499 These options are defined for NDS32 implementations:
18504 @opindex mbig-endian
18505 Generate code in big-endian mode.
18507 @item -mlittle-endian
18508 @opindex mlittle-endian
18509 Generate code in little-endian mode.
18511 @item -mreduced-regs
18512 @opindex mreduced-regs
18513 Use reduced-set registers for register allocation.
18516 @opindex mfull-regs
18517 Use full-set registers for register allocation.
18521 Generate conditional move instructions.
18525 Do not generate conditional move instructions.
18529 Generate performance extension instructions.
18531 @item -mno-perf-ext
18532 @opindex mno-perf-ext
18533 Do not generate performance extension instructions.
18537 Generate v3 push25/pop25 instructions.
18540 @opindex mno-v3push
18541 Do not generate v3 push25/pop25 instructions.
18545 Generate 16-bit instructions.
18548 @opindex mno-16-bit
18549 Do not generate 16-bit instructions.
18552 @opindex mgp-direct
18553 Generate GP base instructions directly.
18555 @item -mno-gp-direct
18556 @opindex mno-gp-direct
18557 Do no generate GP base instructions directly.
18559 @item -misr-vector-size=@var{num}
18560 @opindex misr-vector-size
18561 Specify the size of each interrupt vector, which must be 4 or 16.
18563 @item -mcache-block-size=@var{num}
18564 @opindex mcache-block-size
18565 Specify the size of each cache block,
18566 which must be a power of 2 between 4 and 512.
18568 @item -march=@var{arch}
18570 Specify the name of the target architecture.
18572 @item -mforce-fp-as-gp
18573 @opindex mforce-fp-as-gp
18574 Prevent $fp being allocated during register allocation so that compiler
18575 is able to force performing fp-as-gp optimization.
18577 @item -mforbid-fp-as-gp
18578 @opindex mforbid-fp-as-gp
18579 Forbid using $fp to access static and global variables.
18580 This option strictly forbids fp-as-gp optimization
18581 regardless of @option{-mforce-fp-as-gp}.
18585 Use special directives to guide linker doing ex9 optimization.
18588 @opindex mctor-dtor
18589 Enable constructor/destructor feature.
18593 Guide linker to relax instructions.
18597 @node Nios II Options
18598 @subsection Nios II Options
18599 @cindex Nios II options
18600 @cindex Altera Nios II options
18602 These are the options defined for the Altera Nios II processor.
18608 @cindex smaller data references
18609 Put global and static objects less than or equal to @var{num} bytes
18610 into the small data or BSS sections instead of the normal data or BSS
18611 sections. The default value of @var{num} is 8.
18617 Generate (do not generate) GP-relative accesses for objects in the
18618 small data or BSS sections. The default is @option{-mgpopt} except
18619 when @option{-fpic} or @option{-fPIC} is specified to generate
18620 position-independent code. Note that the Nios II ABI does not permit
18621 GP-relative accesses from shared libraries.
18623 You may need to specify @option{-mno-gpopt} explicitly when building
18624 programs that include large amounts of small data, including large
18625 GOT data sections. In this case, the 16-bit offset for GP-relative
18626 addressing may not be large enough to allow access to the entire
18627 small data section.
18633 Generate little-endian (default) or big-endian (experimental) code,
18636 @item -mbypass-cache
18637 @itemx -mno-bypass-cache
18638 @opindex mno-bypass-cache
18639 @opindex mbypass-cache
18640 Force all load and store instructions to always bypass cache by
18641 using I/O variants of the instructions. The default is not to
18644 @item -mno-cache-volatile
18645 @itemx -mcache-volatile
18646 @opindex mcache-volatile
18647 @opindex mno-cache-volatile
18648 Volatile memory access bypass the cache using the I/O variants of
18649 the load and store instructions. The default is not to bypass the cache.
18651 @item -mno-fast-sw-div
18652 @itemx -mfast-sw-div
18653 @opindex mno-fast-sw-div
18654 @opindex mfast-sw-div
18655 Do not use table-based fast divide for small numbers. The default
18656 is to use the fast divide at @option{-O3} and above.
18660 @itemx -mno-hw-mulx
18664 @opindex mno-hw-mul
18666 @opindex mno-hw-mulx
18668 @opindex mno-hw-div
18670 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18671 instructions by the compiler. The default is to emit @code{mul}
18672 and not emit @code{div} and @code{mulx}.
18674 @item -mcustom-@var{insn}=@var{N}
18675 @itemx -mno-custom-@var{insn}
18676 @opindex mcustom-@var{insn}
18677 @opindex mno-custom-@var{insn}
18678 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18679 custom instruction with encoding @var{N} when generating code that uses
18680 @var{insn}. For example, @code{-mcustom-fadds=253} generates custom
18681 instruction 253 for single-precision floating-point add operations instead
18682 of the default behavior of using a library call.
18684 The following values of @var{insn} are supported. Except as otherwise
18685 noted, floating-point operations are expected to be implemented with
18686 normal IEEE 754 semantics and correspond directly to the C operators or the
18687 equivalent GCC built-in functions (@pxref{Other Builtins}).
18689 Single-precision floating point:
18692 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18693 Binary arithmetic operations.
18699 Unary absolute value.
18701 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18702 Comparison operations.
18704 @item @samp{fmins}, @samp{fmaxs}
18705 Floating-point minimum and maximum. These instructions are only
18706 generated if @option{-ffinite-math-only} is specified.
18708 @item @samp{fsqrts}
18709 Unary square root operation.
18711 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18712 Floating-point trigonometric and exponential functions. These instructions
18713 are only generated if @option{-funsafe-math-optimizations} is also specified.
18717 Double-precision floating point:
18720 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18721 Binary arithmetic operations.
18727 Unary absolute value.
18729 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18730 Comparison operations.
18732 @item @samp{fmind}, @samp{fmaxd}
18733 Double-precision minimum and maximum. These instructions are only
18734 generated if @option{-ffinite-math-only} is specified.
18736 @item @samp{fsqrtd}
18737 Unary square root operation.
18739 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18740 Double-precision trigonometric and exponential functions. These instructions
18741 are only generated if @option{-funsafe-math-optimizations} is also specified.
18747 @item @samp{fextsd}
18748 Conversion from single precision to double precision.
18750 @item @samp{ftruncds}
18751 Conversion from double precision to single precision.
18753 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18754 Conversion from floating point to signed or unsigned integer types, with
18755 truncation towards zero.
18758 Conversion from single-precision floating point to signed integer,
18759 rounding to the nearest integer and ties away from zero.
18760 This corresponds to the @code{__builtin_lroundf} function when
18761 @option{-fno-math-errno} is used.
18763 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18764 Conversion from signed or unsigned integer types to floating-point types.
18768 In addition, all of the following transfer instructions for internal
18769 registers X and Y must be provided to use any of the double-precision
18770 floating-point instructions. Custom instructions taking two
18771 double-precision source operands expect the first operand in the
18772 64-bit register X. The other operand (or only operand of a unary
18773 operation) is given to the custom arithmetic instruction with the
18774 least significant half in source register @var{src1} and the most
18775 significant half in @var{src2}. A custom instruction that returns a
18776 double-precision result returns the most significant 32 bits in the
18777 destination register and the other half in 32-bit register Y.
18778 GCC automatically generates the necessary code sequences to write
18779 register X and/or read register Y when double-precision floating-point
18780 instructions are used.
18785 Write @var{src1} into the least significant half of X and @var{src2} into
18786 the most significant half of X.
18789 Write @var{src1} into Y.
18791 @item @samp{frdxhi}, @samp{frdxlo}
18792 Read the most or least (respectively) significant half of X and store it in
18796 Read the value of Y and store it into @var{dest}.
18799 Note that you can gain more local control over generation of Nios II custom
18800 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18801 and @code{target("no-custom-@var{insn}")} function attributes
18802 (@pxref{Function Attributes})
18803 or pragmas (@pxref{Function Specific Option Pragmas}).
18805 @item -mcustom-fpu-cfg=@var{name}
18806 @opindex mcustom-fpu-cfg
18808 This option enables a predefined, named set of custom instruction encodings
18809 (see @option{-mcustom-@var{insn}} above).
18810 Currently, the following sets are defined:
18812 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18813 @gccoptlist{-mcustom-fmuls=252 @gol
18814 -mcustom-fadds=253 @gol
18815 -mcustom-fsubs=254 @gol
18816 -fsingle-precision-constant}
18818 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18819 @gccoptlist{-mcustom-fmuls=252 @gol
18820 -mcustom-fadds=253 @gol
18821 -mcustom-fsubs=254 @gol
18822 -mcustom-fdivs=255 @gol
18823 -fsingle-precision-constant}
18825 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18826 @gccoptlist{-mcustom-floatus=243 @gol
18827 -mcustom-fixsi=244 @gol
18828 -mcustom-floatis=245 @gol
18829 -mcustom-fcmpgts=246 @gol
18830 -mcustom-fcmples=249 @gol
18831 -mcustom-fcmpeqs=250 @gol
18832 -mcustom-fcmpnes=251 @gol
18833 -mcustom-fmuls=252 @gol
18834 -mcustom-fadds=253 @gol
18835 -mcustom-fsubs=254 @gol
18836 -mcustom-fdivs=255 @gol
18837 -fsingle-precision-constant}
18839 Custom instruction assignments given by individual
18840 @option{-mcustom-@var{insn}=} options override those given by
18841 @option{-mcustom-fpu-cfg=}, regardless of the
18842 order of the options on the command line.
18844 Note that you can gain more local control over selection of a FPU
18845 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18846 function attribute (@pxref{Function Attributes})
18847 or pragma (@pxref{Function Specific Option Pragmas}).
18851 These additional @samp{-m} options are available for the Altera Nios II
18852 ELF (bare-metal) target:
18858 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18859 startup and termination code, and is typically used in conjunction with
18860 @option{-msys-crt0=} to specify the location of the alternate startup code
18861 provided by the HAL BSP.
18865 Link with a limited version of the C library, @option{-lsmallc}, rather than
18868 @item -msys-crt0=@var{startfile}
18870 @var{startfile} is the file name of the startfile (crt0) to use
18871 when linking. This option is only useful in conjunction with @option{-mhal}.
18873 @item -msys-lib=@var{systemlib}
18875 @var{systemlib} is the library name of the library that provides
18876 low-level system calls required by the C library,
18877 e.g. @code{read} and @code{write}.
18878 This option is typically used to link with a library provided by a HAL BSP.
18882 @node PDP-11 Options
18883 @subsection PDP-11 Options
18884 @cindex PDP-11 Options
18886 These options are defined for the PDP-11:
18891 Use hardware FPP floating point. This is the default. (FIS floating
18892 point on the PDP-11/40 is not supported.)
18895 @opindex msoft-float
18896 Do not use hardware floating point.
18900 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18904 Return floating-point results in memory. This is the default.
18908 Generate code for a PDP-11/40.
18912 Generate code for a PDP-11/45. This is the default.
18916 Generate code for a PDP-11/10.
18918 @item -mbcopy-builtin
18919 @opindex mbcopy-builtin
18920 Use inline @code{movmemhi} patterns for copying memory. This is the
18925 Do not use inline @code{movmemhi} patterns for copying memory.
18931 Use 16-bit @code{int}. This is the default.
18937 Use 32-bit @code{int}.
18940 @itemx -mno-float32
18942 @opindex mno-float32
18943 Use 64-bit @code{float}. This is the default.
18946 @itemx -mno-float64
18948 @opindex mno-float64
18949 Use 32-bit @code{float}.
18953 Use @code{abshi2} pattern. This is the default.
18957 Do not use @code{abshi2} pattern.
18959 @item -mbranch-expensive
18960 @opindex mbranch-expensive
18961 Pretend that branches are expensive. This is for experimenting with
18962 code generation only.
18964 @item -mbranch-cheap
18965 @opindex mbranch-cheap
18966 Do not pretend that branches are expensive. This is the default.
18970 Use Unix assembler syntax. This is the default when configured for
18971 @samp{pdp11-*-bsd}.
18975 Use DEC assembler syntax. This is the default when configured for any
18976 PDP-11 target other than @samp{pdp11-*-bsd}.
18979 @node picoChip Options
18980 @subsection picoChip Options
18981 @cindex picoChip options
18983 These @samp{-m} options are defined for picoChip implementations:
18987 @item -mae=@var{ae_type}
18989 Set the instruction set, register set, and instruction scheduling
18990 parameters for array element type @var{ae_type}. Supported values
18991 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18993 @option{-mae=ANY} selects a completely generic AE type. Code
18994 generated with this option runs on any of the other AE types. The
18995 code is not as efficient as it would be if compiled for a specific
18996 AE type, and some types of operation (e.g., multiplication) do not
18997 work properly on all types of AE.
18999 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19000 for compiled code, and is the default.
19002 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19003 option may suffer from poor performance of byte (char) manipulation,
19004 since the DSP AE does not provide hardware support for byte load/stores.
19006 @item -msymbol-as-address
19007 Enable the compiler to directly use a symbol name as an address in a
19008 load/store instruction, without first loading it into a
19009 register. Typically, the use of this option generates larger
19010 programs, which run faster than when the option isn't used. However, the
19011 results vary from program to program, so it is left as a user option,
19012 rather than being permanently enabled.
19014 @item -mno-inefficient-warnings
19015 Disables warnings about the generation of inefficient code. These
19016 warnings can be generated, for example, when compiling code that
19017 performs byte-level memory operations on the MAC AE type. The MAC AE has
19018 no hardware support for byte-level memory operations, so all byte
19019 load/stores must be synthesized from word load/store operations. This is
19020 inefficient and a warning is generated to indicate
19021 that you should rewrite the code to avoid byte operations, or to target
19022 an AE type that has the necessary hardware support. This option disables
19027 @node PowerPC Options
19028 @subsection PowerPC Options
19029 @cindex PowerPC options
19031 These are listed under @xref{RS/6000 and PowerPC Options}.
19034 @subsection RL78 Options
19035 @cindex RL78 Options
19041 Links in additional target libraries to support operation within a
19048 Specifies the type of hardware multiplication support to be used. The
19049 default is @code{none}, which uses software multiplication functions.
19050 The @code{g13} option is for the hardware multiply/divide peripheral
19051 only on the RL78/G13 targets. The @code{rl78} option is for the
19052 standard hardware multiplication defined in the RL78 software manual.
19054 @item -m64bit-doubles
19055 @itemx -m32bit-doubles
19056 @opindex m64bit-doubles
19057 @opindex m32bit-doubles
19058 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19059 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19060 @option{-m32bit-doubles}.
19064 @node RS/6000 and PowerPC Options
19065 @subsection IBM RS/6000 and PowerPC Options
19066 @cindex RS/6000 and PowerPC Options
19067 @cindex IBM RS/6000 and PowerPC Options
19069 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19071 @item -mpowerpc-gpopt
19072 @itemx -mno-powerpc-gpopt
19073 @itemx -mpowerpc-gfxopt
19074 @itemx -mno-powerpc-gfxopt
19077 @itemx -mno-powerpc64
19081 @itemx -mno-popcntb
19083 @itemx -mno-popcntd
19092 @itemx -mno-hard-dfp
19093 @opindex mpowerpc-gpopt
19094 @opindex mno-powerpc-gpopt
19095 @opindex mpowerpc-gfxopt
19096 @opindex mno-powerpc-gfxopt
19097 @opindex mpowerpc64
19098 @opindex mno-powerpc64
19102 @opindex mno-popcntb
19104 @opindex mno-popcntd
19110 @opindex mno-mfpgpr
19112 @opindex mno-hard-dfp
19113 You use these options to specify which instructions are available on the
19114 processor you are using. The default value of these options is
19115 determined when configuring GCC@. Specifying the
19116 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19117 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19118 rather than the options listed above.
19120 Specifying @option{-mpowerpc-gpopt} allows
19121 GCC to use the optional PowerPC architecture instructions in the
19122 General Purpose group, including floating-point square root. Specifying
19123 @option{-mpowerpc-gfxopt} allows GCC to
19124 use the optional PowerPC architecture instructions in the Graphics
19125 group, including floating-point select.
19127 The @option{-mmfcrf} option allows GCC to generate the move from
19128 condition register field instruction implemented on the POWER4
19129 processor and other processors that support the PowerPC V2.01
19131 The @option{-mpopcntb} option allows GCC to generate the popcount and
19132 double-precision FP reciprocal estimate instruction implemented on the
19133 POWER5 processor and other processors that support the PowerPC V2.02
19135 The @option{-mpopcntd} option allows GCC to generate the popcount
19136 instruction implemented on the POWER7 processor and other processors
19137 that support the PowerPC V2.06 architecture.
19138 The @option{-mfprnd} option allows GCC to generate the FP round to
19139 integer instructions implemented on the POWER5+ processor and other
19140 processors that support the PowerPC V2.03 architecture.
19141 The @option{-mcmpb} option allows GCC to generate the compare bytes
19142 instruction implemented on the POWER6 processor and other processors
19143 that support the PowerPC V2.05 architecture.
19144 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19145 general-purpose register instructions implemented on the POWER6X
19146 processor and other processors that support the extended PowerPC V2.05
19148 The @option{-mhard-dfp} option allows GCC to generate the decimal
19149 floating-point instructions implemented on some POWER processors.
19151 The @option{-mpowerpc64} option allows GCC to generate the additional
19152 64-bit instructions that are found in the full PowerPC64 architecture
19153 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19154 @option{-mno-powerpc64}.
19156 @item -mcpu=@var{cpu_type}
19158 Set architecture type, register usage, and
19159 instruction scheduling parameters for machine type @var{cpu_type}.
19160 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19161 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19162 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19163 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19164 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19165 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19166 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19167 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19168 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19169 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
19170 @samp{powerpc64}, and @samp{rs64}.
19172 @option{-mcpu=powerpc}, and @option{-mcpu=powerpc64} specify pure 32-bit
19173 PowerPC and 64-bit PowerPC architecture machine
19174 types, with an appropriate, generic processor model assumed for
19175 scheduling purposes.
19177 The other options specify a specific processor. Code generated under
19178 those options runs best on that processor, and may not run at all on
19181 The @option{-mcpu} options automatically enable or disable the
19184 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19185 -mpopcntb -mpopcntd -mpowerpc64 @gol
19186 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19187 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19188 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19189 -mquad-memory -mquad-memory-atomic}
19191 The particular options set for any particular CPU varies between
19192 compiler versions, depending on what setting seems to produce optimal
19193 code for that CPU; it doesn't necessarily reflect the actual hardware's
19194 capabilities. If you wish to set an individual option to a particular
19195 value, you may specify it after the @option{-mcpu} option, like
19196 @option{-mcpu=970 -mno-altivec}.
19198 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19199 not enabled or disabled by the @option{-mcpu} option at present because
19200 AIX does not have full support for these options. You may still
19201 enable or disable them individually if you're sure it'll work in your
19204 @item -mtune=@var{cpu_type}
19206 Set the instruction scheduling parameters for machine type
19207 @var{cpu_type}, but do not set the architecture type or register usage,
19208 as @option{-mcpu=@var{cpu_type}} does. The same
19209 values for @var{cpu_type} are used for @option{-mtune} as for
19210 @option{-mcpu}. If both are specified, the code generated uses the
19211 architecture and registers set by @option{-mcpu}, but the
19212 scheduling parameters set by @option{-mtune}.
19214 @item -mcmodel=small
19215 @opindex mcmodel=small
19216 Generate PowerPC64 code for the small model: The TOC is limited to
19219 @item -mcmodel=medium
19220 @opindex mcmodel=medium
19221 Generate PowerPC64 code for the medium model: The TOC and other static
19222 data may be up to a total of 4G in size.
19224 @item -mcmodel=large
19225 @opindex mcmodel=large
19226 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19227 in size. Other data and code is only limited by the 64-bit address
19231 @itemx -mno-altivec
19233 @opindex mno-altivec
19234 Generate code that uses (does not use) AltiVec instructions, and also
19235 enable the use of built-in functions that allow more direct access to
19236 the AltiVec instruction set. You may also need to set
19237 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19240 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19241 @option{-maltivec=be}, the element order for Altivec intrinsics such
19242 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert} will
19243 match array element order corresponding to the endianness of the
19244 target. That is, element zero identifies the leftmost element in a
19245 vector register when targeting a big-endian platform, and identifies
19246 the rightmost element in a vector register when targeting a
19247 little-endian platform.
19250 @opindex maltivec=be
19251 Generate Altivec instructions using big-endian element order,
19252 regardless of whether the target is big- or little-endian. This is
19253 the default when targeting a big-endian platform.
19255 The element order is used to interpret element numbers in Altivec
19256 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19257 @code{vec_insert}. By default, these will match array element order
19258 corresponding to the endianness for the target.
19261 @opindex maltivec=le
19262 Generate Altivec instructions using little-endian element order,
19263 regardless of whether the target is big- or little-endian. This is
19264 the default when targeting a little-endian platform. This option is
19265 currently ignored when targeting a big-endian platform.
19267 The element order is used to interpret element numbers in Altivec
19268 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19269 @code{vec_insert}. By default, these will match array element order
19270 corresponding to the endianness for the target.
19275 @opindex mno-vrsave
19276 Generate VRSAVE instructions when generating AltiVec code.
19278 @item -mgen-cell-microcode
19279 @opindex mgen-cell-microcode
19280 Generate Cell microcode instructions.
19282 @item -mwarn-cell-microcode
19283 @opindex mwarn-cell-microcode
19284 Warn when a Cell microcode instruction is emitted. An example
19285 of a Cell microcode instruction is a variable shift.
19288 @opindex msecure-plt
19289 Generate code that allows @command{ld} and @command{ld.so}
19290 to build executables and shared
19291 libraries with non-executable @code{.plt} and @code{.got} sections.
19293 32-bit SYSV ABI option.
19297 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19299 requires @code{.plt} and @code{.got}
19300 sections that are both writable and executable.
19301 This is a PowerPC 32-bit SYSV ABI option.
19307 This switch enables or disables the generation of ISEL instructions.
19309 @item -misel=@var{yes/no}
19310 This switch has been deprecated. Use @option{-misel} and
19311 @option{-mno-isel} instead.
19317 This switch enables or disables the generation of SPE simd
19323 @opindex mno-paired
19324 This switch enables or disables the generation of PAIRED simd
19327 @item -mspe=@var{yes/no}
19328 This option has been deprecated. Use @option{-mspe} and
19329 @option{-mno-spe} instead.
19335 Generate code that uses (does not use) vector/scalar (VSX)
19336 instructions, and also enable the use of built-in functions that allow
19337 more direct access to the VSX instruction set.
19342 @opindex mno-crypto
19343 Enable the use (disable) of the built-in functions that allow direct
19344 access to the cryptographic instructions that were added in version
19345 2.07 of the PowerPC ISA.
19347 @item -mdirect-move
19348 @itemx -mno-direct-move
19349 @opindex mdirect-move
19350 @opindex mno-direct-move
19351 Generate code that uses (does not use) the instructions to move data
19352 between the general purpose registers and the vector/scalar (VSX)
19353 registers that were added in version 2.07 of the PowerPC ISA.
19355 @item -mpower8-fusion
19356 @itemx -mno-power8-fusion
19357 @opindex mpower8-fusion
19358 @opindex mno-power8-fusion
19359 Generate code that keeps (does not keeps) some integer operations
19360 adjacent so that the instructions can be fused together on power8 and
19363 @item -mpower8-vector
19364 @itemx -mno-power8-vector
19365 @opindex mpower8-vector
19366 @opindex mno-power8-vector
19367 Generate code that uses (does not use) the vector and scalar
19368 instructions that were added in version 2.07 of the PowerPC ISA. Also
19369 enable the use of built-in functions that allow more direct access to
19370 the vector instructions.
19372 @item -mquad-memory
19373 @itemx -mno-quad-memory
19374 @opindex mquad-memory
19375 @opindex mno-quad-memory
19376 Generate code that uses (does not use) the non-atomic quad word memory
19377 instructions. The @option{-mquad-memory} option requires use of
19380 @item -mquad-memory-atomic
19381 @itemx -mno-quad-memory-atomic
19382 @opindex mquad-memory-atomic
19383 @opindex mno-quad-memory-atomic
19384 Generate code that uses (does not use) the atomic quad word memory
19385 instructions. The @option{-mquad-memory-atomic} option requires use of
19388 @item -mfloat-gprs=@var{yes/single/double/no}
19389 @itemx -mfloat-gprs
19390 @opindex mfloat-gprs
19391 This switch enables or disables the generation of floating-point
19392 operations on the general-purpose registers for architectures that
19395 The argument @var{yes} or @var{single} enables the use of
19396 single-precision floating-point operations.
19398 The argument @var{double} enables the use of single and
19399 double-precision floating-point operations.
19401 The argument @var{no} disables floating-point operations on the
19402 general-purpose registers.
19404 This option is currently only available on the MPC854x.
19410 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19411 targets (including GNU/Linux). The 32-bit environment sets int, long
19412 and pointer to 32 bits and generates code that runs on any PowerPC
19413 variant. The 64-bit environment sets int to 32 bits and long and
19414 pointer to 64 bits, and generates code for PowerPC64, as for
19415 @option{-mpowerpc64}.
19418 @itemx -mno-fp-in-toc
19419 @itemx -mno-sum-in-toc
19420 @itemx -mminimal-toc
19422 @opindex mno-fp-in-toc
19423 @opindex mno-sum-in-toc
19424 @opindex mminimal-toc
19425 Modify generation of the TOC (Table Of Contents), which is created for
19426 every executable file. The @option{-mfull-toc} option is selected by
19427 default. In that case, GCC allocates at least one TOC entry for
19428 each unique non-automatic variable reference in your program. GCC
19429 also places floating-point constants in the TOC@. However, only
19430 16,384 entries are available in the TOC@.
19432 If you receive a linker error message that saying you have overflowed
19433 the available TOC space, you can reduce the amount of TOC space used
19434 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19435 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19436 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19437 generate code to calculate the sum of an address and a constant at
19438 run time instead of putting that sum into the TOC@. You may specify one
19439 or both of these options. Each causes GCC to produce very slightly
19440 slower and larger code at the expense of conserving TOC space.
19442 If you still run out of space in the TOC even when you specify both of
19443 these options, specify @option{-mminimal-toc} instead. This option causes
19444 GCC to make only one TOC entry for every file. When you specify this
19445 option, GCC produces code that is slower and larger but which
19446 uses extremely little TOC space. You may wish to use this option
19447 only on files that contain less frequently-executed code.
19453 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19454 @code{long} type, and the infrastructure needed to support them.
19455 Specifying @option{-maix64} implies @option{-mpowerpc64},
19456 while @option{-maix32} disables the 64-bit ABI and
19457 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19460 @itemx -mno-xl-compat
19461 @opindex mxl-compat
19462 @opindex mno-xl-compat
19463 Produce code that conforms more closely to IBM XL compiler semantics
19464 when using AIX-compatible ABI@. Pass floating-point arguments to
19465 prototyped functions beyond the register save area (RSA) on the stack
19466 in addition to argument FPRs. Do not assume that most significant
19467 double in 128-bit long double value is properly rounded when comparing
19468 values and converting to double. Use XL symbol names for long double
19471 The AIX calling convention was extended but not initially documented to
19472 handle an obscure K&R C case of calling a function that takes the
19473 address of its arguments with fewer arguments than declared. IBM XL
19474 compilers access floating-point arguments that do not fit in the
19475 RSA from the stack when a subroutine is compiled without
19476 optimization. Because always storing floating-point arguments on the
19477 stack is inefficient and rarely needed, this option is not enabled by
19478 default and only is necessary when calling subroutines compiled by IBM
19479 XL compilers without optimization.
19483 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19484 application written to use message passing with special startup code to
19485 enable the application to run. The system must have PE installed in the
19486 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19487 must be overridden with the @option{-specs=} option to specify the
19488 appropriate directory location. The Parallel Environment does not
19489 support threads, so the @option{-mpe} option and the @option{-pthread}
19490 option are incompatible.
19492 @item -malign-natural
19493 @itemx -malign-power
19494 @opindex malign-natural
19495 @opindex malign-power
19496 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19497 @option{-malign-natural} overrides the ABI-defined alignment of larger
19498 types, such as floating-point doubles, on their natural size-based boundary.
19499 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19500 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19502 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19506 @itemx -mhard-float
19507 @opindex msoft-float
19508 @opindex mhard-float
19509 Generate code that does not use (uses) the floating-point register set.
19510 Software floating-point emulation is provided if you use the
19511 @option{-msoft-float} option, and pass the option to GCC when linking.
19513 @item -msingle-float
19514 @itemx -mdouble-float
19515 @opindex msingle-float
19516 @opindex mdouble-float
19517 Generate code for single- or double-precision floating-point operations.
19518 @option{-mdouble-float} implies @option{-msingle-float}.
19521 @opindex msimple-fpu
19522 Do not generate @code{sqrt} and @code{div} instructions for hardware
19523 floating-point unit.
19525 @item -mfpu=@var{name}
19527 Specify type of floating-point unit. Valid values for @var{name} are
19528 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19529 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19530 @samp{sp_full} (equivalent to @option{-msingle-float}),
19531 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19534 @opindex mxilinx-fpu
19535 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19538 @itemx -mno-multiple
19540 @opindex mno-multiple
19541 Generate code that uses (does not use) the load multiple word
19542 instructions and the store multiple word instructions. These
19543 instructions are generated by default on POWER systems, and not
19544 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19545 PowerPC systems, since those instructions do not work when the
19546 processor is in little-endian mode. The exceptions are PPC740 and
19547 PPC750 which permit these instructions in little-endian mode.
19552 @opindex mno-string
19553 Generate code that uses (does not use) the load string instructions
19554 and the store string word instructions to save multiple registers and
19555 do small block moves. These instructions are generated by default on
19556 POWER systems, and not generated on PowerPC systems. Do not use
19557 @option{-mstring} on little-endian PowerPC systems, since those
19558 instructions do not work when the processor is in little-endian mode.
19559 The exceptions are PPC740 and PPC750 which permit these instructions
19560 in little-endian mode.
19565 @opindex mno-update
19566 Generate code that uses (does not use) the load or store instructions
19567 that update the base register to the address of the calculated memory
19568 location. These instructions are generated by default. If you use
19569 @option{-mno-update}, there is a small window between the time that the
19570 stack pointer is updated and the address of the previous frame is
19571 stored, which means code that walks the stack frame across interrupts or
19572 signals may get corrupted data.
19574 @item -mavoid-indexed-addresses
19575 @itemx -mno-avoid-indexed-addresses
19576 @opindex mavoid-indexed-addresses
19577 @opindex mno-avoid-indexed-addresses
19578 Generate code that tries to avoid (not avoid) the use of indexed load
19579 or store instructions. These instructions can incur a performance
19580 penalty on Power6 processors in certain situations, such as when
19581 stepping through large arrays that cross a 16M boundary. This option
19582 is enabled by default when targeting Power6 and disabled otherwise.
19585 @itemx -mno-fused-madd
19586 @opindex mfused-madd
19587 @opindex mno-fused-madd
19588 Generate code that uses (does not use) the floating-point multiply and
19589 accumulate instructions. These instructions are generated by default
19590 if hardware floating point is used. The machine-dependent
19591 @option{-mfused-madd} option is now mapped to the machine-independent
19592 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19593 mapped to @option{-ffp-contract=off}.
19599 Generate code that uses (does not use) the half-word multiply and
19600 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19601 These instructions are generated by default when targeting those
19608 Generate code that uses (does not use) the string-search @samp{dlmzb}
19609 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19610 generated by default when targeting those processors.
19612 @item -mno-bit-align
19614 @opindex mno-bit-align
19615 @opindex mbit-align
19616 On System V.4 and embedded PowerPC systems do not (do) force structures
19617 and unions that contain bit-fields to be aligned to the base type of the
19620 For example, by default a structure containing nothing but 8
19621 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19622 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19623 the structure is aligned to a 1-byte boundary and is 1 byte in
19626 @item -mno-strict-align
19627 @itemx -mstrict-align
19628 @opindex mno-strict-align
19629 @opindex mstrict-align
19630 On System V.4 and embedded PowerPC systems do not (do) assume that
19631 unaligned memory references are handled by the system.
19633 @item -mrelocatable
19634 @itemx -mno-relocatable
19635 @opindex mrelocatable
19636 @opindex mno-relocatable
19637 Generate code that allows (does not allow) a static executable to be
19638 relocated to a different address at run time. A simple embedded
19639 PowerPC system loader should relocate the entire contents of
19640 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19641 a table of 32-bit addresses generated by this option. For this to
19642 work, all objects linked together must be compiled with
19643 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19644 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19646 @item -mrelocatable-lib
19647 @itemx -mno-relocatable-lib
19648 @opindex mrelocatable-lib
19649 @opindex mno-relocatable-lib
19650 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19651 @code{.fixup} section to allow static executables to be relocated at
19652 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19653 alignment of @option{-mrelocatable}. Objects compiled with
19654 @option{-mrelocatable-lib} may be linked with objects compiled with
19655 any combination of the @option{-mrelocatable} options.
19661 On System V.4 and embedded PowerPC systems do not (do) assume that
19662 register 2 contains a pointer to a global area pointing to the addresses
19663 used in the program.
19666 @itemx -mlittle-endian
19668 @opindex mlittle-endian
19669 On System V.4 and embedded PowerPC systems compile code for the
19670 processor in little-endian mode. The @option{-mlittle-endian} option is
19671 the same as @option{-mlittle}.
19674 @itemx -mbig-endian
19676 @opindex mbig-endian
19677 On System V.4 and embedded PowerPC systems compile code for the
19678 processor in big-endian mode. The @option{-mbig-endian} option is
19679 the same as @option{-mbig}.
19681 @item -mdynamic-no-pic
19682 @opindex mdynamic-no-pic
19683 On Darwin and Mac OS X systems, compile code so that it is not
19684 relocatable, but that its external references are relocatable. The
19685 resulting code is suitable for applications, but not shared
19688 @item -msingle-pic-base
19689 @opindex msingle-pic-base
19690 Treat the register used for PIC addressing as read-only, rather than
19691 loading it in the prologue for each function. The runtime system is
19692 responsible for initializing this register with an appropriate value
19693 before execution begins.
19695 @item -mprioritize-restricted-insns=@var{priority}
19696 @opindex mprioritize-restricted-insns
19697 This option controls the priority that is assigned to
19698 dispatch-slot restricted instructions during the second scheduling
19699 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19700 or @samp{2} to assign no, highest, or second-highest (respectively)
19701 priority to dispatch-slot restricted
19704 @item -msched-costly-dep=@var{dependence_type}
19705 @opindex msched-costly-dep
19706 This option controls which dependences are considered costly
19707 by the target during instruction scheduling. The argument
19708 @var{dependence_type} takes one of the following values:
19712 No dependence is costly.
19715 All dependences are costly.
19717 @item @samp{true_store_to_load}
19718 A true dependence from store to load is costly.
19720 @item @samp{store_to_load}
19721 Any dependence from store to load is costly.
19724 Any dependence for which the latency is greater than or equal to
19725 @var{number} is costly.
19728 @item -minsert-sched-nops=@var{scheme}
19729 @opindex minsert-sched-nops
19730 This option controls which NOP insertion scheme is used during
19731 the second scheduling pass. The argument @var{scheme} takes one of the
19739 Pad with NOPs any dispatch group that has vacant issue slots,
19740 according to the scheduler's grouping.
19742 @item @samp{regroup_exact}
19743 Insert NOPs to force costly dependent insns into
19744 separate groups. Insert exactly as many NOPs as needed to force an insn
19745 to a new group, according to the estimated processor grouping.
19748 Insert NOPs to force costly dependent insns into
19749 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19753 @opindex mcall-sysv
19754 On System V.4 and embedded PowerPC systems compile code using calling
19755 conventions that adhere to the March 1995 draft of the System V
19756 Application Binary Interface, PowerPC processor supplement. This is the
19757 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19759 @item -mcall-sysv-eabi
19761 @opindex mcall-sysv-eabi
19762 @opindex mcall-eabi
19763 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19765 @item -mcall-sysv-noeabi
19766 @opindex mcall-sysv-noeabi
19767 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19769 @item -mcall-aixdesc
19771 On System V.4 and embedded PowerPC systems compile code for the AIX
19775 @opindex mcall-linux
19776 On System V.4 and embedded PowerPC systems compile code for the
19777 Linux-based GNU system.
19779 @item -mcall-freebsd
19780 @opindex mcall-freebsd
19781 On System V.4 and embedded PowerPC systems compile code for the
19782 FreeBSD operating system.
19784 @item -mcall-netbsd
19785 @opindex mcall-netbsd
19786 On System V.4 and embedded PowerPC systems compile code for the
19787 NetBSD operating system.
19789 @item -mcall-openbsd
19790 @opindex mcall-netbsd
19791 On System V.4 and embedded PowerPC systems compile code for the
19792 OpenBSD operating system.
19794 @item -maix-struct-return
19795 @opindex maix-struct-return
19796 Return all structures in memory (as specified by the AIX ABI)@.
19798 @item -msvr4-struct-return
19799 @opindex msvr4-struct-return
19800 Return structures smaller than 8 bytes in registers (as specified by the
19803 @item -mabi=@var{abi-type}
19805 Extend the current ABI with a particular extension, or remove such extension.
19806 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
19807 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble},
19808 @var{elfv1}, @var{elfv2}@.
19812 Extend the current ABI with SPE ABI extensions. This does not change
19813 the default ABI, instead it adds the SPE ABI extensions to the current
19817 @opindex mabi=no-spe
19818 Disable Book-E SPE ABI extensions for the current ABI@.
19820 @item -mabi=ibmlongdouble
19821 @opindex mabi=ibmlongdouble
19822 Change the current ABI to use IBM extended-precision long double.
19823 This is a PowerPC 32-bit SYSV ABI option.
19825 @item -mabi=ieeelongdouble
19826 @opindex mabi=ieeelongdouble
19827 Change the current ABI to use IEEE extended-precision long double.
19828 This is a PowerPC 32-bit Linux ABI option.
19831 @opindex mabi=elfv1
19832 Change the current ABI to use the ELFv1 ABI.
19833 This is the default ABI for big-endian PowerPC 64-bit Linux.
19834 Overriding the default ABI requires special system support and is
19835 likely to fail in spectacular ways.
19838 @opindex mabi=elfv2
19839 Change the current ABI to use the ELFv2 ABI.
19840 This is the default ABI for little-endian PowerPC 64-bit Linux.
19841 Overriding the default ABI requires special system support and is
19842 likely to fail in spectacular ways.
19845 @itemx -mno-prototype
19846 @opindex mprototype
19847 @opindex mno-prototype
19848 On System V.4 and embedded PowerPC systems assume that all calls to
19849 variable argument functions are properly prototyped. Otherwise, the
19850 compiler must insert an instruction before every non-prototyped call to
19851 set or clear bit 6 of the condition code register (@var{CR}) to
19852 indicate whether floating-point values are passed in the floating-point
19853 registers in case the function takes variable arguments. With
19854 @option{-mprototype}, only calls to prototyped variable argument functions
19855 set or clear the bit.
19859 On embedded PowerPC systems, assume that the startup module is called
19860 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19861 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19866 On embedded PowerPC systems, assume that the startup module is called
19867 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19872 On embedded PowerPC systems, assume that the startup module is called
19873 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19876 @item -myellowknife
19877 @opindex myellowknife
19878 On embedded PowerPC systems, assume that the startup module is called
19879 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19884 On System V.4 and embedded PowerPC systems, specify that you are
19885 compiling for a VxWorks system.
19889 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
19890 header to indicate that @samp{eabi} extended relocations are used.
19896 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19897 Embedded Applications Binary Interface (EABI), which is a set of
19898 modifications to the System V.4 specifications. Selecting @option{-meabi}
19899 means that the stack is aligned to an 8-byte boundary, a function
19900 @code{__eabi} is called from @code{main} to set up the EABI
19901 environment, and the @option{-msdata} option can use both @code{r2} and
19902 @code{r13} to point to two separate small data areas. Selecting
19903 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19904 no EABI initialization function is called from @code{main}, and the
19905 @option{-msdata} option only uses @code{r13} to point to a single
19906 small data area. The @option{-meabi} option is on by default if you
19907 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19910 @opindex msdata=eabi
19911 On System V.4 and embedded PowerPC systems, put small initialized
19912 @code{const} global and static data in the @samp{.sdata2} section, which
19913 is pointed to by register @code{r2}. Put small initialized
19914 non-@code{const} global and static data in the @samp{.sdata} section,
19915 which is pointed to by register @code{r13}. Put small uninitialized
19916 global and static data in the @samp{.sbss} section, which is adjacent to
19917 the @samp{.sdata} section. The @option{-msdata=eabi} option is
19918 incompatible with the @option{-mrelocatable} option. The
19919 @option{-msdata=eabi} option also sets the @option{-memb} option.
19922 @opindex msdata=sysv
19923 On System V.4 and embedded PowerPC systems, put small global and static
19924 data in the @samp{.sdata} section, which is pointed to by register
19925 @code{r13}. Put small uninitialized global and static data in the
19926 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
19927 The @option{-msdata=sysv} option is incompatible with the
19928 @option{-mrelocatable} option.
19930 @item -msdata=default
19932 @opindex msdata=default
19934 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19935 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19936 same as @option{-msdata=sysv}.
19939 @opindex msdata=data
19940 On System V.4 and embedded PowerPC systems, put small global
19941 data in the @samp{.sdata} section. Put small uninitialized global
19942 data in the @samp{.sbss} section. Do not use register @code{r13}
19943 to address small data however. This is the default behavior unless
19944 other @option{-msdata} options are used.
19948 @opindex msdata=none
19950 On embedded PowerPC systems, put all initialized global and static data
19951 in the @samp{.data} section, and all uninitialized data in the
19952 @samp{.bss} section.
19954 @item -mblock-move-inline-limit=@var{num}
19955 @opindex mblock-move-inline-limit
19956 Inline all block moves (such as calls to @code{memcpy} or structure
19957 copies) less than or equal to @var{num} bytes. The minimum value for
19958 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19959 targets. The default value is target-specific.
19963 @cindex smaller data references (PowerPC)
19964 @cindex .sdata/.sdata2 references (PowerPC)
19965 On embedded PowerPC systems, put global and static items less than or
19966 equal to @var{num} bytes into the small data or BSS sections instead of
19967 the normal data or BSS section. By default, @var{num} is 8. The
19968 @option{-G @var{num}} switch is also passed to the linker.
19969 All modules should be compiled with the same @option{-G @var{num}} value.
19972 @itemx -mno-regnames
19974 @opindex mno-regnames
19975 On System V.4 and embedded PowerPC systems do (do not) emit register
19976 names in the assembly language output using symbolic forms.
19979 @itemx -mno-longcall
19981 @opindex mno-longcall
19982 By default assume that all calls are far away so that a longer and more
19983 expensive calling sequence is required. This is required for calls
19984 farther than 32 megabytes (33,554,432 bytes) from the current location.
19985 A short call is generated if the compiler knows
19986 the call cannot be that far away. This setting can be overridden by
19987 the @code{shortcall} function attribute, or by @code{#pragma
19990 Some linkers are capable of detecting out-of-range calls and generating
19991 glue code on the fly. On these systems, long calls are unnecessary and
19992 generate slower code. As of this writing, the AIX linker can do this,
19993 as can the GNU linker for PowerPC/64. It is planned to add this feature
19994 to the GNU linker for 32-bit PowerPC systems as well.
19996 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
19997 callee, L42}, plus a @dfn{branch island} (glue code). The two target
19998 addresses represent the callee and the branch island. The
19999 Darwin/PPC linker prefers the first address and generates a @code{bl
20000 callee} if the PPC @code{bl} instruction reaches the callee directly;
20001 otherwise, the linker generates @code{bl L42} to call the branch
20002 island. The branch island is appended to the body of the
20003 calling function; it computes the full 32-bit address of the callee
20006 On Mach-O (Darwin) systems, this option directs the compiler emit to
20007 the glue for every direct call, and the Darwin linker decides whether
20008 to use or discard it.
20010 In the future, GCC may ignore all longcall specifications
20011 when the linker is known to generate glue.
20013 @item -mtls-markers
20014 @itemx -mno-tls-markers
20015 @opindex mtls-markers
20016 @opindex mno-tls-markers
20017 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20018 specifying the function argument. The relocation allows the linker to
20019 reliably associate function call with argument setup instructions for
20020 TLS optimization, which in turn allows GCC to better schedule the
20025 Adds support for multithreading with the @dfn{pthreads} library.
20026 This option sets flags for both the preprocessor and linker.
20031 This option enables use of the reciprocal estimate and
20032 reciprocal square root estimate instructions with additional
20033 Newton-Raphson steps to increase precision instead of doing a divide or
20034 square root and divide for floating-point arguments. You should use
20035 the @option{-ffast-math} option when using @option{-mrecip} (or at
20036 least @option{-funsafe-math-optimizations},
20037 @option{-finite-math-only}, @option{-freciprocal-math} and
20038 @option{-fno-trapping-math}). Note that while the throughput of the
20039 sequence is generally higher than the throughput of the non-reciprocal
20040 instruction, the precision of the sequence can be decreased by up to 2
20041 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20044 @item -mrecip=@var{opt}
20045 @opindex mrecip=opt
20046 This option controls which reciprocal estimate instructions
20047 may be used. @var{opt} is a comma-separated list of options, which may
20048 be preceded by a @code{!} to invert the option:
20049 @code{all}: enable all estimate instructions,
20050 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
20051 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
20052 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
20053 @code{divf}: enable the single-precision reciprocal approximation instructions;
20054 @code{divd}: enable the double-precision reciprocal approximation instructions;
20055 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
20056 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
20057 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
20059 So, for example, @option{-mrecip=all,!rsqrtd} enables
20060 all of the reciprocal estimate instructions, except for the
20061 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20062 which handle the double-precision reciprocal square root calculations.
20064 @item -mrecip-precision
20065 @itemx -mno-recip-precision
20066 @opindex mrecip-precision
20067 Assume (do not assume) that the reciprocal estimate instructions
20068 provide higher-precision estimates than is mandated by the PowerPC
20069 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20070 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20071 The double-precision square root estimate instructions are not generated by
20072 default on low-precision machines, since they do not provide an
20073 estimate that converges after three steps.
20075 @item -mveclibabi=@var{type}
20076 @opindex mveclibabi
20077 Specifies the ABI type to use for vectorizing intrinsics using an
20078 external library. The only type supported at present is @code{mass},
20079 which specifies to use IBM's Mathematical Acceleration Subsystem
20080 (MASS) libraries for vectorizing intrinsics using external libraries.
20081 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20082 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20083 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20084 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20085 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20086 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20087 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20088 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20089 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20090 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20091 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20092 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20093 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20094 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20095 for power7. Both @option{-ftree-vectorize} and
20096 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20097 libraries must be specified at link time.
20102 Generate (do not generate) the @code{friz} instruction when the
20103 @option{-funsafe-math-optimizations} option is used to optimize
20104 rounding of floating-point values to 64-bit integer and back to floating
20105 point. The @code{friz} instruction does not return the same value if
20106 the floating-point number is too large to fit in an integer.
20108 @item -mpointers-to-nested-functions
20109 @itemx -mno-pointers-to-nested-functions
20110 @opindex mpointers-to-nested-functions
20111 Generate (do not generate) code to load up the static chain register
20112 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
20113 systems where a function pointer points to a 3-word descriptor giving
20114 the function address, TOC value to be loaded in register @var{r2}, and
20115 static chain value to be loaded in register @var{r11}. The
20116 @option{-mpointers-to-nested-functions} is on by default. You cannot
20117 call through pointers to nested functions or pointers
20118 to functions compiled in other languages that use the static chain if
20119 you use the @option{-mno-pointers-to-nested-functions}.
20121 @item -msave-toc-indirect
20122 @itemx -mno-save-toc-indirect
20123 @opindex msave-toc-indirect
20124 Generate (do not generate) code to save the TOC value in the reserved
20125 stack location in the function prologue if the function calls through
20126 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20127 saved in the prologue, it is saved just before the call through the
20128 pointer. The @option{-mno-save-toc-indirect} option is the default.
20130 @item -mcompat-align-parm
20131 @itemx -mno-compat-align-parm
20132 @opindex mcompat-align-parm
20133 Generate (do not generate) code to pass structure parameters with a
20134 maximum alignment of 64 bits, for compatibility with older versions
20137 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20138 structure parameter on a 128-bit boundary when that structure contained
20139 a member requiring 128-bit alignment. This is corrected in more
20140 recent versions of GCC. This option may be used to generate code
20141 that is compatible with functions compiled with older versions of
20144 The @option{-mno-compat-align-parm} option is the default.
20148 @subsection RX Options
20151 These command-line options are defined for RX targets:
20154 @item -m64bit-doubles
20155 @itemx -m32bit-doubles
20156 @opindex m64bit-doubles
20157 @opindex m32bit-doubles
20158 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20159 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20160 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20161 works on 32-bit values, which is why the default is
20162 @option{-m32bit-doubles}.
20168 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20169 floating-point hardware. The default is enabled for the @var{RX600}
20170 series and disabled for the @var{RX200} series.
20172 Floating-point instructions are only generated for 32-bit floating-point
20173 values, however, so the FPU hardware is not used for doubles if the
20174 @option{-m64bit-doubles} option is used.
20176 @emph{Note} If the @option{-fpu} option is enabled then
20177 @option{-funsafe-math-optimizations} is also enabled automatically.
20178 This is because the RX FPU instructions are themselves unsafe.
20180 @item -mcpu=@var{name}
20182 Selects the type of RX CPU to be targeted. Currently three types are
20183 supported, the generic @var{RX600} and @var{RX200} series hardware and
20184 the specific @var{RX610} CPU. The default is @var{RX600}.
20186 The only difference between @var{RX600} and @var{RX610} is that the
20187 @var{RX610} does not support the @code{MVTIPL} instruction.
20189 The @var{RX200} series does not have a hardware floating-point unit
20190 and so @option{-nofpu} is enabled by default when this type is
20193 @item -mbig-endian-data
20194 @itemx -mlittle-endian-data
20195 @opindex mbig-endian-data
20196 @opindex mlittle-endian-data
20197 Store data (but not code) in the big-endian format. The default is
20198 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20201 @item -msmall-data-limit=@var{N}
20202 @opindex msmall-data-limit
20203 Specifies the maximum size in bytes of global and static variables
20204 which can be placed into the small data area. Using the small data
20205 area can lead to smaller and faster code, but the size of area is
20206 limited and it is up to the programmer to ensure that the area does
20207 not overflow. Also when the small data area is used one of the RX's
20208 registers (usually @code{r13}) is reserved for use pointing to this
20209 area, so it is no longer available for use by the compiler. This
20210 could result in slower and/or larger code if variables are pushed onto
20211 the stack instead of being held in this register.
20213 Note, common variables (variables that have not been initialized) and
20214 constants are not placed into the small data area as they are assigned
20215 to other sections in the output executable.
20217 The default value is zero, which disables this feature. Note, this
20218 feature is not enabled by default with higher optimization levels
20219 (@option{-O2} etc) because of the potentially detrimental effects of
20220 reserving a register. It is up to the programmer to experiment and
20221 discover whether this feature is of benefit to their program. See the
20222 description of the @option{-mpid} option for a description of how the
20223 actual register to hold the small data area pointer is chosen.
20229 Use the simulator runtime. The default is to use the libgloss
20230 board-specific runtime.
20232 @item -mas100-syntax
20233 @itemx -mno-as100-syntax
20234 @opindex mas100-syntax
20235 @opindex mno-as100-syntax
20236 When generating assembler output use a syntax that is compatible with
20237 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20238 assembler, but it has some restrictions so it is not generated by default.
20240 @item -mmax-constant-size=@var{N}
20241 @opindex mmax-constant-size
20242 Specifies the maximum size, in bytes, of a constant that can be used as
20243 an operand in a RX instruction. Although the RX instruction set does
20244 allow constants of up to 4 bytes in length to be used in instructions,
20245 a longer value equates to a longer instruction. Thus in some
20246 circumstances it can be beneficial to restrict the size of constants
20247 that are used in instructions. Constants that are too big are instead
20248 placed into a constant pool and referenced via register indirection.
20250 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20251 or 4 means that constants of any size are allowed.
20255 Enable linker relaxation. Linker relaxation is a process whereby the
20256 linker attempts to reduce the size of a program by finding shorter
20257 versions of various instructions. Disabled by default.
20259 @item -mint-register=@var{N}
20260 @opindex mint-register
20261 Specify the number of registers to reserve for fast interrupt handler
20262 functions. The value @var{N} can be between 0 and 4. A value of 1
20263 means that register @code{r13} is reserved for the exclusive use
20264 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20265 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20266 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20267 A value of 0, the default, does not reserve any registers.
20269 @item -msave-acc-in-interrupts
20270 @opindex msave-acc-in-interrupts
20271 Specifies that interrupt handler functions should preserve the
20272 accumulator register. This is only necessary if normal code might use
20273 the accumulator register, for example because it performs 64-bit
20274 multiplications. The default is to ignore the accumulator as this
20275 makes the interrupt handlers faster.
20281 Enables the generation of position independent data. When enabled any
20282 access to constant data is done via an offset from a base address
20283 held in a register. This allows the location of constant data to be
20284 determined at run time without requiring the executable to be
20285 relocated, which is a benefit to embedded applications with tight
20286 memory constraints. Data that can be modified is not affected by this
20289 Note, using this feature reserves a register, usually @code{r13}, for
20290 the constant data base address. This can result in slower and/or
20291 larger code, especially in complicated functions.
20293 The actual register chosen to hold the constant data base address
20294 depends upon whether the @option{-msmall-data-limit} and/or the
20295 @option{-mint-register} command-line options are enabled. Starting
20296 with register @code{r13} and proceeding downwards, registers are
20297 allocated first to satisfy the requirements of @option{-mint-register},
20298 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20299 is possible for the small data area register to be @code{r8} if both
20300 @option{-mint-register=4} and @option{-mpid} are specified on the
20303 By default this feature is not enabled. The default can be restored
20304 via the @option{-mno-pid} command-line option.
20306 @item -mno-warn-multiple-fast-interrupts
20307 @itemx -mwarn-multiple-fast-interrupts
20308 @opindex mno-warn-multiple-fast-interrupts
20309 @opindex mwarn-multiple-fast-interrupts
20310 Prevents GCC from issuing a warning message if it finds more than one
20311 fast interrupt handler when it is compiling a file. The default is to
20312 issue a warning for each extra fast interrupt handler found, as the RX
20313 only supports one such interrupt.
20317 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20318 has special significance to the RX port when used with the
20319 @code{interrupt} function attribute. This attribute indicates a
20320 function intended to process fast interrupts. GCC ensures
20321 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20322 and/or @code{r13} and only provided that the normal use of the
20323 corresponding registers have been restricted via the
20324 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20327 @node S/390 and zSeries Options
20328 @subsection S/390 and zSeries Options
20329 @cindex S/390 and zSeries Options
20331 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20335 @itemx -msoft-float
20336 @opindex mhard-float
20337 @opindex msoft-float
20338 Use (do not use) the hardware floating-point instructions and registers
20339 for floating-point operations. When @option{-msoft-float} is specified,
20340 functions in @file{libgcc.a} are used to perform floating-point
20341 operations. When @option{-mhard-float} is specified, the compiler
20342 generates IEEE floating-point instructions. This is the default.
20345 @itemx -mno-hard-dfp
20347 @opindex mno-hard-dfp
20348 Use (do not use) the hardware decimal-floating-point instructions for
20349 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20350 specified, functions in @file{libgcc.a} are used to perform
20351 decimal-floating-point operations. When @option{-mhard-dfp} is
20352 specified, the compiler generates decimal-floating-point hardware
20353 instructions. This is the default for @option{-march=z9-ec} or higher.
20355 @item -mlong-double-64
20356 @itemx -mlong-double-128
20357 @opindex mlong-double-64
20358 @opindex mlong-double-128
20359 These switches control the size of @code{long double} type. A size
20360 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20361 type. This is the default.
20364 @itemx -mno-backchain
20365 @opindex mbackchain
20366 @opindex mno-backchain
20367 Store (do not store) the address of the caller's frame as backchain pointer
20368 into the callee's stack frame.
20369 A backchain may be needed to allow debugging using tools that do not understand
20370 DWARF 2 call frame information.
20371 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20372 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20373 the backchain is placed into the topmost word of the 96/160 byte register
20376 In general, code compiled with @option{-mbackchain} is call-compatible with
20377 code compiled with @option{-mmo-backchain}; however, use of the backchain
20378 for debugging purposes usually requires that the whole binary is built with
20379 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20380 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20381 to build a linux kernel use @option{-msoft-float}.
20383 The default is to not maintain the backchain.
20385 @item -mpacked-stack
20386 @itemx -mno-packed-stack
20387 @opindex mpacked-stack
20388 @opindex mno-packed-stack
20389 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20390 specified, the compiler uses the all fields of the 96/160 byte register save
20391 area only for their default purpose; unused fields still take up stack space.
20392 When @option{-mpacked-stack} is specified, register save slots are densely
20393 packed at the top of the register save area; unused space is reused for other
20394 purposes, allowing for more efficient use of the available stack space.
20395 However, when @option{-mbackchain} is also in effect, the topmost word of
20396 the save area is always used to store the backchain, and the return address
20397 register is always saved two words below the backchain.
20399 As long as the stack frame backchain is not used, code generated with
20400 @option{-mpacked-stack} is call-compatible with code generated with
20401 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20402 S/390 or zSeries generated code that uses the stack frame backchain at run
20403 time, not just for debugging purposes. Such code is not call-compatible
20404 with code compiled with @option{-mpacked-stack}. Also, note that the
20405 combination of @option{-mbackchain},
20406 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20407 to build a linux kernel use @option{-msoft-float}.
20409 The default is to not use the packed stack layout.
20412 @itemx -mno-small-exec
20413 @opindex msmall-exec
20414 @opindex mno-small-exec
20415 Generate (or do not generate) code using the @code{bras} instruction
20416 to do subroutine calls.
20417 This only works reliably if the total executable size does not
20418 exceed 64k. The default is to use the @code{basr} instruction instead,
20419 which does not have this limitation.
20425 When @option{-m31} is specified, generate code compliant to the
20426 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20427 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20428 particular to generate 64-bit instructions. For the @samp{s390}
20429 targets, the default is @option{-m31}, while the @samp{s390x}
20430 targets default to @option{-m64}.
20436 When @option{-mzarch} is specified, generate code using the
20437 instructions available on z/Architecture.
20438 When @option{-mesa} is specified, generate code using the
20439 instructions available on ESA/390. Note that @option{-mesa} is
20440 not possible with @option{-m64}.
20441 When generating code compliant to the GNU/Linux for S/390 ABI,
20442 the default is @option{-mesa}. When generating code compliant
20443 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20449 Generate (or do not generate) code using the @code{mvcle} instruction
20450 to perform block moves. When @option{-mno-mvcle} is specified,
20451 use a @code{mvc} loop instead. This is the default unless optimizing for
20458 Print (or do not print) additional debug information when compiling.
20459 The default is to not print debug information.
20461 @item -march=@var{cpu-type}
20463 Generate code that runs on @var{cpu-type}, which is the name of a system
20464 representing a certain processor type. Possible values for
20465 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20466 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
20467 When generating code using the instructions available on z/Architecture,
20468 the default is @option{-march=z900}. Otherwise, the default is
20469 @option{-march=g5}.
20471 @item -mtune=@var{cpu-type}
20473 Tune to @var{cpu-type} everything applicable about the generated code,
20474 except for the ABI and the set of available instructions.
20475 The list of @var{cpu-type} values is the same as for @option{-march}.
20476 The default is the value used for @option{-march}.
20479 @itemx -mno-tpf-trace
20480 @opindex mtpf-trace
20481 @opindex mno-tpf-trace
20482 Generate code that adds (does not add) in TPF OS specific branches to trace
20483 routines in the operating system. This option is off by default, even
20484 when compiling for the TPF OS@.
20487 @itemx -mno-fused-madd
20488 @opindex mfused-madd
20489 @opindex mno-fused-madd
20490 Generate code that uses (does not use) the floating-point multiply and
20491 accumulate instructions. These instructions are generated by default if
20492 hardware floating point is used.
20494 @item -mwarn-framesize=@var{framesize}
20495 @opindex mwarn-framesize
20496 Emit a warning if the current function exceeds the given frame size. Because
20497 this is a compile-time check it doesn't need to be a real problem when the program
20498 runs. It is intended to identify functions that most probably cause
20499 a stack overflow. It is useful to be used in an environment with limited stack
20500 size e.g.@: the linux kernel.
20502 @item -mwarn-dynamicstack
20503 @opindex mwarn-dynamicstack
20504 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20505 arrays. This is generally a bad idea with a limited stack size.
20507 @item -mstack-guard=@var{stack-guard}
20508 @itemx -mstack-size=@var{stack-size}
20509 @opindex mstack-guard
20510 @opindex mstack-size
20511 If these options are provided the S/390 back end emits additional instructions in
20512 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20513 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20514 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20515 the frame size of the compiled function is chosen.
20516 These options are intended to be used to help debugging stack overflow problems.
20517 The additionally emitted code causes only little overhead and hence can also be
20518 used in production-like systems without greater performance degradation. The given
20519 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20520 @var{stack-guard} without exceeding 64k.
20521 In order to be efficient the extra code makes the assumption that the stack starts
20522 at an address aligned to the value given by @var{stack-size}.
20523 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20525 @item -mhotpatch[=@var{halfwords}]
20526 @itemx -mno-hotpatch
20528 If the hotpatch option is enabled, a ``hot-patching'' function
20529 prologue is generated for all functions in the compilation unit.
20530 The funtion label is prepended with the given number of two-byte
20531 Nop instructions (@var{halfwords}, maximum 1000000) or 12 Nop
20532 instructions if no argument is present. Functions with a
20533 hot-patching prologue are never inlined automatically, and a
20534 hot-patching prologue is never generated for functions functions
20535 that are explicitly inline.
20537 This option can be overridden for individual functions with the
20538 @code{hotpatch} attribute.
20541 @node Score Options
20542 @subsection Score Options
20543 @cindex Score Options
20545 These options are defined for Score implementations:
20550 Compile code for big-endian mode. This is the default.
20554 Compile code for little-endian mode.
20558 Disable generation of @code{bcnz} instructions.
20562 Enable generation of unaligned load and store instructions.
20566 Enable the use of multiply-accumulate instructions. Disabled by default.
20570 Specify the SCORE5 as the target architecture.
20574 Specify the SCORE5U of the target architecture.
20578 Specify the SCORE7 as the target architecture. This is the default.
20582 Specify the SCORE7D as the target architecture.
20586 @subsection SH Options
20588 These @samp{-m} options are defined for the SH implementations:
20593 Generate code for the SH1.
20597 Generate code for the SH2.
20600 Generate code for the SH2e.
20604 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20605 that the floating-point unit is not used.
20607 @item -m2a-single-only
20608 @opindex m2a-single-only
20609 Generate code for the SH2a-FPU, in such a way that no double-precision
20610 floating-point operations are used.
20613 @opindex m2a-single
20614 Generate code for the SH2a-FPU assuming the floating-point unit is in
20615 single-precision mode by default.
20619 Generate code for the SH2a-FPU assuming the floating-point unit is in
20620 double-precision mode by default.
20624 Generate code for the SH3.
20628 Generate code for the SH3e.
20632 Generate code for the SH4 without a floating-point unit.
20634 @item -m4-single-only
20635 @opindex m4-single-only
20636 Generate code for the SH4 with a floating-point unit that only
20637 supports single-precision arithmetic.
20641 Generate code for the SH4 assuming the floating-point unit is in
20642 single-precision mode by default.
20646 Generate code for the SH4.
20650 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20651 floating-point unit is not used.
20653 @item -m4a-single-only
20654 @opindex m4a-single-only
20655 Generate code for the SH4a, in such a way that no double-precision
20656 floating-point operations are used.
20659 @opindex m4a-single
20660 Generate code for the SH4a assuming the floating-point unit is in
20661 single-precision mode by default.
20665 Generate code for the SH4a.
20669 Same as @option{-m4a-nofpu}, except that it implicitly passes
20670 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20671 instructions at the moment.
20675 Compile code for the processor in big-endian mode.
20679 Compile code for the processor in little-endian mode.
20683 Align doubles at 64-bit boundaries. Note that this changes the calling
20684 conventions, and thus some functions from the standard C library do
20685 not work unless you recompile it first with @option{-mdalign}.
20689 Shorten some address references at link time, when possible; uses the
20690 linker option @option{-relax}.
20694 Use 32-bit offsets in @code{switch} tables. The default is to use
20699 Enable the use of bit manipulation instructions on SH2A.
20703 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20704 alignment constraints.
20708 Comply with the calling conventions defined by Renesas.
20712 Comply with the calling conventions defined by Renesas.
20716 Comply with the calling conventions defined for GCC before the Renesas
20717 conventions were available. This option is the default for all
20718 targets of the SH toolchain.
20721 @opindex mnomacsave
20722 Mark the @code{MAC} register as call-clobbered, even if
20723 @option{-mhitachi} is given.
20729 Control the IEEE compliance of floating-point comparisons, which affects the
20730 handling of cases where the result of a comparison is unordered. By default
20731 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20732 enabled @option{-mno-ieee} is implicitly set, which results in faster
20733 floating-point greater-equal and less-equal comparisons. The implcit settings
20734 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20736 @item -minline-ic_invalidate
20737 @opindex minline-ic_invalidate
20738 Inline code to invalidate instruction cache entries after setting up
20739 nested function trampolines.
20740 This option has no effect if @option{-musermode} is in effect and the selected
20741 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20743 If the selected code generation option does not allow the use of the @code{icbi}
20744 instruction, and @option{-musermode} is not in effect, the inlined code
20745 manipulates the instruction cache address array directly with an associative
20746 write. This not only requires privileged mode at run time, but it also
20747 fails if the cache line had been mapped via the TLB and has become unmapped.
20751 Dump instruction size and location in the assembly code.
20754 @opindex mpadstruct
20755 This option is deprecated. It pads structures to multiple of 4 bytes,
20756 which is incompatible with the SH ABI@.
20758 @item -matomic-model=@var{model}
20759 @opindex matomic-model=@var{model}
20760 Sets the model of atomic operations and additional parameters as a comma
20761 separated list. For details on the atomic built-in functions see
20762 @ref{__atomic Builtins}. The following models and parameters are supported:
20767 Disable compiler generated atomic sequences and emit library calls for atomic
20768 operations. This is the default if the target is not @code{sh-*-linux*}.
20771 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20772 built-in functions. The generated atomic sequences require additional support
20773 from the interrupt/exception handling code of the system and are only suitable
20774 for SH3* and SH4* single-core systems. This option is enabled by default when
20775 the target is @code{sh-*-linux*} and SH3* or SH4*. When the target is SH4A,
20776 this option will also partially utilize the hardware atomic instructions
20777 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20778 @samp{strict} is specified.
20781 Generate software atomic sequences that use a variable in the thread control
20782 block. This is a variation of the gUSA sequences which can also be used on
20783 SH1* and SH2* targets. The generated atomic sequences require additional
20784 support from the interrupt/exception handling code of the system and are only
20785 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20786 parameter has to be specified as well.
20789 Generate software atomic sequences that temporarily disable interrupts by
20790 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20791 in privileged mode and is only suitable for single-core systems. Additional
20792 support from the interrupt/exception handling code of the system is not
20793 required. This model is enabled by default when the target is
20794 @code{sh-*-linux*} and SH1* or SH2*.
20797 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20798 instructions only. This is only available on SH4A and is suitable for
20799 multi-core systems. Since the hardware instructions support only 32 bit atomic
20800 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20801 Code compiled with this option will also be compatible with other software
20802 atomic model interrupt/exception handling systems if executed on an SH4A
20803 system. Additional support from the interrupt/exception handling code of the
20804 system is not required for this model.
20807 This parameter specifies the offset in bytes of the variable in the thread
20808 control block structure that should be used by the generated atomic sequences
20809 when the @samp{soft-tcb} model has been selected. For other models this
20810 parameter is ignored. The specified value must be an integer multiple of four
20811 and in the range 0-1020.
20814 This parameter prevents mixed usage of multiple atomic models, even though they
20815 would be compatible, and will make the compiler generate atomic sequences of the
20816 specified model only.
20822 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20823 Notice that depending on the particular hardware and software configuration
20824 this can degrade overall performance due to the operand cache line flushes
20825 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20826 processors the @code{tas.b} instruction must be used with caution since it
20827 can result in data corruption for certain cache configurations.
20831 Optimize for space instead of speed. Implied by @option{-Os}.
20834 @opindex mprefergot
20835 When generating position-independent code, emit function calls using
20836 the Global Offset Table instead of the Procedure Linkage Table.
20840 Don't generate privileged mode only code. This option
20841 implies @option{-mno-inline-ic_invalidate}
20842 if the inlined code would not work in user mode.
20843 This is the default when the target is @code{sh-*-linux*}.
20845 @item -multcost=@var{number}
20846 @opindex multcost=@var{number}
20847 Set the cost to assume for a multiply insn.
20849 @item -mdiv=@var{strategy}
20850 @opindex mdiv=@var{strategy}
20851 Set the division strategy to be used for integer division operations.
20852 For SHmedia @var{strategy} can be one of:
20857 Performs the operation in floating point. This has a very high latency,
20858 but needs only a few instructions, so it might be a good choice if
20859 your code has enough easily-exploitable ILP to allow the compiler to
20860 schedule the floating-point instructions together with other instructions.
20861 Division by zero causes a floating-point exception.
20864 Uses integer operations to calculate the inverse of the divisor,
20865 and then multiplies the dividend with the inverse. This strategy allows
20866 CSE and hoisting of the inverse calculation. Division by zero calculates
20867 an unspecified result, but does not trap.
20870 A variant of @samp{inv} where, if no CSE or hoisting opportunities
20871 have been found, or if the entire operation has been hoisted to the same
20872 place, the last stages of the inverse calculation are intertwined with the
20873 final multiply to reduce the overall latency, at the expense of using a few
20874 more instructions, and thus offering fewer scheduling opportunities with
20878 Calls a library function that usually implements the @samp{inv:minlat}
20880 This gives high code density for @code{m5-*media-nofpu} compilations.
20883 Uses a different entry point of the same library function, where it
20884 assumes that a pointer to a lookup table has already been set up, which
20885 exposes the pointer load to CSE and code hoisting optimizations.
20890 Use the @samp{inv} algorithm for initial
20891 code generation, but if the code stays unoptimized, revert to the @samp{call},
20892 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
20893 potentially-trapping side effect of division by zero is carried by a
20894 separate instruction, so it is possible that all the integer instructions
20895 are hoisted out, but the marker for the side effect stays where it is.
20896 A recombination to floating-point operations or a call is not possible
20901 Variants of the @samp{inv:minlat} strategy. In the case
20902 that the inverse calculation is not separated from the multiply, they speed
20903 up division where the dividend fits into 20 bits (plus sign where applicable)
20904 by inserting a test to skip a number of operations in this case; this test
20905 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
20906 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
20910 For targets other than SHmedia @var{strategy} can be one of:
20915 Calls a library function that uses the single-step division instruction
20916 @code{div1} to perform the operation. Division by zero calculates an
20917 unspecified result and does not trap. This is the default except for SH4,
20918 SH2A and SHcompact.
20921 Calls a library function that performs the operation in double precision
20922 floating point. Division by zero causes a floating-point exception. This is
20923 the default for SHcompact with FPU. Specifying this for targets that do not
20924 have a double precision FPU will default to @code{call-div1}.
20927 Calls a library function that uses a lookup table for small divisors and
20928 the @code{div1} instruction with case distinction for larger divisors. Division
20929 by zero calculates an unspecified result and does not trap. This is the default
20930 for SH4. Specifying this for targets that do not have dynamic shift
20931 instructions will default to @code{call-div1}.
20935 When a division strategy has not been specified the default strategy will be
20936 selected based on the current target. For SH2A the default strategy is to
20937 use the @code{divs} and @code{divu} instructions instead of library function
20940 @item -maccumulate-outgoing-args
20941 @opindex maccumulate-outgoing-args
20942 Reserve space once for outgoing arguments in the function prologue rather
20943 than around each call. Generally beneficial for performance and size. Also
20944 needed for unwinding to avoid changing the stack frame around conditional code.
20946 @item -mdivsi3_libfunc=@var{name}
20947 @opindex mdivsi3_libfunc=@var{name}
20948 Set the name of the library function used for 32-bit signed division to
20950 This only affects the name used in the @samp{call} and @samp{inv:call}
20951 division strategies, and the compiler still expects the same
20952 sets of input/output/clobbered registers as if this option were not present.
20954 @item -mfixed-range=@var{register-range}
20955 @opindex mfixed-range
20956 Generate code treating the given register range as fixed registers.
20957 A fixed register is one that the register allocator can not use. This is
20958 useful when compiling kernel code. A register range is specified as
20959 two registers separated by a dash. Multiple register ranges can be
20960 specified separated by a comma.
20962 @item -mindexed-addressing
20963 @opindex mindexed-addressing
20964 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
20965 This is only safe if the hardware and/or OS implement 32-bit wrap-around
20966 semantics for the indexed addressing mode. The architecture allows the
20967 implementation of processors with 64-bit MMU, which the OS could use to
20968 get 32-bit addressing, but since no current hardware implementation supports
20969 this or any other way to make the indexed addressing mode safe to use in
20970 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
20972 @item -mgettrcost=@var{number}
20973 @opindex mgettrcost=@var{number}
20974 Set the cost assumed for the @code{gettr} instruction to @var{number}.
20975 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
20979 Assume @code{pt*} instructions won't trap. This generally generates
20980 better-scheduled code, but is unsafe on current hardware.
20981 The current architecture
20982 definition says that @code{ptabs} and @code{ptrel} trap when the target
20984 This has the unintentional effect of making it unsafe to schedule these
20985 instructions before a branch, or hoist them out of a loop. For example,
20986 @code{__do_global_ctors}, a part of @file{libgcc}
20987 that runs constructors at program
20988 startup, calls functions in a list which is delimited by @minus{}1. With the
20989 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
20990 That means that all the constructors run a bit more quickly, but when
20991 the loop comes to the end of the list, the program crashes because @code{ptabs}
20992 loads @minus{}1 into a target register.
20994 Since this option is unsafe for any
20995 hardware implementing the current architecture specification, the default
20996 is @option{-mno-pt-fixed}. Unless specified explicitly with
20997 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
20998 this deters register allocation from using target registers for storing
21001 @item -minvalid-symbols
21002 @opindex minvalid-symbols
21003 Assume symbols might be invalid. Ordinary function symbols generated by
21004 the compiler are always valid to load with
21005 @code{movi}/@code{shori}/@code{ptabs} or
21006 @code{movi}/@code{shori}/@code{ptrel},
21007 but with assembler and/or linker tricks it is possible
21008 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
21009 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
21010 It prevents cross-basic-block CSE, hoisting and most scheduling
21011 of symbol loads. The default is @option{-mno-invalid-symbols}.
21013 @item -mbranch-cost=@var{num}
21014 @opindex mbranch-cost=@var{num}
21015 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21016 make the compiler try to generate more branch-free code if possible.
21017 If not specified the value is selected depending on the processor type that
21018 is being compiled for.
21021 @itemx -mno-zdcbranch
21022 @opindex mzdcbranch
21023 @opindex mno-zdcbranch
21024 Assume (do not assume) that zero displacement conditional branch instructions
21025 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21026 compiler will try to prefer zero displacement branch code sequences. This is
21027 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21028 disabled by specifying @option{-mno-zdcbranch}.
21031 @itemx -mno-fused-madd
21032 @opindex mfused-madd
21033 @opindex mno-fused-madd
21034 Generate code that uses (does not use) the floating-point multiply and
21035 accumulate instructions. These instructions are generated by default
21036 if hardware floating point is used. The machine-dependent
21037 @option{-mfused-madd} option is now mapped to the machine-independent
21038 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21039 mapped to @option{-ffp-contract=off}.
21045 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21046 and cosine approximations. The option @code{-mfsca} must be used in
21047 combination with @code{-funsafe-math-optimizations}. It is enabled by default
21048 when generating code for SH4A. Using @code{-mno-fsca} disables sine and cosine
21049 approximations even if @code{-funsafe-math-optimizations} is in effect.
21055 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21056 reciprocal square root approximations. The option @code{-mfsrra} must be used
21057 in combination with @code{-funsafe-math-optimizations} and
21058 @code{-ffinite-math-only}. It is enabled by default when generating code for
21059 SH4A. Using @code{-mno-fsrra} disables reciprocal square root approximations
21060 even if @code{-funsafe-math-optimizations} and @code{-ffinite-math-only} are
21063 @item -mpretend-cmove
21064 @opindex mpretend-cmove
21065 Prefer zero-displacement conditional branches for conditional move instruction
21066 patterns. This can result in faster code on the SH4 processor.
21070 @node Solaris 2 Options
21071 @subsection Solaris 2 Options
21072 @cindex Solaris 2 options
21074 These @samp{-m} options are supported on Solaris 2:
21077 @item -mclear-hwcap
21078 @opindex mclear-hwcap
21079 @option{-mclear-hwcap} tells the compiler to remove the hardware
21080 capabilities generated by the Solaris assembler. This is only necessary
21081 when object files use ISA extensions not supported by the current
21082 machine, but check at runtime whether or not to use them.
21084 @item -mimpure-text
21085 @opindex mimpure-text
21086 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21087 the compiler to not pass @option{-z text} to the linker when linking a
21088 shared object. Using this option, you can link position-dependent
21089 code into a shared object.
21091 @option{-mimpure-text} suppresses the ``relocations remain against
21092 allocatable but non-writable sections'' linker error message.
21093 However, the necessary relocations trigger copy-on-write, and the
21094 shared object is not actually shared across processes. Instead of
21095 using @option{-mimpure-text}, you should compile all source code with
21096 @option{-fpic} or @option{-fPIC}.
21100 These switches are supported in addition to the above on Solaris 2:
21105 Add support for multithreading using the POSIX threads library. This
21106 option sets flags for both the preprocessor and linker. This option does
21107 not affect the thread safety of object code produced by the compiler or
21108 that of libraries supplied with it.
21112 This is a synonym for @option{-pthreads}.
21115 @node SPARC Options
21116 @subsection SPARC Options
21117 @cindex SPARC options
21119 These @samp{-m} options are supported on the SPARC:
21122 @item -mno-app-regs
21124 @opindex mno-app-regs
21126 Specify @option{-mapp-regs} to generate output using the global registers
21127 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21128 global register 1, each global register 2 through 4 is then treated as an
21129 allocable register that is clobbered by function calls. This is the default.
21131 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21132 specify @option{-mno-app-regs}. You should compile libraries and system
21133 software with this option.
21139 With @option{-mflat}, the compiler does not generate save/restore instructions
21140 and uses a ``flat'' or single register window model. This model is compatible
21141 with the regular register window model. The local registers and the input
21142 registers (0--5) are still treated as ``call-saved'' registers and are
21143 saved on the stack as needed.
21145 With @option{-mno-flat} (the default), the compiler generates save/restore
21146 instructions (except for leaf functions). This is the normal operating mode.
21149 @itemx -mhard-float
21151 @opindex mhard-float
21152 Generate output containing floating-point instructions. This is the
21156 @itemx -msoft-float
21158 @opindex msoft-float
21159 Generate output containing library calls for floating point.
21160 @strong{Warning:} the requisite libraries are not available for all SPARC
21161 targets. Normally the facilities of the machine's usual C compiler are
21162 used, but this cannot be done directly in cross-compilation. You must make
21163 your own arrangements to provide suitable library functions for
21164 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21165 @samp{sparclite-*-*} do provide software floating-point support.
21167 @option{-msoft-float} changes the calling convention in the output file;
21168 therefore, it is only useful if you compile @emph{all} of a program with
21169 this option. In particular, you need to compile @file{libgcc.a}, the
21170 library that comes with GCC, with @option{-msoft-float} in order for
21173 @item -mhard-quad-float
21174 @opindex mhard-quad-float
21175 Generate output containing quad-word (long double) floating-point
21178 @item -msoft-quad-float
21179 @opindex msoft-quad-float
21180 Generate output containing library calls for quad-word (long double)
21181 floating-point instructions. The functions called are those specified
21182 in the SPARC ABI@. This is the default.
21184 As of this writing, there are no SPARC implementations that have hardware
21185 support for the quad-word floating-point instructions. They all invoke
21186 a trap handler for one of these instructions, and then the trap handler
21187 emulates the effect of the instruction. Because of the trap handler overhead,
21188 this is much slower than calling the ABI library routines. Thus the
21189 @option{-msoft-quad-float} option is the default.
21191 @item -mno-unaligned-doubles
21192 @itemx -munaligned-doubles
21193 @opindex mno-unaligned-doubles
21194 @opindex munaligned-doubles
21195 Assume that doubles have 8-byte alignment. This is the default.
21197 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21198 alignment only if they are contained in another type, or if they have an
21199 absolute address. Otherwise, it assumes they have 4-byte alignment.
21200 Specifying this option avoids some rare compatibility problems with code
21201 generated by other compilers. It is not the default because it results
21202 in a performance loss, especially for floating-point code.
21205 @itemx -mno-user-mode
21206 @opindex muser-mode
21207 @opindex mno-user-mode
21208 Do not generate code that can only run in supervisor mode. This is relevant
21209 only for the @code{casa} instruction emitted for the LEON3 processor. The
21210 default is @option{-mno-user-mode}.
21212 @item -mno-faster-structs
21213 @itemx -mfaster-structs
21214 @opindex mno-faster-structs
21215 @opindex mfaster-structs
21216 With @option{-mfaster-structs}, the compiler assumes that structures
21217 should have 8-byte alignment. This enables the use of pairs of
21218 @code{ldd} and @code{std} instructions for copies in structure
21219 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21220 However, the use of this changed alignment directly violates the SPARC
21221 ABI@. Thus, it's intended only for use on targets where the developer
21222 acknowledges that their resulting code is not directly in line with
21223 the rules of the ABI@.
21225 @item -mcpu=@var{cpu_type}
21227 Set the instruction set, register set, and instruction scheduling parameters
21228 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21229 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21230 @samp{leon}, @samp{leon3}, @samp{sparclite}, @samp{f930}, @samp{f934},
21231 @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21232 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21233 @samp{niagara3} and @samp{niagara4}.
21235 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21236 which selects the best architecture option for the host processor.
21237 @option{-mcpu=native} has no effect if GCC does not recognize
21240 Default instruction scheduling parameters are used for values that select
21241 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21242 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21244 Here is a list of each supported architecture and their supported
21252 supersparc, hypersparc, leon, leon3
21255 f930, f934, sparclite86x
21261 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21264 By default (unless configured otherwise), GCC generates code for the V7
21265 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21266 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21267 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21268 SPARCStation 1, 2, IPX etc.
21270 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21271 architecture. The only difference from V7 code is that the compiler emits
21272 the integer multiply and integer divide instructions which exist in SPARC-V8
21273 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21274 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21277 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21278 the SPARC architecture. This adds the integer multiply, integer divide step
21279 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21280 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21281 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21282 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21283 MB86934 chip, which is the more recent SPARClite with FPU@.
21285 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21286 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21287 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21288 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21289 optimizes it for the TEMIC SPARClet chip.
21291 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21292 architecture. This adds 64-bit integer and floating-point move instructions,
21293 3 additional floating-point condition code registers and conditional move
21294 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21295 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21296 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21297 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21298 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21299 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21300 additionally optimizes it for Sun UltraSPARC T2 chips. With
21301 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21302 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21303 additionally optimizes it for Sun UltraSPARC T4 chips.
21305 @item -mtune=@var{cpu_type}
21307 Set the instruction scheduling parameters for machine type
21308 @var{cpu_type}, but do not set the instruction set or register set that the
21309 option @option{-mcpu=@var{cpu_type}} does.
21311 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21312 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21313 that select a particular CPU implementation. Those are @samp{cypress},
21314 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3}, @samp{f930},
21315 @samp{f934}, @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
21316 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3} and
21317 @samp{niagara4}. With native Solaris and GNU/Linux toolchains, @samp{native}
21323 @opindex mno-v8plus
21324 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21325 difference from the V8 ABI is that the global and out registers are
21326 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21327 mode for all SPARC-V9 processors.
21333 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21334 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21340 With @option{-mvis2}, GCC generates code that takes advantage of
21341 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21342 default is @option{-mvis2} when targeting a cpu that supports such
21343 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21344 also sets @option{-mvis}.
21350 With @option{-mvis3}, GCC generates code that takes advantage of
21351 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21352 default is @option{-mvis3} when targeting a cpu that supports such
21353 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21354 also sets @option{-mvis2} and @option{-mvis}.
21359 @opindex mno-cbcond
21360 With @option{-mcbcond}, GCC generates code that takes advantage of
21361 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21362 The default is @option{-mcbcond} when targeting a cpu that supports such
21363 instructions, such as niagara-4 and later.
21369 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21370 population count instruction. The default is @option{-mpopc}
21371 when targeting a cpu that supports such instructions, such as Niagara-2 and
21378 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21379 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21380 when targeting a cpu that supports such instructions, such as Niagara-3 and
21384 @opindex mfix-at697f
21385 Enable the documented workaround for the single erratum of the Atmel AT697F
21386 processor (which corresponds to erratum #13 of the AT697E processor).
21389 @opindex mfix-ut699
21390 Enable the documented workarounds for the floating-point errata and the data
21391 cache nullify errata of the UT699 processor.
21394 These @samp{-m} options are supported in addition to the above
21395 on SPARC-V9 processors in 64-bit environments:
21402 Generate code for a 32-bit or 64-bit environment.
21403 The 32-bit environment sets int, long and pointer to 32 bits.
21404 The 64-bit environment sets int to 32 bits and long and pointer
21407 @item -mcmodel=@var{which}
21409 Set the code model to one of
21413 The Medium/Low code model: 64-bit addresses, programs
21414 must be linked in the low 32 bits of memory. Programs can be statically
21415 or dynamically linked.
21418 The Medium/Middle code model: 64-bit addresses, programs
21419 must be linked in the low 44 bits of memory, the text and data segments must
21420 be less than 2GB in size and the data segment must be located within 2GB of
21424 The Medium/Anywhere code model: 64-bit addresses, programs
21425 may be linked anywhere in memory, the text and data segments must be less
21426 than 2GB in size and the data segment must be located within 2GB of the
21430 The Medium/Anywhere code model for embedded systems:
21431 64-bit addresses, the text and data segments must be less than 2GB in
21432 size, both starting anywhere in memory (determined at link time). The
21433 global register %g4 points to the base of the data segment. Programs
21434 are statically linked and PIC is not supported.
21437 @item -mmemory-model=@var{mem-model}
21438 @opindex mmemory-model
21439 Set the memory model in force on the processor to one of
21443 The default memory model for the processor and operating system.
21446 Relaxed Memory Order
21449 Partial Store Order
21455 Sequential Consistency
21458 These memory models are formally defined in Appendix D of the Sparc V9
21459 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21462 @itemx -mno-stack-bias
21463 @opindex mstack-bias
21464 @opindex mno-stack-bias
21465 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21466 frame pointer if present, are offset by @minus{}2047 which must be added back
21467 when making stack frame references. This is the default in 64-bit mode.
21468 Otherwise, assume no such offset is present.
21472 @subsection SPU Options
21473 @cindex SPU options
21475 These @samp{-m} options are supported on the SPU:
21479 @itemx -merror-reloc
21480 @opindex mwarn-reloc
21481 @opindex merror-reloc
21483 The loader for SPU does not handle dynamic relocations. By default, GCC
21484 gives an error when it generates code that requires a dynamic
21485 relocation. @option{-mno-error-reloc} disables the error,
21486 @option{-mwarn-reloc} generates a warning instead.
21489 @itemx -munsafe-dma
21491 @opindex munsafe-dma
21493 Instructions that initiate or test completion of DMA must not be
21494 reordered with respect to loads and stores of the memory that is being
21496 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21497 memory accesses, but that can lead to inefficient code in places where the
21498 memory is known to not change. Rather than mark the memory as volatile,
21499 you can use @option{-msafe-dma} to tell the compiler to treat
21500 the DMA instructions as potentially affecting all memory.
21502 @item -mbranch-hints
21503 @opindex mbranch-hints
21505 By default, GCC generates a branch hint instruction to avoid
21506 pipeline stalls for always-taken or probably-taken branches. A hint
21507 is not generated closer than 8 instructions away from its branch.
21508 There is little reason to disable them, except for debugging purposes,
21509 or to make an object a little bit smaller.
21513 @opindex msmall-mem
21514 @opindex mlarge-mem
21516 By default, GCC generates code assuming that addresses are never larger
21517 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21518 a full 32-bit address.
21523 By default, GCC links against startup code that assumes the SPU-style
21524 main function interface (which has an unconventional parameter list).
21525 With @option{-mstdmain}, GCC links your program against startup
21526 code that assumes a C99-style interface to @code{main}, including a
21527 local copy of @code{argv} strings.
21529 @item -mfixed-range=@var{register-range}
21530 @opindex mfixed-range
21531 Generate code treating the given register range as fixed registers.
21532 A fixed register is one that the register allocator cannot use. This is
21533 useful when compiling kernel code. A register range is specified as
21534 two registers separated by a dash. Multiple register ranges can be
21535 specified separated by a comma.
21541 Compile code assuming that pointers to the PPU address space accessed
21542 via the @code{__ea} named address space qualifier are either 32 or 64
21543 bits wide. The default is 32 bits. As this is an ABI-changing option,
21544 all object code in an executable must be compiled with the same setting.
21546 @item -maddress-space-conversion
21547 @itemx -mno-address-space-conversion
21548 @opindex maddress-space-conversion
21549 @opindex mno-address-space-conversion
21550 Allow/disallow treating the @code{__ea} address space as superset
21551 of the generic address space. This enables explicit type casts
21552 between @code{__ea} and generic pointer as well as implicit
21553 conversions of generic pointers to @code{__ea} pointers. The
21554 default is to allow address space pointer conversions.
21556 @item -mcache-size=@var{cache-size}
21557 @opindex mcache-size
21558 This option controls the version of libgcc that the compiler links to an
21559 executable and selects a software-managed cache for accessing variables
21560 in the @code{__ea} address space with a particular cache size. Possible
21561 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21562 and @samp{128}. The default cache size is 64KB.
21564 @item -matomic-updates
21565 @itemx -mno-atomic-updates
21566 @opindex matomic-updates
21567 @opindex mno-atomic-updates
21568 This option controls the version of libgcc that the compiler links to an
21569 executable and selects whether atomic updates to the software-managed
21570 cache of PPU-side variables are used. If you use atomic updates, changes
21571 to a PPU variable from SPU code using the @code{__ea} named address space
21572 qualifier do not interfere with changes to other PPU variables residing
21573 in the same cache line from PPU code. If you do not use atomic updates,
21574 such interference may occur; however, writing back cache lines is
21575 more efficient. The default behavior is to use atomic updates.
21578 @itemx -mdual-nops=@var{n}
21579 @opindex mdual-nops
21580 By default, GCC inserts nops to increase dual issue when it expects
21581 it to increase performance. @var{n} can be a value from 0 to 10. A
21582 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21583 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21585 @item -mhint-max-nops=@var{n}
21586 @opindex mhint-max-nops
21587 Maximum number of nops to insert for a branch hint. A branch hint must
21588 be at least 8 instructions away from the branch it is affecting. GCC
21589 inserts up to @var{n} nops to enforce this, otherwise it does not
21590 generate the branch hint.
21592 @item -mhint-max-distance=@var{n}
21593 @opindex mhint-max-distance
21594 The encoding of the branch hint instruction limits the hint to be within
21595 256 instructions of the branch it is affecting. By default, GCC makes
21596 sure it is within 125.
21599 @opindex msafe-hints
21600 Work around a hardware bug that causes the SPU to stall indefinitely.
21601 By default, GCC inserts the @code{hbrp} instruction to make sure
21602 this stall won't happen.
21606 @node System V Options
21607 @subsection Options for System V
21609 These additional options are available on System V Release 4 for
21610 compatibility with other compilers on those systems:
21615 Create a shared object.
21616 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21620 Identify the versions of each tool used by the compiler, in a
21621 @code{.ident} assembler directive in the output.
21625 Refrain from adding @code{.ident} directives to the output file (this is
21628 @item -YP,@var{dirs}
21630 Search the directories @var{dirs}, and no others, for libraries
21631 specified with @option{-l}.
21633 @item -Ym,@var{dir}
21635 Look in the directory @var{dir} to find the M4 preprocessor.
21636 The assembler uses this option.
21637 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21638 @c the generic assembler that comes with Solaris takes just -Ym.
21641 @node TILE-Gx Options
21642 @subsection TILE-Gx Options
21643 @cindex TILE-Gx options
21645 These @samp{-m} options are supported on the TILE-Gx:
21648 @item -mcmodel=small
21649 @opindex mcmodel=small
21650 Generate code for the small model. The distance for direct calls is
21651 limited to 500M in either direction. PC-relative addresses are 32
21652 bits. Absolute addresses support the full address range.
21654 @item -mcmodel=large
21655 @opindex mcmodel=large
21656 Generate code for the large model. There is no limitation on call
21657 distance, pc-relative addresses, or absolute addresses.
21659 @item -mcpu=@var{name}
21661 Selects the type of CPU to be targeted. Currently the only supported
21662 type is @samp{tilegx}.
21668 Generate code for a 32-bit or 64-bit environment. The 32-bit
21669 environment sets int, long, and pointer to 32 bits. The 64-bit
21670 environment sets int to 32 bits and long and pointer to 64 bits.
21673 @itemx -mlittle-endian
21674 @opindex mbig-endian
21675 @opindex mlittle-endian
21676 Generate code in big/little endian mode, respectively.
21679 @node TILEPro Options
21680 @subsection TILEPro Options
21681 @cindex TILEPro options
21683 These @samp{-m} options are supported on the TILEPro:
21686 @item -mcpu=@var{name}
21688 Selects the type of CPU to be targeted. Currently the only supported
21689 type is @samp{tilepro}.
21693 Generate code for a 32-bit environment, which sets int, long, and
21694 pointer to 32 bits. This is the only supported behavior so the flag
21695 is essentially ignored.
21699 @subsection V850 Options
21700 @cindex V850 Options
21702 These @samp{-m} options are defined for V850 implementations:
21706 @itemx -mno-long-calls
21707 @opindex mlong-calls
21708 @opindex mno-long-calls
21709 Treat all calls as being far away (near). If calls are assumed to be
21710 far away, the compiler always loads the function's address into a
21711 register, and calls indirect through the pointer.
21717 Do not optimize (do optimize) basic blocks that use the same index
21718 pointer 4 or more times to copy pointer into the @code{ep} register, and
21719 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21720 option is on by default if you optimize.
21722 @item -mno-prolog-function
21723 @itemx -mprolog-function
21724 @opindex mno-prolog-function
21725 @opindex mprolog-function
21726 Do not use (do use) external functions to save and restore registers
21727 at the prologue and epilogue of a function. The external functions
21728 are slower, but use less code space if more than one function saves
21729 the same number of registers. The @option{-mprolog-function} option
21730 is on by default if you optimize.
21734 Try to make the code as small as possible. At present, this just turns
21735 on the @option{-mep} and @option{-mprolog-function} options.
21737 @item -mtda=@var{n}
21739 Put static or global variables whose size is @var{n} bytes or less into
21740 the tiny data area that register @code{ep} points to. The tiny data
21741 area can hold up to 256 bytes in total (128 bytes for byte references).
21743 @item -msda=@var{n}
21745 Put static or global variables whose size is @var{n} bytes or less into
21746 the small data area that register @code{gp} points to. The small data
21747 area can hold up to 64 kilobytes.
21749 @item -mzda=@var{n}
21751 Put static or global variables whose size is @var{n} bytes or less into
21752 the first 32 kilobytes of memory.
21756 Specify that the target processor is the V850.
21760 Specify that the target processor is the V850E3V5. The preprocessor
21761 constant @samp{__v850e3v5__} is defined if this option is used.
21765 Specify that the target processor is the V850E3V5. This is an alias for
21766 the @option{-mv850e3v5} option.
21770 Specify that the target processor is the V850E2V3. The preprocessor
21771 constant @samp{__v850e2v3__} is defined if this option is used.
21775 Specify that the target processor is the V850E2. The preprocessor
21776 constant @samp{__v850e2__} is defined if this option is used.
21780 Specify that the target processor is the V850E1. The preprocessor
21781 constants @samp{__v850e1__} and @samp{__v850e__} are defined if
21782 this option is used.
21786 Specify that the target processor is the V850ES. This is an alias for
21787 the @option{-mv850e1} option.
21791 Specify that the target processor is the V850E@. The preprocessor
21792 constant @samp{__v850e__} is defined if this option is used.
21794 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21795 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21796 are defined then a default target processor is chosen and the
21797 relevant @samp{__v850*__} preprocessor constant is defined.
21799 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
21800 defined, regardless of which processor variant is the target.
21802 @item -mdisable-callt
21803 @itemx -mno-disable-callt
21804 @opindex mdisable-callt
21805 @opindex mno-disable-callt
21806 This option suppresses generation of the @code{CALLT} instruction for the
21807 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21810 This option is enabled by default when the RH850 ABI is
21811 in use (see @option{-mrh850-abi}), and disabled by default when the
21812 GCC ABI is in use. If @code{CALLT} instructions are being generated
21813 then the C preprocessor symbol @code{__V850_CALLT__} will be defined.
21819 Pass on (or do not pass on) the @option{-mrelax} command line option
21823 @itemx -mno-long-jumps
21824 @opindex mlong-jumps
21825 @opindex mno-long-jumps
21826 Disable (or re-enable) the generation of PC-relative jump instructions.
21829 @itemx -mhard-float
21830 @opindex msoft-float
21831 @opindex mhard-float
21832 Disable (or re-enable) the generation of hardware floating point
21833 instructions. This option is only significant when the target
21834 architecture is @samp{V850E2V3} or higher. If hardware floating point
21835 instructions are being generated then the C preprocessor symbol
21836 @code{__FPU_OK__} will be defined, otherwise the symbol
21837 @code{__NO_FPU__} will be defined.
21841 Enables the use of the e3v5 LOOP instruction. The use of this
21842 instruction is not enabled by default when the e3v5 architecture is
21843 selected because its use is still experimental.
21847 @opindex mrh850-abi
21849 Enables support for the RH850 version of the V850 ABI. This is the
21850 default. With this version of the ABI the following rules apply:
21854 Integer sized structures and unions are returned via a memory pointer
21855 rather than a register.
21858 Large structures and unions (more than 8 bytes in size) are passed by
21862 Functions are aligned to 16-bit boundaries.
21865 The @option{-m8byte-align} command line option is supported.
21868 The @option{-mdisable-callt} command line option is enabled by
21869 default. The @option{-mno-disable-callt} command line option is not
21873 When this version of the ABI is enabled the C preprocessor symbol
21874 @code{__V850_RH850_ABI__} is defined.
21878 Enables support for the old GCC version of the V850 ABI. With this
21879 version of the ABI the following rules apply:
21883 Integer sized structures and unions are returned in register @code{r10}.
21886 Large structures and unions (more than 8 bytes in size) are passed by
21890 Functions are aligned to 32-bit boundaries, unless optimizing for
21894 The @option{-m8byte-align} command line option is not supported.
21897 The @option{-mdisable-callt} command line option is supported but not
21898 enabled by default.
21901 When this version of the ABI is enabled the C preprocessor symbol
21902 @code{__V850_GCC_ABI__} is defined.
21904 @item -m8byte-align
21905 @itemx -mno-8byte-align
21906 @opindex m8byte-align
21907 @opindex mno-8byte-align
21908 Enables support for @code{doubles} and @code{long long} types to be
21909 aligned on 8-byte boundaries. The default is to restrict the
21910 alignment of all objects to at most 4-bytes. When
21911 @option{-m8byte-align} is in effect the C preprocessor symbol
21912 @code{__V850_8BYTE_ALIGN__} will be defined.
21915 @opindex mbig-switch
21916 Generate code suitable for big switch tables. Use this option only if
21917 the assembler/linker complain about out of range branches within a switch
21922 This option causes r2 and r5 to be used in the code generated by
21923 the compiler. This setting is the default.
21925 @item -mno-app-regs
21926 @opindex mno-app-regs
21927 This option causes r2 and r5 to be treated as fixed registers.
21932 @subsection VAX Options
21933 @cindex VAX options
21935 These @samp{-m} options are defined for the VAX:
21940 Do not output certain jump instructions (@code{aobleq} and so on)
21941 that the Unix assembler for the VAX cannot handle across long
21946 Do output those jump instructions, on the assumption that the
21947 GNU assembler is being used.
21951 Output code for G-format floating-point numbers instead of D-format.
21955 @subsection VMS Options
21957 These @samp{-m} options are defined for the VMS implementations:
21960 @item -mvms-return-codes
21961 @opindex mvms-return-codes
21962 Return VMS condition codes from @code{main}. The default is to return POSIX-style
21963 condition (e.g.@ error) codes.
21965 @item -mdebug-main=@var{prefix}
21966 @opindex mdebug-main=@var{prefix}
21967 Flag the first routine whose name starts with @var{prefix} as the main
21968 routine for the debugger.
21972 Default to 64-bit memory allocation routines.
21974 @item -mpointer-size=@var{size}
21975 @opindex -mpointer-size=@var{size}
21976 Set the default size of pointers. Possible options for @var{size} are
21977 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
21978 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
21979 The later option disables @code{pragma pointer_size}.
21982 @node VxWorks Options
21983 @subsection VxWorks Options
21984 @cindex VxWorks Options
21986 The options in this section are defined for all VxWorks targets.
21987 Options specific to the target hardware are listed with the other
21988 options for that target.
21993 GCC can generate code for both VxWorks kernels and real time processes
21994 (RTPs). This option switches from the former to the latter. It also
21995 defines the preprocessor macro @code{__RTP__}.
21998 @opindex non-static
21999 Link an RTP executable against shared libraries rather than static
22000 libraries. The options @option{-static} and @option{-shared} can
22001 also be used for RTPs (@pxref{Link Options}); @option{-static}
22008 These options are passed down to the linker. They are defined for
22009 compatibility with Diab.
22012 @opindex Xbind-lazy
22013 Enable lazy binding of function calls. This option is equivalent to
22014 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22018 Disable lazy binding of function calls. This option is the default and
22019 is defined for compatibility with Diab.
22022 @node x86-64 Options
22023 @subsection x86-64 Options
22024 @cindex x86-64 options
22026 These are listed under @xref{i386 and x86-64 Options}.
22028 @node Xstormy16 Options
22029 @subsection Xstormy16 Options
22030 @cindex Xstormy16 Options
22032 These options are defined for Xstormy16:
22037 Choose startup files and linker script suitable for the simulator.
22040 @node Xtensa Options
22041 @subsection Xtensa Options
22042 @cindex Xtensa Options
22044 These options are supported for Xtensa targets:
22048 @itemx -mno-const16
22050 @opindex mno-const16
22051 Enable or disable use of @code{CONST16} instructions for loading
22052 constant values. The @code{CONST16} instruction is currently not a
22053 standard option from Tensilica. When enabled, @code{CONST16}
22054 instructions are always used in place of the standard @code{L32R}
22055 instructions. The use of @code{CONST16} is enabled by default only if
22056 the @code{L32R} instruction is not available.
22059 @itemx -mno-fused-madd
22060 @opindex mfused-madd
22061 @opindex mno-fused-madd
22062 Enable or disable use of fused multiply/add and multiply/subtract
22063 instructions in the floating-point option. This has no effect if the
22064 floating-point option is not also enabled. Disabling fused multiply/add
22065 and multiply/subtract instructions forces the compiler to use separate
22066 instructions for the multiply and add/subtract operations. This may be
22067 desirable in some cases where strict IEEE 754-compliant results are
22068 required: the fused multiply add/subtract instructions do not round the
22069 intermediate result, thereby producing results with @emph{more} bits of
22070 precision than specified by the IEEE standard. Disabling fused multiply
22071 add/subtract instructions also ensures that the program output is not
22072 sensitive to the compiler's ability to combine multiply and add/subtract
22075 @item -mserialize-volatile
22076 @itemx -mno-serialize-volatile
22077 @opindex mserialize-volatile
22078 @opindex mno-serialize-volatile
22079 When this option is enabled, GCC inserts @code{MEMW} instructions before
22080 @code{volatile} memory references to guarantee sequential consistency.
22081 The default is @option{-mserialize-volatile}. Use
22082 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
22084 @item -mforce-no-pic
22085 @opindex mforce-no-pic
22086 For targets, like GNU/Linux, where all user-mode Xtensa code must be
22087 position-independent code (PIC), this option disables PIC for compiling
22090 @item -mtext-section-literals
22091 @itemx -mno-text-section-literals
22092 @opindex mtext-section-literals
22093 @opindex mno-text-section-literals
22094 Control the treatment of literal pools. The default is
22095 @option{-mno-text-section-literals}, which places literals in a separate
22096 section in the output file. This allows the literal pool to be placed
22097 in a data RAM/ROM, and it also allows the linker to combine literal
22098 pools from separate object files to remove redundant literals and
22099 improve code size. With @option{-mtext-section-literals}, the literals
22100 are interspersed in the text section in order to keep them as close as
22101 possible to their references. This may be necessary for large assembly
22104 @item -mtarget-align
22105 @itemx -mno-target-align
22106 @opindex mtarget-align
22107 @opindex mno-target-align
22108 When this option is enabled, GCC instructs the assembler to
22109 automatically align instructions to reduce branch penalties at the
22110 expense of some code density. The assembler attempts to widen density
22111 instructions to align branch targets and the instructions following call
22112 instructions. If there are not enough preceding safe density
22113 instructions to align a target, no widening is performed. The
22114 default is @option{-mtarget-align}. These options do not affect the
22115 treatment of auto-aligned instructions like @code{LOOP}, which the
22116 assembler always aligns, either by widening density instructions or
22117 by inserting NOP instructions.
22120 @itemx -mno-longcalls
22121 @opindex mlongcalls
22122 @opindex mno-longcalls
22123 When this option is enabled, GCC instructs the assembler to translate
22124 direct calls to indirect calls unless it can determine that the target
22125 of a direct call is in the range allowed by the call instruction. This
22126 translation typically occurs for calls to functions in other source
22127 files. Specifically, the assembler translates a direct @code{CALL}
22128 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
22129 The default is @option{-mno-longcalls}. This option should be used in
22130 programs where the call target can potentially be out of range. This
22131 option is implemented in the assembler, not the compiler, so the
22132 assembly code generated by GCC still shows direct call
22133 instructions---look at the disassembled object code to see the actual
22134 instructions. Note that the assembler uses an indirect call for
22135 every cross-file call, not just those that really are out of range.
22138 @node zSeries Options
22139 @subsection zSeries Options
22140 @cindex zSeries options
22142 These are listed under @xref{S/390 and zSeries Options}.
22144 @node Code Gen Options
22145 @section Options for Code Generation Conventions
22146 @cindex code generation conventions
22147 @cindex options, code generation
22148 @cindex run-time options
22150 These machine-independent options control the interface conventions
22151 used in code generation.
22153 Most of them have both positive and negative forms; the negative form
22154 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
22155 one of the forms is listed---the one that is not the default. You
22156 can figure out the other form by either removing @samp{no-} or adding
22160 @item -fbounds-check
22161 @opindex fbounds-check
22162 For front ends that support it, generate additional code to check that
22163 indices used to access arrays are within the declared range. This is
22164 currently only supported by the Java and Fortran front ends, where
22165 this option defaults to true and false respectively.
22167 @item -fstack-reuse=@var{reuse-level}
22168 @opindex fstack_reuse
22169 This option controls stack space reuse for user declared local/auto variables
22170 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
22171 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
22172 local variables and temporaries, @samp{named_vars} enables the reuse only for
22173 user defined local variables with names, and @samp{none} disables stack reuse
22174 completely. The default value is @samp{all}. The option is needed when the
22175 program extends the lifetime of a scoped local variable or a compiler generated
22176 temporary beyond the end point defined by the language. When a lifetime of
22177 a variable ends, and if the variable lives in memory, the optimizing compiler
22178 has the freedom to reuse its stack space with other temporaries or scoped
22179 local variables whose live range does not overlap with it. Legacy code extending
22180 local lifetime will likely to break with the stack reuse optimization.
22199 if (*p == 10) // out of scope use of local1
22210 A(int k) : i(k), j(k) @{ @}
22217 void foo(const A& ar)
22224 foo(A(10)); // temp object's lifetime ends when foo returns
22230 ap->i+= 10; // ap references out of scope temp whose space
22231 // is reused with a. What is the value of ap->i?
22236 The lifetime of a compiler generated temporary is well defined by the C++
22237 standard. When a lifetime of a temporary ends, and if the temporary lives
22238 in memory, the optimizing compiler has the freedom to reuse its stack
22239 space with other temporaries or scoped local variables whose live range
22240 does not overlap with it. However some of the legacy code relies on
22241 the behavior of older compilers in which temporaries' stack space is
22242 not reused, the aggressive stack reuse can lead to runtime errors. This
22243 option is used to control the temporary stack reuse optimization.
22247 This option generates traps for signed overflow on addition, subtraction,
22248 multiplication operations.
22252 This option instructs the compiler to assume that signed arithmetic
22253 overflow of addition, subtraction and multiplication wraps around
22254 using twos-complement representation. This flag enables some optimizations
22255 and disables others. This option is enabled by default for the Java
22256 front end, as required by the Java language specification.
22259 @opindex fexceptions
22260 Enable exception handling. Generates extra code needed to propagate
22261 exceptions. For some targets, this implies GCC generates frame
22262 unwind information for all functions, which can produce significant data
22263 size overhead, although it does not affect execution. If you do not
22264 specify this option, GCC enables it by default for languages like
22265 C++ that normally require exception handling, and disables it for
22266 languages like C that do not normally require it. However, you may need
22267 to enable this option when compiling C code that needs to interoperate
22268 properly with exception handlers written in C++. You may also wish to
22269 disable this option if you are compiling older C++ programs that don't
22270 use exception handling.
22272 @item -fnon-call-exceptions
22273 @opindex fnon-call-exceptions
22274 Generate code that allows trapping instructions to throw exceptions.
22275 Note that this requires platform-specific runtime support that does
22276 not exist everywhere. Moreover, it only allows @emph{trapping}
22277 instructions to throw exceptions, i.e.@: memory references or floating-point
22278 instructions. It does not allow exceptions to be thrown from
22279 arbitrary signal handlers such as @code{SIGALRM}.
22281 @item -fdelete-dead-exceptions
22282 @opindex fdelete-dead-exceptions
22283 Consider that instructions that may throw exceptions but don't otherwise
22284 contribute to the execution of the program can be optimized away.
22285 This option is enabled by default for the Ada front end, as permitted by
22286 the Ada language specification.
22287 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
22289 @item -funwind-tables
22290 @opindex funwind-tables
22291 Similar to @option{-fexceptions}, except that it just generates any needed
22292 static data, but does not affect the generated code in any other way.
22293 You normally do not need to enable this option; instead, a language processor
22294 that needs this handling enables it on your behalf.
22296 @item -fasynchronous-unwind-tables
22297 @opindex fasynchronous-unwind-tables
22298 Generate unwind table in DWARF 2 format, if supported by target machine. The
22299 table is exact at each instruction boundary, so it can be used for stack
22300 unwinding from asynchronous events (such as debugger or garbage collector).
22302 @item -fno-gnu-unique
22303 @opindex fno-gnu-unique
22304 On systems with recent GNU assembler and C library, the C++ compiler
22305 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
22306 of template static data members and static local variables in inline
22307 functions are unique even in the presence of @code{RTLD_LOCAL}; this
22308 is necessary to avoid problems with a library used by two different
22309 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
22310 therefore disagreeing with the other one about the binding of the
22311 symbol. But this causes @code{dlclose} to be ignored for affected
22312 DSOs; if your program relies on reinitialization of a DSO via
22313 @code{dlclose} and @code{dlopen}, you can use
22314 @option{-fno-gnu-unique}.
22316 @item -fpcc-struct-return
22317 @opindex fpcc-struct-return
22318 Return ``short'' @code{struct} and @code{union} values in memory like
22319 longer ones, rather than in registers. This convention is less
22320 efficient, but it has the advantage of allowing intercallability between
22321 GCC-compiled files and files compiled with other compilers, particularly
22322 the Portable C Compiler (pcc).
22324 The precise convention for returning structures in memory depends
22325 on the target configuration macros.
22327 Short structures and unions are those whose size and alignment match
22328 that of some integer type.
22330 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
22331 switch is not binary compatible with code compiled with the
22332 @option{-freg-struct-return} switch.
22333 Use it to conform to a non-default application binary interface.
22335 @item -freg-struct-return
22336 @opindex freg-struct-return
22337 Return @code{struct} and @code{union} values in registers when possible.
22338 This is more efficient for small structures than
22339 @option{-fpcc-struct-return}.
22341 If you specify neither @option{-fpcc-struct-return} nor
22342 @option{-freg-struct-return}, GCC defaults to whichever convention is
22343 standard for the target. If there is no standard convention, GCC
22344 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
22345 the principal compiler. In those cases, we can choose the standard, and
22346 we chose the more efficient register return alternative.
22348 @strong{Warning:} code compiled with the @option{-freg-struct-return}
22349 switch is not binary compatible with code compiled with the
22350 @option{-fpcc-struct-return} switch.
22351 Use it to conform to a non-default application binary interface.
22353 @item -fshort-enums
22354 @opindex fshort-enums
22355 Allocate to an @code{enum} type only as many bytes as it needs for the
22356 declared range of possible values. Specifically, the @code{enum} type
22357 is equivalent to the smallest integer type that has enough room.
22359 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
22360 code that is not binary compatible with code generated without that switch.
22361 Use it to conform to a non-default application binary interface.
22363 @item -fshort-double
22364 @opindex fshort-double
22365 Use the same size for @code{double} as for @code{float}.
22367 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
22368 code that is not binary compatible with code generated without that switch.
22369 Use it to conform to a non-default application binary interface.
22371 @item -fshort-wchar
22372 @opindex fshort-wchar
22373 Override the underlying type for @samp{wchar_t} to be @samp{short
22374 unsigned int} instead of the default for the target. This option is
22375 useful for building programs to run under WINE@.
22377 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
22378 code that is not binary compatible with code generated without that switch.
22379 Use it to conform to a non-default application binary interface.
22382 @opindex fno-common
22383 In C code, controls the placement of uninitialized global variables.
22384 Unix C compilers have traditionally permitted multiple definitions of
22385 such variables in different compilation units by placing the variables
22387 This is the behavior specified by @option{-fcommon}, and is the default
22388 for GCC on most targets.
22389 On the other hand, this behavior is not required by ISO C, and on some
22390 targets may carry a speed or code size penalty on variable references.
22391 The @option{-fno-common} option specifies that the compiler should place
22392 uninitialized global variables in the data section of the object file,
22393 rather than generating them as common blocks.
22394 This has the effect that if the same variable is declared
22395 (without @code{extern}) in two different compilations,
22396 you get a multiple-definition error when you link them.
22397 In this case, you must compile with @option{-fcommon} instead.
22398 Compiling with @option{-fno-common} is useful on targets for which
22399 it provides better performance, or if you wish to verify that the
22400 program will work on other systems that always treat uninitialized
22401 variable declarations this way.
22405 Ignore the @samp{#ident} directive.
22407 @item -finhibit-size-directive
22408 @opindex finhibit-size-directive
22409 Don't output a @code{.size} assembler directive, or anything else that
22410 would cause trouble if the function is split in the middle, and the
22411 two halves are placed at locations far apart in memory. This option is
22412 used when compiling @file{crtstuff.c}; you should not need to use it
22415 @item -fverbose-asm
22416 @opindex fverbose-asm
22417 Put extra commentary information in the generated assembly code to
22418 make it more readable. This option is generally only of use to those
22419 who actually need to read the generated assembly code (perhaps while
22420 debugging the compiler itself).
22422 @option{-fno-verbose-asm}, the default, causes the
22423 extra information to be omitted and is useful when comparing two assembler
22426 @item -frecord-gcc-switches
22427 @opindex frecord-gcc-switches
22428 This switch causes the command line used to invoke the
22429 compiler to be recorded into the object file that is being created.
22430 This switch is only implemented on some targets and the exact format
22431 of the recording is target and binary file format dependent, but it
22432 usually takes the form of a section containing ASCII text. This
22433 switch is related to the @option{-fverbose-asm} switch, but that
22434 switch only records information in the assembler output file as
22435 comments, so it never reaches the object file.
22436 See also @option{-grecord-gcc-switches} for another
22437 way of storing compiler options into the object file.
22441 @cindex global offset table
22443 Generate position-independent code (PIC) suitable for use in a shared
22444 library, if supported for the target machine. Such code accesses all
22445 constant addresses through a global offset table (GOT)@. The dynamic
22446 loader resolves the GOT entries when the program starts (the dynamic
22447 loader is not part of GCC; it is part of the operating system). If
22448 the GOT size for the linked executable exceeds a machine-specific
22449 maximum size, you get an error message from the linker indicating that
22450 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
22451 instead. (These maximums are 8k on the SPARC and 32k
22452 on the m68k and RS/6000. The 386 has no such limit.)
22454 Position-independent code requires special support, and therefore works
22455 only on certain machines. For the 386, GCC supports PIC for System V
22456 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
22457 position-independent.
22459 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22464 If supported for the target machine, emit position-independent code,
22465 suitable for dynamic linking and avoiding any limit on the size of the
22466 global offset table. This option makes a difference on the m68k,
22467 PowerPC and SPARC@.
22469 Position-independent code requires special support, and therefore works
22470 only on certain machines.
22472 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
22479 These options are similar to @option{-fpic} and @option{-fPIC}, but
22480 generated position independent code can be only linked into executables.
22481 Usually these options are used when @option{-pie} GCC option is
22482 used during linking.
22484 @option{-fpie} and @option{-fPIE} both define the macros
22485 @code{__pie__} and @code{__PIE__}. The macros have the value 1
22486 for @option{-fpie} and 2 for @option{-fPIE}.
22488 @item -fno-jump-tables
22489 @opindex fno-jump-tables
22490 Do not use jump tables for switch statements even where it would be
22491 more efficient than other code generation strategies. This option is
22492 of use in conjunction with @option{-fpic} or @option{-fPIC} for
22493 building code that forms part of a dynamic linker and cannot
22494 reference the address of a jump table. On some targets, jump tables
22495 do not require a GOT and this option is not needed.
22497 @item -ffixed-@var{reg}
22499 Treat the register named @var{reg} as a fixed register; generated code
22500 should never refer to it (except perhaps as a stack pointer, frame
22501 pointer or in some other fixed role).
22503 @var{reg} must be the name of a register. The register names accepted
22504 are machine-specific and are defined in the @code{REGISTER_NAMES}
22505 macro in the machine description macro file.
22507 This flag does not have a negative form, because it specifies a
22510 @item -fcall-used-@var{reg}
22511 @opindex fcall-used
22512 Treat the register named @var{reg} as an allocable register that is
22513 clobbered by function calls. It may be allocated for temporaries or
22514 variables that do not live across a call. Functions compiled this way
22515 do not save and restore the register @var{reg}.
22517 It is an error to use this flag with the frame pointer or stack pointer.
22518 Use of this flag for other registers that have fixed pervasive roles in
22519 the machine's execution model produces disastrous results.
22521 This flag does not have a negative form, because it specifies a
22524 @item -fcall-saved-@var{reg}
22525 @opindex fcall-saved
22526 Treat the register named @var{reg} as an allocable register saved by
22527 functions. It may be allocated even for temporaries or variables that
22528 live across a call. Functions compiled this way save and restore
22529 the register @var{reg} if they use it.
22531 It is an error to use this flag with the frame pointer or stack pointer.
22532 Use of this flag for other registers that have fixed pervasive roles in
22533 the machine's execution model produces disastrous results.
22535 A different sort of disaster results from the use of this flag for
22536 a register in which function values may be returned.
22538 This flag does not have a negative form, because it specifies a
22541 @item -fpack-struct[=@var{n}]
22542 @opindex fpack-struct
22543 Without a value specified, pack all structure members together without
22544 holes. When a value is specified (which must be a small power of two), pack
22545 structure members according to this value, representing the maximum
22546 alignment (that is, objects with default alignment requirements larger than
22547 this are output potentially unaligned at the next fitting location.
22549 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
22550 code that is not binary compatible with code generated without that switch.
22551 Additionally, it makes the code suboptimal.
22552 Use it to conform to a non-default application binary interface.
22554 @item -finstrument-functions
22555 @opindex finstrument-functions
22556 Generate instrumentation calls for entry and exit to functions. Just
22557 after function entry and just before function exit, the following
22558 profiling functions are called with the address of the current
22559 function and its call site. (On some platforms,
22560 @code{__builtin_return_address} does not work beyond the current
22561 function, so the call site information may not be available to the
22562 profiling functions otherwise.)
22565 void __cyg_profile_func_enter (void *this_fn,
22567 void __cyg_profile_func_exit (void *this_fn,
22571 The first argument is the address of the start of the current function,
22572 which may be looked up exactly in the symbol table.
22574 This instrumentation is also done for functions expanded inline in other
22575 functions. The profiling calls indicate where, conceptually, the
22576 inline function is entered and exited. This means that addressable
22577 versions of such functions must be available. If all your uses of a
22578 function are expanded inline, this may mean an additional expansion of
22579 code size. If you use @samp{extern inline} in your C code, an
22580 addressable version of such functions must be provided. (This is
22581 normally the case anyway, but if you get lucky and the optimizer always
22582 expands the functions inline, you might have gotten away without
22583 providing static copies.)
22585 A function may be given the attribute @code{no_instrument_function}, in
22586 which case this instrumentation is not done. This can be used, for
22587 example, for the profiling functions listed above, high-priority
22588 interrupt routines, and any functions from which the profiling functions
22589 cannot safely be called (perhaps signal handlers, if the profiling
22590 routines generate output or allocate memory).
22592 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
22593 @opindex finstrument-functions-exclude-file-list
22595 Set the list of functions that are excluded from instrumentation (see
22596 the description of @code{-finstrument-functions}). If the file that
22597 contains a function definition matches with one of @var{file}, then
22598 that function is not instrumented. The match is done on substrings:
22599 if the @var{file} parameter is a substring of the file name, it is
22600 considered to be a match.
22605 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
22609 excludes any inline function defined in files whose pathnames
22610 contain @code{/bits/stl} or @code{include/sys}.
22612 If, for some reason, you want to include letter @code{','} in one of
22613 @var{sym}, write @code{'\,'}. For example,
22614 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
22615 (note the single quote surrounding the option).
22617 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
22618 @opindex finstrument-functions-exclude-function-list
22620 This is similar to @code{-finstrument-functions-exclude-file-list},
22621 but this option sets the list of function names to be excluded from
22622 instrumentation. The function name to be matched is its user-visible
22623 name, such as @code{vector<int> blah(const vector<int> &)}, not the
22624 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
22625 match is done on substrings: if the @var{sym} parameter is a substring
22626 of the function name, it is considered to be a match. For C99 and C++
22627 extended identifiers, the function name must be given in UTF-8, not
22628 using universal character names.
22630 @item -fstack-check
22631 @opindex fstack-check
22632 Generate code to verify that you do not go beyond the boundary of the
22633 stack. You should specify this flag if you are running in an
22634 environment with multiple threads, but you only rarely need to specify it in
22635 a single-threaded environment since stack overflow is automatically
22636 detected on nearly all systems if there is only one stack.
22638 Note that this switch does not actually cause checking to be done; the
22639 operating system or the language runtime must do that. The switch causes
22640 generation of code to ensure that they see the stack being extended.
22642 You can additionally specify a string parameter: @code{no} means no
22643 checking, @code{generic} means force the use of old-style checking,
22644 @code{specific} means use the best checking method and is equivalent
22645 to bare @option{-fstack-check}.
22647 Old-style checking is a generic mechanism that requires no specific
22648 target support in the compiler but comes with the following drawbacks:
22652 Modified allocation strategy for large objects: they are always
22653 allocated dynamically if their size exceeds a fixed threshold.
22656 Fixed limit on the size of the static frame of functions: when it is
22657 topped by a particular function, stack checking is not reliable and
22658 a warning is issued by the compiler.
22661 Inefficiency: because of both the modified allocation strategy and the
22662 generic implementation, code performance is hampered.
22665 Note that old-style stack checking is also the fallback method for
22666 @code{specific} if no target support has been added in the compiler.
22668 @item -fstack-limit-register=@var{reg}
22669 @itemx -fstack-limit-symbol=@var{sym}
22670 @itemx -fno-stack-limit
22671 @opindex fstack-limit-register
22672 @opindex fstack-limit-symbol
22673 @opindex fno-stack-limit
22674 Generate code to ensure that the stack does not grow beyond a certain value,
22675 either the value of a register or the address of a symbol. If a larger
22676 stack is required, a signal is raised at run time. For most targets,
22677 the signal is raised before the stack overruns the boundary, so
22678 it is possible to catch the signal without taking special precautions.
22680 For instance, if the stack starts at absolute address @samp{0x80000000}
22681 and grows downwards, you can use the flags
22682 @option{-fstack-limit-symbol=__stack_limit} and
22683 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
22684 of 128KB@. Note that this may only work with the GNU linker.
22686 @item -fsplit-stack
22687 @opindex fsplit-stack
22688 Generate code to automatically split the stack before it overflows.
22689 The resulting program has a discontiguous stack which can only
22690 overflow if the program is unable to allocate any more memory. This
22691 is most useful when running threaded programs, as it is no longer
22692 necessary to calculate a good stack size to use for each thread. This
22693 is currently only implemented for the i386 and x86_64 back ends running
22696 When code compiled with @option{-fsplit-stack} calls code compiled
22697 without @option{-fsplit-stack}, there may not be much stack space
22698 available for the latter code to run. If compiling all code,
22699 including library code, with @option{-fsplit-stack} is not an option,
22700 then the linker can fix up these calls so that the code compiled
22701 without @option{-fsplit-stack} always has a large stack. Support for
22702 this is implemented in the gold linker in GNU binutils release 2.21
22705 @item -fleading-underscore
22706 @opindex fleading-underscore
22707 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
22708 change the way C symbols are represented in the object file. One use
22709 is to help link with legacy assembly code.
22711 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
22712 generate code that is not binary compatible with code generated without that
22713 switch. Use it to conform to a non-default application binary interface.
22714 Not all targets provide complete support for this switch.
22716 @item -ftls-model=@var{model}
22717 @opindex ftls-model
22718 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
22719 The @var{model} argument should be one of @code{global-dynamic},
22720 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
22721 Note that the choice is subject to optimization: the compiler may use
22722 a more efficient model for symbols not visible outside of the translation
22723 unit, or if @option{-fpic} is not given on the command line.
22725 The default without @option{-fpic} is @code{initial-exec}; with
22726 @option{-fpic} the default is @code{global-dynamic}.
22728 @item -fvisibility=@var{default|internal|hidden|protected}
22729 @opindex fvisibility
22730 Set the default ELF image symbol visibility to the specified option---all
22731 symbols are marked with this unless overridden within the code.
22732 Using this feature can very substantially improve linking and
22733 load times of shared object libraries, produce more optimized
22734 code, provide near-perfect API export and prevent symbol clashes.
22735 It is @strong{strongly} recommended that you use this in any shared objects
22738 Despite the nomenclature, @code{default} always means public; i.e.,
22739 available to be linked against from outside the shared object.
22740 @code{protected} and @code{internal} are pretty useless in real-world
22741 usage so the only other commonly used option is @code{hidden}.
22742 The default if @option{-fvisibility} isn't specified is
22743 @code{default}, i.e., make every
22744 symbol public---this causes the same behavior as previous versions of
22747 A good explanation of the benefits offered by ensuring ELF
22748 symbols have the correct visibility is given by ``How To Write
22749 Shared Libraries'' by Ulrich Drepper (which can be found at
22750 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
22751 solution made possible by this option to marking things hidden when
22752 the default is public is to make the default hidden and mark things
22753 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
22754 and @code{__attribute__ ((visibility("default")))} instead of
22755 @code{__declspec(dllexport)} you get almost identical semantics with
22756 identical syntax. This is a great boon to those working with
22757 cross-platform projects.
22759 For those adding visibility support to existing code, you may find
22760 @samp{#pragma GCC visibility} of use. This works by you enclosing
22761 the declarations you wish to set visibility for with (for example)
22762 @samp{#pragma GCC visibility push(hidden)} and
22763 @samp{#pragma GCC visibility pop}.
22764 Bear in mind that symbol visibility should be viewed @strong{as
22765 part of the API interface contract} and thus all new code should
22766 always specify visibility when it is not the default; i.e., declarations
22767 only for use within the local DSO should @strong{always} be marked explicitly
22768 as hidden as so to avoid PLT indirection overheads---making this
22769 abundantly clear also aids readability and self-documentation of the code.
22770 Note that due to ISO C++ specification requirements, @code{operator new} and
22771 @code{operator delete} must always be of default visibility.
22773 Be aware that headers from outside your project, in particular system
22774 headers and headers from any other library you use, may not be
22775 expecting to be compiled with visibility other than the default. You
22776 may need to explicitly say @samp{#pragma GCC visibility push(default)}
22777 before including any such headers.
22779 @samp{extern} declarations are not affected by @option{-fvisibility}, so
22780 a lot of code can be recompiled with @option{-fvisibility=hidden} with
22781 no modifications. However, this means that calls to @code{extern}
22782 functions with no explicit visibility use the PLT, so it is more
22783 effective to use @code{__attribute ((visibility))} and/or
22784 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
22785 declarations should be treated as hidden.
22787 Note that @option{-fvisibility} does affect C++ vague linkage
22788 entities. This means that, for instance, an exception class that is
22789 be thrown between DSOs must be explicitly marked with default
22790 visibility so that the @samp{type_info} nodes are unified between
22793 An overview of these techniques, their benefits and how to use them
22794 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
22796 @item -fstrict-volatile-bitfields
22797 @opindex fstrict-volatile-bitfields
22798 This option should be used if accesses to volatile bit-fields (or other
22799 structure fields, although the compiler usually honors those types
22800 anyway) should use a single access of the width of the
22801 field's type, aligned to a natural alignment if possible. For
22802 example, targets with memory-mapped peripheral registers might require
22803 all such accesses to be 16 bits wide; with this flag you can
22804 declare all peripheral bit-fields as @code{unsigned short} (assuming short
22805 is 16 bits on these targets) to force GCC to use 16-bit accesses
22806 instead of, perhaps, a more efficient 32-bit access.
22808 If this option is disabled, the compiler uses the most efficient
22809 instruction. In the previous example, that might be a 32-bit load
22810 instruction, even though that accesses bytes that do not contain
22811 any portion of the bit-field, or memory-mapped registers unrelated to
22812 the one being updated.
22814 In some cases, such as when the @code{packed} attribute is applied to a
22815 structure field, it may not be possible to access the field with a single
22816 read or write that is correctly aligned for the target machine. In this
22817 case GCC falls back to generating multiple accesses rather than code that
22818 will fault or truncate the result at run time.
22820 Note: Due to restrictions of the C/C++11 memory model, write accesses are
22821 not allowed to touch non bit-field members. It is therefore recommended
22822 to define all bits of the field's type as bit-field members.
22824 The default value of this option is determined by the application binary
22825 interface for the target processor.
22827 @item -fsync-libcalls
22828 @opindex fsync-libcalls
22829 This option controls whether any out-of-line instance of the @code{__sync}
22830 family of functions may be used to implement the C++11 @code{__atomic}
22831 family of functions.
22833 The default value of this option is enabled, thus the only useful form
22834 of the option is @option{-fno-sync-libcalls}. This option is used in
22835 the implementation of the @file{libatomic} runtime library.
22841 @node Environment Variables
22842 @section Environment Variables Affecting GCC
22843 @cindex environment variables
22845 @c man begin ENVIRONMENT
22846 This section describes several environment variables that affect how GCC
22847 operates. Some of them work by specifying directories or prefixes to use
22848 when searching for various kinds of files. Some are used to specify other
22849 aspects of the compilation environment.
22851 Note that you can also specify places to search using options such as
22852 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
22853 take precedence over places specified using environment variables, which
22854 in turn take precedence over those specified by the configuration of GCC@.
22855 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
22856 GNU Compiler Collection (GCC) Internals}.
22861 @c @itemx LC_COLLATE
22863 @c @itemx LC_MONETARY
22864 @c @itemx LC_NUMERIC
22869 @c @findex LC_COLLATE
22870 @findex LC_MESSAGES
22871 @c @findex LC_MONETARY
22872 @c @findex LC_NUMERIC
22876 These environment variables control the way that GCC uses
22877 localization information which allows GCC to work with different
22878 national conventions. GCC inspects the locale categories
22879 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
22880 so. These locale categories can be set to any value supported by your
22881 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
22882 Kingdom encoded in UTF-8.
22884 The @env{LC_CTYPE} environment variable specifies character
22885 classification. GCC uses it to determine the character boundaries in
22886 a string; this is needed for some multibyte encodings that contain quote
22887 and escape characters that are otherwise interpreted as a string
22890 The @env{LC_MESSAGES} environment variable specifies the language to
22891 use in diagnostic messages.
22893 If the @env{LC_ALL} environment variable is set, it overrides the value
22894 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
22895 and @env{LC_MESSAGES} default to the value of the @env{LANG}
22896 environment variable. If none of these variables are set, GCC
22897 defaults to traditional C English behavior.
22901 If @env{TMPDIR} is set, it specifies the directory to use for temporary
22902 files. GCC uses temporary files to hold the output of one stage of
22903 compilation which is to be used as input to the next stage: for example,
22904 the output of the preprocessor, which is the input to the compiler
22907 @item GCC_COMPARE_DEBUG
22908 @findex GCC_COMPARE_DEBUG
22909 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
22910 @option{-fcompare-debug} to the compiler driver. See the documentation
22911 of this option for more details.
22913 @item GCC_EXEC_PREFIX
22914 @findex GCC_EXEC_PREFIX
22915 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
22916 names of the subprograms executed by the compiler. No slash is added
22917 when this prefix is combined with the name of a subprogram, but you can
22918 specify a prefix that ends with a slash if you wish.
22920 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
22921 an appropriate prefix to use based on the pathname it is invoked with.
22923 If GCC cannot find the subprogram using the specified prefix, it
22924 tries looking in the usual places for the subprogram.
22926 The default value of @env{GCC_EXEC_PREFIX} is
22927 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
22928 the installed compiler. In many cases @var{prefix} is the value
22929 of @code{prefix} when you ran the @file{configure} script.
22931 Other prefixes specified with @option{-B} take precedence over this prefix.
22933 This prefix is also used for finding files such as @file{crt0.o} that are
22936 In addition, the prefix is used in an unusual way in finding the
22937 directories to search for header files. For each of the standard
22938 directories whose name normally begins with @samp{/usr/local/lib/gcc}
22939 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
22940 replacing that beginning with the specified prefix to produce an
22941 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
22942 @file{foo/bar} just before it searches the standard directory
22943 @file{/usr/local/lib/bar}.
22944 If a standard directory begins with the configured
22945 @var{prefix} then the value of @var{prefix} is replaced by
22946 @env{GCC_EXEC_PREFIX} when looking for header files.
22948 @item COMPILER_PATH
22949 @findex COMPILER_PATH
22950 The value of @env{COMPILER_PATH} is a colon-separated list of
22951 directories, much like @env{PATH}. GCC tries the directories thus
22952 specified when searching for subprograms, if it can't find the
22953 subprograms using @env{GCC_EXEC_PREFIX}.
22956 @findex LIBRARY_PATH
22957 The value of @env{LIBRARY_PATH} is a colon-separated list of
22958 directories, much like @env{PATH}. When configured as a native compiler,
22959 GCC tries the directories thus specified when searching for special
22960 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
22961 using GCC also uses these directories when searching for ordinary
22962 libraries for the @option{-l} option (but directories specified with
22963 @option{-L} come first).
22967 @cindex locale definition
22968 This variable is used to pass locale information to the compiler. One way in
22969 which this information is used is to determine the character set to be used
22970 when character literals, string literals and comments are parsed in C and C++.
22971 When the compiler is configured to allow multibyte characters,
22972 the following values for @env{LANG} are recognized:
22976 Recognize JIS characters.
22978 Recognize SJIS characters.
22980 Recognize EUCJP characters.
22983 If @env{LANG} is not defined, or if it has some other value, then the
22984 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
22985 recognize and translate multibyte characters.
22989 Some additional environment variables affect the behavior of the
22992 @include cppenv.texi
22996 @node Precompiled Headers
22997 @section Using Precompiled Headers
22998 @cindex precompiled headers
22999 @cindex speed of compilation
23001 Often large projects have many header files that are included in every
23002 source file. The time the compiler takes to process these header files
23003 over and over again can account for nearly all of the time required to
23004 build the project. To make builds faster, GCC allows you to
23005 @dfn{precompile} a header file.
23007 To create a precompiled header file, simply compile it as you would any
23008 other file, if necessary using the @option{-x} option to make the driver
23009 treat it as a C or C++ header file. You may want to use a
23010 tool like @command{make} to keep the precompiled header up-to-date when
23011 the headers it contains change.
23013 A precompiled header file is searched for when @code{#include} is
23014 seen in the compilation. As it searches for the included file
23015 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
23016 compiler looks for a precompiled header in each directory just before it
23017 looks for the include file in that directory. The name searched for is
23018 the name specified in the @code{#include} with @samp{.gch} appended. If
23019 the precompiled header file can't be used, it is ignored.
23021 For instance, if you have @code{#include "all.h"}, and you have
23022 @file{all.h.gch} in the same directory as @file{all.h}, then the
23023 precompiled header file is used if possible, and the original
23024 header is used otherwise.
23026 Alternatively, you might decide to put the precompiled header file in a
23027 directory and use @option{-I} to ensure that directory is searched
23028 before (or instead of) the directory containing the original header.
23029 Then, if you want to check that the precompiled header file is always
23030 used, you can put a file of the same name as the original header in this
23031 directory containing an @code{#error} command.
23033 This also works with @option{-include}. So yet another way to use
23034 precompiled headers, good for projects not designed with precompiled
23035 header files in mind, is to simply take most of the header files used by
23036 a project, include them from another header file, precompile that header
23037 file, and @option{-include} the precompiled header. If the header files
23038 have guards against multiple inclusion, they are skipped because
23039 they've already been included (in the precompiled header).
23041 If you need to precompile the same header file for different
23042 languages, targets, or compiler options, you can instead make a
23043 @emph{directory} named like @file{all.h.gch}, and put each precompiled
23044 header in the directory, perhaps using @option{-o}. It doesn't matter
23045 what you call the files in the directory; every precompiled header in
23046 the directory is considered. The first precompiled header
23047 encountered in the directory that is valid for this compilation is
23048 used; they're searched in no particular order.
23050 There are many other possibilities, limited only by your imagination,
23051 good sense, and the constraints of your build system.
23053 A precompiled header file can be used only when these conditions apply:
23057 Only one precompiled header can be used in a particular compilation.
23060 A precompiled header can't be used once the first C token is seen. You
23061 can have preprocessor directives before a precompiled header; you cannot
23062 include a precompiled header from inside another header.
23065 The precompiled header file must be produced for the same language as
23066 the current compilation. You can't use a C precompiled header for a C++
23070 The precompiled header file must have been produced by the same compiler
23071 binary as the current compilation is using.
23074 Any macros defined before the precompiled header is included must
23075 either be defined in the same way as when the precompiled header was
23076 generated, or must not affect the precompiled header, which usually
23077 means that they don't appear in the precompiled header at all.
23079 The @option{-D} option is one way to define a macro before a
23080 precompiled header is included; using a @code{#define} can also do it.
23081 There are also some options that define macros implicitly, like
23082 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
23085 @item If debugging information is output when using the precompiled
23086 header, using @option{-g} or similar, the same kind of debugging information
23087 must have been output when building the precompiled header. However,
23088 a precompiled header built using @option{-g} can be used in a compilation
23089 when no debugging information is being output.
23091 @item The same @option{-m} options must generally be used when building
23092 and using the precompiled header. @xref{Submodel Options},
23093 for any cases where this rule is relaxed.
23095 @item Each of the following options must be the same when building and using
23096 the precompiled header:
23098 @gccoptlist{-fexceptions}
23101 Some other command-line options starting with @option{-f},
23102 @option{-p}, or @option{-O} must be defined in the same way as when
23103 the precompiled header was generated. At present, it's not clear
23104 which options are safe to change and which are not; the safest choice
23105 is to use exactly the same options when generating and using the
23106 precompiled header. The following are known to be safe:
23108 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
23109 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
23110 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
23115 For all of these except the last, the compiler automatically
23116 ignores the precompiled header if the conditions aren't met. If you
23117 find an option combination that doesn't work and doesn't cause the
23118 precompiled header to be ignored, please consider filing a bug report,
23121 If you do use differing options when generating and using the
23122 precompiled header, the actual behavior is a mixture of the
23123 behavior for the options. For instance, if you use @option{-g} to
23124 generate the precompiled header but not when using it, you may or may
23125 not get debugging information for routines in the precompiled header.