1 @c Copyright (C) 1988-2015 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{} 1988-2015 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. @command{g++} accepts mostly the same options as @command{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 -fsized-deallocation @gol
191 -fstats -ftemplate-backtrace-limit=@var{n} @gol
192 -ftemplate-depth=@var{n} @gol
193 -fno-threadsafe-statics -fuse-cxa-atexit @gol
194 -fno-weak -nostdinc++ @gol
195 -fvisibility-inlines-hidden @gol
196 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
197 -fvtv-counts -fvtv-debug @gol
198 -fvisibility-ms-compat @gol
199 -fext-numeric-literals @gol
200 -Wabi=@var{n} -Wconversion-null -Wctor-dtor-privacy @gol
201 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wnarrowing @gol
202 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
203 -Weffc++ -Wstrict-null-sentinel @gol
204 -Wno-non-template-friend -Wold-style-cast @gol
205 -Woverloaded-virtual -Wno-pmf-conversions @gol
208 @item Objective-C and Objective-C++ Language Options
209 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
210 Objective-C and Objective-C++ Dialects}.
211 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
212 -fgnu-runtime -fnext-runtime @gol
213 -fno-nil-receivers @gol
214 -fobjc-abi-version=@var{n} @gol
215 -fobjc-call-cxx-cdtors @gol
216 -fobjc-direct-dispatch @gol
217 -fobjc-exceptions @gol
220 -fobjc-std=objc1 @gol
221 -fno-local-ivars @gol
222 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
223 -freplace-objc-classes @gol
226 -Wassign-intercept @gol
227 -Wno-protocol -Wselector @gol
228 -Wstrict-selector-match @gol
229 -Wundeclared-selector}
231 @item Language Independent Options
232 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
233 @gccoptlist{-fmessage-length=@var{n} @gol
234 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
235 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
236 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
238 @item Warning Options
239 @xref{Warning Options,,Options to Request or Suppress Warnings}.
240 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
241 -pedantic-errors @gol
242 -w -Wextra -Wall -Waddress -Waggregate-return @gol
243 -Waggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
245 -Wno-attributes -Wno-builtin-macro-redefined @gol
246 -Wc90-c99-compat -Wc99-c11-compat @gol
247 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
248 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
249 -Wconversion -Wcoverage-mismatch -Wdate-time -Wdelete-incomplete -Wno-cpp @gol
250 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
251 -Wdisabled-optimization @gol
252 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
253 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
254 -Wno-endif-labels -Werror -Werror=* @gol
255 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
256 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
257 -Wformat-security -Wformat-signedness -Wformat-y2k @gol
258 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
259 -Wignored-qualifiers -Wincompatible-pointer-types @gol
260 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
261 -Winit-self -Winline -Wno-int-conversion @gol
262 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
263 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
264 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
265 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args -Wmissing-braces @gol
266 -Wmissing-field-initializers -Wmissing-include-dirs @gol
267 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
268 -Wodr -Wno-overflow -Wopenmp-simd @gol
269 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
270 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
271 -Wpointer-arith -Wno-pointer-to-int-cast @gol
272 -Wredundant-decls -Wno-return-local-addr @gol
273 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
274 -Wshift-count-negative -Wshift-count-overflow @gol
275 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
276 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
277 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
278 -Wstrict-aliasing=n @gol -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
279 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
280 -Wsuggest-final-types @gol -Wsuggest-final-methods @gol -Wsuggest-override @gol
281 -Wmissing-format-attribute @gol
282 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
283 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
284 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
285 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
286 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
287 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
288 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
289 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
290 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
291 -Wzero-as-null-pointer-constant}
293 @item C and Objective-C-only Warning Options
294 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
295 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
296 -Wold-style-declaration -Wold-style-definition @gol
297 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
298 -Wdeclaration-after-statement -Wpointer-sign}
300 @item Debugging Options
301 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
302 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
303 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
304 -fasan-shadow-offset=@var{number} -fsanitize-undefined-trap-on-error @gol
305 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
306 -fdisable-ipa-@var{pass_name} @gol
307 -fdisable-rtl-@var{pass_name} @gol
308 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
309 -fdisable-tree-@var{pass_name} @gol
310 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
311 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
312 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
313 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
314 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
316 -fdump-statistics @gol
318 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
319 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
320 -fdump-tree-cfg -fdump-tree-alias @gol
322 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
323 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
324 -fdump-tree-gimple@r{[}-raw@r{]} @gol
325 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
326 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
327 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
328 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
329 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
330 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
331 -fdump-tree-nrv -fdump-tree-vect @gol
332 -fdump-tree-sink @gol
333 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
334 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
335 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
336 -fdump-tree-vtable-verify @gol
337 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
338 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
339 -fdump-final-insns=@var{file} @gol
340 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
341 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
342 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
343 -fenable-@var{kind}-@var{pass} @gol
344 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
345 -fdebug-types-section -fmem-report-wpa @gol
346 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
348 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
349 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
350 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
351 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
352 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
353 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
354 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
355 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
356 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
357 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
358 -fdebug-prefix-map=@var{old}=@var{new} @gol
359 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
360 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
361 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
362 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
363 -print-prog-name=@var{program} -print-search-dirs -Q @gol
364 -print-sysroot -print-sysroot-headers-suffix @gol
365 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
367 @item Optimization Options
368 @xref{Optimize Options,,Options that Control Optimization}.
369 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
370 -falign-jumps[=@var{n}] @gol
371 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
372 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
373 -fauto-inc-dec -fbranch-probabilities @gol
374 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
375 -fbtr-bb-exclusive -fcaller-saves @gol
376 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
377 -fcompare-elim -fcprop-registers -fcrossjumping @gol
378 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
379 -fcx-limited-range @gol
380 -fdata-sections -fdce -fdelayed-branch @gol
381 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
382 -fdevirtualize-at-ltrans -fdse @gol
383 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
384 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
385 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
386 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
387 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
388 -fif-conversion2 -findirect-inlining @gol
389 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
390 -finline-small-functions -fipa-cp -fipa-cp-clone @gol
391 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
392 -fira-algorithm=@var{algorithm} @gol
393 -fira-region=@var{region} -fira-hoist-pressure @gol
394 -fira-loop-pressure -fno-ira-share-save-slots @gol
395 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
396 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
397 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
398 -flive-range-shrinkage @gol
399 -floop-block -floop-interchange -floop-strip-mine @gol
400 -floop-unroll-and-jam -floop-nest-optimize @gol
401 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
402 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
403 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
404 -fmove-loop-invariants -fno-branch-count-reg @gol
405 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
406 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
407 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
408 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
409 -fomit-frame-pointer -foptimize-sibling-calls @gol
410 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
411 -fprefetch-loop-arrays -fprofile-report @gol
412 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
413 -fprofile-generate=@var{path} @gol
414 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
415 -fprofile-reorder-functions @gol
416 -freciprocal-math -free -frename-registers -freorder-blocks @gol
417 -freorder-blocks-and-partition -freorder-functions @gol
418 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
419 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
420 -fsched-spec-load -fsched-spec-load-dangerous @gol
421 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
422 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
423 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
424 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
425 -fschedule-fusion @gol
426 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
427 -fselective-scheduling -fselective-scheduling2 @gol
428 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
429 -fsemantic-interposition @gol
430 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
431 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt -fstack-protector @gol
432 -fstack-protector-all -fstack-protector-strong -fstrict-aliasing @gol
433 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
434 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
435 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
436 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
437 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
438 -ftree-loop-if-convert-stores -ftree-loop-im @gol
439 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
440 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
441 -ftree-loop-vectorize @gol
442 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
443 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
444 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
445 -ftree-vectorize -ftree-vrp @gol
446 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
447 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
448 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
449 -fweb -fwhole-program -fwpa -fuse-ld=@var{linker} -fuse-linker-plugin @gol
450 --param @var{name}=@var{value}
451 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
453 @item Preprocessor Options
454 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
455 @gccoptlist{-A@var{question}=@var{answer} @gol
456 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
457 -C -dD -dI -dM -dN @gol
458 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
459 -idirafter @var{dir} @gol
460 -include @var{file} -imacros @var{file} @gol
461 -iprefix @var{file} -iwithprefix @var{dir} @gol
462 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
463 -imultilib @var{dir} -isysroot @var{dir} @gol
464 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
465 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
466 -remap -trigraphs -undef -U@var{macro} @gol
467 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
469 @item Assembler Option
470 @xref{Assembler Options,,Passing Options to the Assembler}.
471 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
474 @xref{Link Options,,Options for Linking}.
475 @gccoptlist{@var{object-file-name} -l@var{library} @gol
476 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
477 -s -static -static-libgcc -static-libstdc++ @gol
478 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
479 -shared -shared-libgcc -symbolic @gol
480 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
481 -u @var{symbol} -z @var{keyword}}
483 @item Directory Options
484 @xref{Directory Options,,Options for Directory Search}.
485 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
486 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
487 --sysroot=@var{dir} --no-sysroot-suffix}
489 @item Machine Dependent Options
490 @xref{Submodel Options,,Hardware Models and Configurations}.
491 @c This list is ordered alphanumerically by subsection name.
492 @c Try and put the significant identifier (CPU or system) first,
493 @c so users have a clue at guessing where the ones they want will be.
495 @emph{AArch64 Options}
496 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
497 -mgeneral-regs-only @gol
498 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
500 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
501 -mtls-dialect=desc -mtls-dialect=traditional @gol
502 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
503 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
505 @emph{Adapteva Epiphany Options}
506 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
507 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
508 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
509 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
510 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
511 -msplit-vecmove-early -m1reg-@var{reg}}
514 @gccoptlist{-mbarrel-shifter @gol
515 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
516 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
517 -mea -mno-mpy -mmul32x16 -mmul64 @gol
518 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
519 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
520 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
521 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
522 -mucb-mcount -mvolatile-cache @gol
523 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
524 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
525 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
526 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
527 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
528 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
531 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
532 -mabi=@var{name} @gol
533 -mapcs-stack-check -mno-apcs-stack-check @gol
534 -mapcs-float -mno-apcs-float @gol
535 -mapcs-reentrant -mno-apcs-reentrant @gol
536 -msched-prolog -mno-sched-prolog @gol
537 -mlittle-endian -mbig-endian @gol
538 -mfloat-abi=@var{name} @gol
539 -mfp16-format=@var{name}
540 -mthumb-interwork -mno-thumb-interwork @gol
541 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
542 -mstructure-size-boundary=@var{n} @gol
543 -mabort-on-noreturn @gol
544 -mlong-calls -mno-long-calls @gol
545 -msingle-pic-base -mno-single-pic-base @gol
546 -mpic-register=@var{reg} @gol
547 -mnop-fun-dllimport @gol
548 -mpoke-function-name @gol
550 -mtpcs-frame -mtpcs-leaf-frame @gol
551 -mcaller-super-interworking -mcallee-super-interworking @gol
552 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
553 -mword-relocations @gol
554 -mfix-cortex-m3-ldrd @gol
555 -munaligned-access @gol
556 -mneon-for-64bits @gol
557 -mslow-flash-data @gol
558 -masm-syntax-unified @gol
562 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
563 -mcall-prologues -mint8 -mno-interrupts -mrelax @gol
564 -mstrict-X -mtiny-stack -Waddr-space-convert}
566 @emph{Blackfin Options}
567 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
568 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
569 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
570 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
571 -mno-id-shared-library -mshared-library-id=@var{n} @gol
572 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
573 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
574 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
578 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
579 -msim -msdata=@var{sdata-type}}
582 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
583 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
584 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
585 -mstack-align -mdata-align -mconst-align @gol
586 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
587 -melf -maout -melinux -mlinux -sim -sim2 @gol
588 -mmul-bug-workaround -mno-mul-bug-workaround}
591 @gccoptlist{-mmac @gol
592 -mcr16cplus -mcr16c @gol
593 -msim -mint32 -mbit-ops
594 -mdata-model=@var{model}}
596 @emph{Darwin Options}
597 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
598 -arch_only -bind_at_load -bundle -bundle_loader @gol
599 -client_name -compatibility_version -current_version @gol
601 -dependency-file -dylib_file -dylinker_install_name @gol
602 -dynamic -dynamiclib -exported_symbols_list @gol
603 -filelist -flat_namespace -force_cpusubtype_ALL @gol
604 -force_flat_namespace -headerpad_max_install_names @gol
606 -image_base -init -install_name -keep_private_externs @gol
607 -multi_module -multiply_defined -multiply_defined_unused @gol
608 -noall_load -no_dead_strip_inits_and_terms @gol
609 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
610 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
611 -private_bundle -read_only_relocs -sectalign @gol
612 -sectobjectsymbols -whyload -seg1addr @gol
613 -sectcreate -sectobjectsymbols -sectorder @gol
614 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
615 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
616 -segprot -segs_read_only_addr -segs_read_write_addr @gol
617 -single_module -static -sub_library -sub_umbrella @gol
618 -twolevel_namespace -umbrella -undefined @gol
619 -unexported_symbols_list -weak_reference_mismatches @gol
620 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
621 -mkernel -mone-byte-bool}
623 @emph{DEC Alpha Options}
624 @gccoptlist{-mno-fp-regs -msoft-float @gol
625 -mieee -mieee-with-inexact -mieee-conformant @gol
626 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
627 -mtrap-precision=@var{mode} -mbuild-constants @gol
628 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
629 -mbwx -mmax -mfix -mcix @gol
630 -mfloat-vax -mfloat-ieee @gol
631 -mexplicit-relocs -msmall-data -mlarge-data @gol
632 -msmall-text -mlarge-text @gol
633 -mmemory-latency=@var{time}}
636 @gccoptlist{-msmall-model -mno-lsim}
639 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
640 -mhard-float -msoft-float @gol
641 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
642 -mdouble -mno-double @gol
643 -mmedia -mno-media -mmuladd -mno-muladd @gol
644 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
645 -mlinked-fp -mlong-calls -malign-labels @gol
646 -mlibrary-pic -macc-4 -macc-8 @gol
647 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
648 -moptimize-membar -mno-optimize-membar @gol
649 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
650 -mvliw-branch -mno-vliw-branch @gol
651 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
652 -mno-nested-cond-exec -mtomcat-stats @gol
656 @emph{GNU/Linux Options}
657 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
658 -tno-android-cc -tno-android-ld}
660 @emph{H8/300 Options}
661 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
664 @gccoptlist{-march=@var{architecture-type} @gol
665 -mdisable-fpregs -mdisable-indexing @gol
666 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
667 -mfixed-range=@var{register-range} @gol
668 -mjump-in-delay -mlinker-opt -mlong-calls @gol
669 -mlong-load-store -mno-disable-fpregs @gol
670 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
671 -mno-jump-in-delay -mno-long-load-store @gol
672 -mno-portable-runtime -mno-soft-float @gol
673 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
674 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
675 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
676 -munix=@var{unix-std} -nolibdld -static -threads}
678 @emph{i386 and x86-64 Options}
679 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
680 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
681 -mfpmath=@var{unit} @gol
682 -masm=@var{dialect} -mno-fancy-math-387 @gol
683 -mno-fp-ret-in-387 -msoft-float @gol
684 -mno-wide-multiply -mrtd -malign-double @gol
685 -mpreferred-stack-boundary=@var{num} @gol
686 -mincoming-stack-boundary=@var{num} @gol
687 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
688 -mrecip -mrecip=@var{opt} @gol
689 -mvzeroupper -mprefer-avx128 @gol
690 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
691 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
692 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
693 -mclflushopt -mxsavec -mxsaves @gol
694 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
695 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mthreads @gol
696 -mno-align-stringops -minline-all-stringops @gol
697 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
698 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
699 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
700 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
701 -mregparm=@var{num} -msseregparm @gol
702 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
703 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
704 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
705 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
706 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
707 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
708 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
709 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
711 @emph{i386 and x86-64 Windows Options}
712 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
713 -mnop-fun-dllimport -mthread @gol
714 -municode -mwin32 -mwindows -fno-set-stack-executable}
717 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
718 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
719 -mconstant-gp -mauto-pic -mfused-madd @gol
720 -minline-float-divide-min-latency @gol
721 -minline-float-divide-max-throughput @gol
722 -mno-inline-float-divide @gol
723 -minline-int-divide-min-latency @gol
724 -minline-int-divide-max-throughput @gol
725 -mno-inline-int-divide @gol
726 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
727 -mno-inline-sqrt @gol
728 -mdwarf2-asm -mearly-stop-bits @gol
729 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
730 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
731 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
732 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
733 -msched-spec-ldc -msched-spec-control-ldc @gol
734 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
735 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
736 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
737 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
740 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
741 -msign-extend-enabled -muser-enabled}
743 @emph{M32R/D Options}
744 @gccoptlist{-m32r2 -m32rx -m32r @gol
746 -malign-loops -mno-align-loops @gol
747 -missue-rate=@var{number} @gol
748 -mbranch-cost=@var{number} @gol
749 -mmodel=@var{code-size-model-type} @gol
750 -msdata=@var{sdata-type} @gol
751 -mno-flush-func -mflush-func=@var{name} @gol
752 -mno-flush-trap -mflush-trap=@var{number} @gol
756 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
758 @emph{M680x0 Options}
759 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
760 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
761 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
762 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
763 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
764 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
765 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
766 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
770 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
771 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
772 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
773 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
774 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
777 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
778 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
779 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
780 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
783 @emph{MicroBlaze Options}
784 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
785 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
786 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
787 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
788 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
791 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
792 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
793 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
794 -mips16 -mno-mips16 -mflip-mips16 @gol
795 -minterlink-compressed -mno-interlink-compressed @gol
796 -minterlink-mips16 -mno-interlink-mips16 @gol
797 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
798 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
799 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
800 -mno-float -msingle-float -mdouble-float @gol
801 -modd-spreg -mno-odd-spreg @gol
802 -mabs=@var{mode} -mnan=@var{encoding} @gol
803 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
806 -mvirt -mno-virt @gol
808 -mmicromips -mno-micromips @gol
809 -mfpu=@var{fpu-type} @gol
810 -msmartmips -mno-smartmips @gol
811 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
812 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
813 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
814 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
815 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
816 -membedded-data -mno-embedded-data @gol
817 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
818 -mcode-readable=@var{setting} @gol
819 -msplit-addresses -mno-split-addresses @gol
820 -mexplicit-relocs -mno-explicit-relocs @gol
821 -mcheck-zero-division -mno-check-zero-division @gol
822 -mdivide-traps -mdivide-breaks @gol
823 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
824 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
825 -mfix-24k -mno-fix-24k @gol
826 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
827 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
828 -mfix-vr4120 -mno-fix-vr4120 @gol
829 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
830 -mflush-func=@var{func} -mno-flush-func @gol
831 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
832 -mfp-exceptions -mno-fp-exceptions @gol
833 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
834 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
837 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
838 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
839 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
840 -mno-base-addresses -msingle-exit -mno-single-exit}
842 @emph{MN10300 Options}
843 @gccoptlist{-mmult-bug -mno-mult-bug @gol
844 -mno-am33 -mam33 -mam33-2 -mam34 @gol
845 -mtune=@var{cpu-type} @gol
846 -mreturn-pointer-on-d0 @gol
847 -mno-crt0 -mrelax -mliw -msetlb}
850 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
852 @emph{MSP430 Options}
853 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
857 @gccoptlist{-mbig-endian -mlittle-endian @gol
858 -mreduced-regs -mfull-regs @gol
859 -mcmov -mno-cmov @gol
860 -mperf-ext -mno-perf-ext @gol
861 -mv3push -mno-v3push @gol
862 -m16bit -mno-16bit @gol
863 -mgp-direct -mno-gp-direct @gol
864 -misr-vector-size=@var{num} @gol
865 -mcache-block-size=@var{num} @gol
866 -march=@var{arch} @gol
867 -mforce-fp-as-gp -mforbid-fp-as-gp @gol
868 -mex9 -mctor-dtor -mrelax}
870 @emph{Nios II Options}
871 @gccoptlist{-G @var{num} -mgpopt -mno-gpopt -mel -meb @gol
872 -mno-bypass-cache -mbypass-cache @gol
873 -mno-cache-volatile -mcache-volatile @gol
874 -mno-fast-sw-div -mfast-sw-div @gol
875 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
876 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
877 -mcustom-fpu-cfg=@var{name} @gol
878 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
880 @emph{PDP-11 Options}
881 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
882 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
883 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
884 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
885 -mbranch-expensive -mbranch-cheap @gol
886 -munix-asm -mdec-asm}
888 @emph{picoChip Options}
889 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
890 -msymbol-as-address -mno-inefficient-warnings}
892 @emph{PowerPC Options}
893 See RS/6000 and PowerPC Options.
896 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78 @gol
897 -m64bit-doubles -m32bit-doubles}
899 @emph{RS/6000 and PowerPC Options}
900 @gccoptlist{-mcpu=@var{cpu-type} @gol
901 -mtune=@var{cpu-type} @gol
902 -mcmodel=@var{code-model} @gol
904 -maltivec -mno-altivec @gol
905 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
906 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
907 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
908 -mfprnd -mno-fprnd @gol
909 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
910 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
911 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
912 -malign-power -malign-natural @gol
913 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
914 -msingle-float -mdouble-float -msimple-fpu @gol
915 -mstring -mno-string -mupdate -mno-update @gol
916 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
917 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
918 -mstrict-align -mno-strict-align -mrelocatable @gol
919 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
920 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
921 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
922 -mprioritize-restricted-insns=@var{priority} @gol
923 -msched-costly-dep=@var{dependence_type} @gol
924 -minsert-sched-nops=@var{scheme} @gol
925 -mcall-sysv -mcall-netbsd @gol
926 -maix-struct-return -msvr4-struct-return @gol
927 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
928 -mblock-move-inline-limit=@var{num} @gol
929 -misel -mno-isel @gol
930 -misel=yes -misel=no @gol
932 -mspe=yes -mspe=no @gol
934 -mgen-cell-microcode -mwarn-cell-microcode @gol
935 -mvrsave -mno-vrsave @gol
936 -mmulhw -mno-mulhw @gol
937 -mdlmzb -mno-dlmzb @gol
938 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
939 -mprototype -mno-prototype @gol
940 -msim -mmvme -mads -myellowknife -memb -msdata @gol
941 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
942 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
943 -mno-recip-precision @gol
944 -mveclibabi=@var{type} -mfriz -mno-friz @gol
945 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
946 -msave-toc-indirect -mno-save-toc-indirect @gol
947 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
948 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
949 -mquad-memory -mno-quad-memory @gol
950 -mquad-memory-atomic -mno-quad-memory-atomic @gol
951 -mcompat-align-parm -mno-compat-align-parm @gol
952 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
953 -mupper-regs -mno-upper-regs}
956 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
958 -mbig-endian-data -mlittle-endian-data @gol
961 -mas100-syntax -mno-as100-syntax@gol
963 -mmax-constant-size=@gol
966 -mno-warn-multiple-fast-interrupts@gol
967 -msave-acc-in-interrupts}
969 @emph{S/390 and zSeries Options}
970 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
971 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
972 -mlong-double-64 -mlong-double-128 @gol
973 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
974 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
975 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
976 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
977 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
978 -mhotpatch[=@var{halfwords}] -mno-hotpatch}
981 @gccoptlist{-meb -mel @gol
985 -mscore5 -mscore5u -mscore7 -mscore7d}
988 @gccoptlist{-m1 -m2 -m2e @gol
989 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
991 -m4-nofpu -m4-single-only -m4-single -m4 @gol
992 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
993 -m5-64media -m5-64media-nofpu @gol
994 -m5-32media -m5-32media-nofpu @gol
995 -m5-compact -m5-compact-nofpu @gol
996 -mb -ml -mdalign -mrelax @gol
997 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
998 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
999 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1000 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1001 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
1002 -maccumulate-outgoing-args -minvalid-symbols @gol
1003 -matomic-model=@var{atomic-model} @gol
1004 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1005 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1006 -mpretend-cmove -mtas}
1008 @emph{Solaris 2 Options}
1009 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1012 @emph{SPARC Options}
1013 @gccoptlist{-mcpu=@var{cpu-type} @gol
1014 -mtune=@var{cpu-type} @gol
1015 -mcmodel=@var{code-model} @gol
1016 -mmemory-model=@var{mem-model} @gol
1017 -m32 -m64 -mapp-regs -mno-app-regs @gol
1018 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1019 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1020 -mhard-quad-float -msoft-quad-float @gol
1021 -mstack-bias -mno-stack-bias @gol
1022 -munaligned-doubles -mno-unaligned-doubles @gol
1023 -muser-mode -mno-user-mode @gol
1024 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1025 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1026 -mcbcond -mno-cbcond @gol
1027 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1028 -mfix-at697f -mfix-ut699}
1031 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1032 -msafe-dma -munsafe-dma @gol
1034 -msmall-mem -mlarge-mem -mstdmain @gol
1035 -mfixed-range=@var{register-range} @gol
1037 -maddress-space-conversion -mno-address-space-conversion @gol
1038 -mcache-size=@var{cache-size} @gol
1039 -matomic-updates -mno-atomic-updates}
1041 @emph{System V Options}
1042 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1044 @emph{TILE-Gx Options}
1045 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1046 -mcmodel=@var{code-model}}
1048 @emph{TILEPro Options}
1049 @gccoptlist{-mcpu=@var{cpu} -m32}
1052 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1053 -mprolog-function -mno-prolog-function -mspace @gol
1054 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1055 -mapp-regs -mno-app-regs @gol
1056 -mdisable-callt -mno-disable-callt @gol
1057 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1058 -mv850e -mv850 -mv850e3v5 @gol
1069 @gccoptlist{-mg -mgnu -munix}
1071 @emph{Visium Options}
1072 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1073 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1076 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1077 -mpointer-size=@var{size}}
1079 @emph{VxWorks Options}
1080 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1081 -Xbind-lazy -Xbind-now}
1083 @emph{x86-64 Options}
1084 See i386 and x86-64 Options.
1086 @emph{Xstormy16 Options}
1089 @emph{Xtensa Options}
1090 @gccoptlist{-mconst16 -mno-const16 @gol
1091 -mfused-madd -mno-fused-madd @gol
1093 -mserialize-volatile -mno-serialize-volatile @gol
1094 -mtext-section-literals -mno-text-section-literals @gol
1095 -mtarget-align -mno-target-align @gol
1096 -mlongcalls -mno-longcalls}
1098 @emph{zSeries Options}
1099 See S/390 and zSeries Options.
1101 @item Code Generation Options
1102 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1103 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1104 -ffixed-@var{reg} -fexceptions @gol
1105 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1106 -fasynchronous-unwind-tables @gol
1107 -fno-gnu-unique @gol
1108 -finhibit-size-directive -finstrument-functions @gol
1109 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1110 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1111 -fno-common -fno-ident @gol
1112 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1113 -fno-jump-tables @gol
1114 -frecord-gcc-switches @gol
1115 -freg-struct-return -fshort-enums @gol
1116 -fshort-double -fshort-wchar @gol
1117 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1118 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1119 -fno-stack-limit -fsplit-stack @gol
1120 -fleading-underscore -ftls-model=@var{model} @gol
1121 -fstack-reuse=@var{reuse_level} @gol
1122 -ftrapv -fwrapv -fbounds-check @gol
1123 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1124 -fstrict-volatile-bitfields -fsync-libcalls}
1128 @node Overall Options
1129 @section Options Controlling the Kind of Output
1131 Compilation can involve up to four stages: preprocessing, compilation
1132 proper, assembly and linking, always in that order. GCC is capable of
1133 preprocessing and compiling several files either into several
1134 assembler input files, or into one assembler input file; then each
1135 assembler input file produces an object file, and linking combines all
1136 the object files (those newly compiled, and those specified as input)
1137 into an executable file.
1139 @cindex file name suffix
1140 For any given input file, the file name suffix determines what kind of
1141 compilation is done:
1145 C source code that must be preprocessed.
1148 C source code that should not be preprocessed.
1151 C++ source code that should not be preprocessed.
1154 Objective-C source code. Note that you must link with the @file{libobjc}
1155 library to make an Objective-C program work.
1158 Objective-C source code that should not be preprocessed.
1162 Objective-C++ source code. Note that you must link with the @file{libobjc}
1163 library to make an Objective-C++ program work. Note that @samp{.M} refers
1164 to a literal capital M@.
1166 @item @var{file}.mii
1167 Objective-C++ source code that should not be preprocessed.
1170 C, C++, Objective-C or Objective-C++ header file to be turned into a
1171 precompiled header (default), or C, C++ header file to be turned into an
1172 Ada spec (via the @option{-fdump-ada-spec} switch).
1175 @itemx @var{file}.cp
1176 @itemx @var{file}.cxx
1177 @itemx @var{file}.cpp
1178 @itemx @var{file}.CPP
1179 @itemx @var{file}.c++
1181 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1182 the last two letters must both be literally @samp{x}. Likewise,
1183 @samp{.C} refers to a literal capital C@.
1187 Objective-C++ source code that must be preprocessed.
1189 @item @var{file}.mii
1190 Objective-C++ source code that should not be preprocessed.
1194 @itemx @var{file}.hp
1195 @itemx @var{file}.hxx
1196 @itemx @var{file}.hpp
1197 @itemx @var{file}.HPP
1198 @itemx @var{file}.h++
1199 @itemx @var{file}.tcc
1200 C++ header file to be turned into a precompiled header or Ada spec.
1203 @itemx @var{file}.for
1204 @itemx @var{file}.ftn
1205 Fixed form Fortran source code that should not be preprocessed.
1208 @itemx @var{file}.FOR
1209 @itemx @var{file}.fpp
1210 @itemx @var{file}.FPP
1211 @itemx @var{file}.FTN
1212 Fixed form Fortran source code that must be preprocessed (with the traditional
1215 @item @var{file}.f90
1216 @itemx @var{file}.f95
1217 @itemx @var{file}.f03
1218 @itemx @var{file}.f08
1219 Free form Fortran source code that should not be preprocessed.
1221 @item @var{file}.F90
1222 @itemx @var{file}.F95
1223 @itemx @var{file}.F03
1224 @itemx @var{file}.F08
1225 Free form Fortran source code that must be preprocessed (with the
1226 traditional preprocessor).
1231 @c FIXME: Descriptions of Java file types.
1237 @item @var{file}.ads
1238 Ada source code file that contains a library unit declaration (a
1239 declaration of a package, subprogram, or generic, or a generic
1240 instantiation), or a library unit renaming declaration (a package,
1241 generic, or subprogram renaming declaration). Such files are also
1244 @item @var{file}.adb
1245 Ada source code file containing a library unit body (a subprogram or
1246 package body). Such files are also called @dfn{bodies}.
1248 @c GCC also knows about some suffixes for languages not yet included:
1259 @itemx @var{file}.sx
1260 Assembler code that must be preprocessed.
1263 An object file to be fed straight into linking.
1264 Any file name with no recognized suffix is treated this way.
1268 You can specify the input language explicitly with the @option{-x} option:
1271 @item -x @var{language}
1272 Specify explicitly the @var{language} for the following input files
1273 (rather than letting the compiler choose a default based on the file
1274 name suffix). This option applies to all following input files until
1275 the next @option{-x} option. Possible values for @var{language} are:
1277 c c-header cpp-output
1278 c++ c++-header c++-cpp-output
1279 objective-c objective-c-header objective-c-cpp-output
1280 objective-c++ objective-c++-header objective-c++-cpp-output
1281 assembler assembler-with-cpp
1283 f77 f77-cpp-input f95 f95-cpp-input
1289 Turn off any specification of a language, so that subsequent files are
1290 handled according to their file name suffixes (as they are if @option{-x}
1291 has not been used at all).
1293 @item -pass-exit-codes
1294 @opindex pass-exit-codes
1295 Normally the @command{gcc} program exits with the code of 1 if any
1296 phase of the compiler returns a non-success return code. If you specify
1297 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1298 the numerically highest error produced by any phase returning an error
1299 indication. The C, C++, and Fortran front ends return 4 if an internal
1300 compiler error is encountered.
1303 If you only want some of the stages of compilation, you can use
1304 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1305 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1306 @command{gcc} is to stop. Note that some combinations (for example,
1307 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1312 Compile or assemble the source files, but do not link. The linking
1313 stage simply is not done. The ultimate output is in the form of an
1314 object file for each source file.
1316 By default, the object file name for a source file is made by replacing
1317 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1319 Unrecognized input files, not requiring compilation or assembly, are
1324 Stop after the stage of compilation proper; do not assemble. The output
1325 is in the form of an assembler code file for each non-assembler input
1328 By default, the assembler file name for a source file is made by
1329 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1331 Input files that don't require compilation are ignored.
1335 Stop after the preprocessing stage; do not run the compiler proper. The
1336 output is in the form of preprocessed source code, which is sent to the
1339 Input files that don't require preprocessing are ignored.
1341 @cindex output file option
1344 Place output in file @var{file}. This applies to whatever
1345 sort of output is being produced, whether it be an executable file,
1346 an object file, an assembler file or preprocessed C code.
1348 If @option{-o} is not specified, the default is to put an executable
1349 file in @file{a.out}, the object file for
1350 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1351 assembler file in @file{@var{source}.s}, a precompiled header file in
1352 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1357 Print (on standard error output) the commands executed to run the stages
1358 of compilation. Also print the version number of the compiler driver
1359 program and of the preprocessor and the compiler proper.
1363 Like @option{-v} except the commands are not executed and arguments
1364 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1365 This is useful for shell scripts to capture the driver-generated command lines.
1369 Use pipes rather than temporary files for communication between the
1370 various stages of compilation. This fails to work on some systems where
1371 the assembler is unable to read from a pipe; but the GNU assembler has
1376 Print (on the standard output) a description of the command-line options
1377 understood by @command{gcc}. If the @option{-v} option is also specified
1378 then @option{--help} is also passed on to the various processes
1379 invoked by @command{gcc}, so that they can display the command-line options
1380 they accept. If the @option{-Wextra} option has also been specified
1381 (prior to the @option{--help} option), then command-line options that
1382 have no documentation associated with them are also displayed.
1385 @opindex target-help
1386 Print (on the standard output) a description of target-specific command-line
1387 options for each tool. For some targets extra target-specific
1388 information may also be printed.
1390 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1391 Print (on the standard output) a description of the command-line
1392 options understood by the compiler that fit into all specified classes
1393 and qualifiers. These are the supported classes:
1396 @item @samp{optimizers}
1397 Display all of the optimization options supported by the
1400 @item @samp{warnings}
1401 Display all of the options controlling warning messages
1402 produced by the compiler.
1405 Display target-specific options. Unlike the
1406 @option{--target-help} option however, target-specific options of the
1407 linker and assembler are not displayed. This is because those
1408 tools do not currently support the extended @option{--help=} syntax.
1411 Display the values recognized by the @option{--param}
1414 @item @var{language}
1415 Display the options supported for @var{language}, where
1416 @var{language} is the name of one of the languages supported in this
1420 Display the options that are common to all languages.
1423 These are the supported qualifiers:
1426 @item @samp{undocumented}
1427 Display only those options that are undocumented.
1430 Display options taking an argument that appears after an equal
1431 sign in the same continuous piece of text, such as:
1432 @samp{--help=target}.
1434 @item @samp{separate}
1435 Display options taking an argument that appears as a separate word
1436 following the original option, such as: @samp{-o output-file}.
1439 Thus for example to display all the undocumented target-specific
1440 switches supported by the compiler, use:
1443 --help=target,undocumented
1446 The sense of a qualifier can be inverted by prefixing it with the
1447 @samp{^} character, so for example to display all binary warning
1448 options (i.e., ones that are either on or off and that do not take an
1449 argument) that have a description, use:
1452 --help=warnings,^joined,^undocumented
1455 The argument to @option{--help=} should not consist solely of inverted
1458 Combining several classes is possible, although this usually
1459 restricts the output so much that there is nothing to display. One
1460 case where it does work, however, is when one of the classes is
1461 @var{target}. For example, to display all the target-specific
1462 optimization options, use:
1465 --help=target,optimizers
1468 The @option{--help=} option can be repeated on the command line. Each
1469 successive use displays its requested class of options, skipping
1470 those that have already been displayed.
1472 If the @option{-Q} option appears on the command line before the
1473 @option{--help=} option, then the descriptive text displayed by
1474 @option{--help=} is changed. Instead of describing the displayed
1475 options, an indication is given as to whether the option is enabled,
1476 disabled or set to a specific value (assuming that the compiler
1477 knows this at the point where the @option{--help=} option is used).
1479 Here is a truncated example from the ARM port of @command{gcc}:
1482 % gcc -Q -mabi=2 --help=target -c
1483 The following options are target specific:
1485 -mabort-on-noreturn [disabled]
1489 The output is sensitive to the effects of previous command-line
1490 options, so for example it is possible to find out which optimizations
1491 are enabled at @option{-O2} by using:
1494 -Q -O2 --help=optimizers
1497 Alternatively you can discover which binary optimizations are enabled
1498 by @option{-O3} by using:
1501 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1502 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1503 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1506 @item -no-canonical-prefixes
1507 @opindex no-canonical-prefixes
1508 Do not expand any symbolic links, resolve references to @samp{/../}
1509 or @samp{/./}, or make the path absolute when generating a relative
1514 Display the version number and copyrights of the invoked GCC@.
1518 Invoke all subcommands under a wrapper program. The name of the
1519 wrapper program and its parameters are passed as a comma separated
1523 gcc -c t.c -wrapper gdb,--args
1527 This invokes all subprograms of @command{gcc} under
1528 @samp{gdb --args}, thus the invocation of @command{cc1} is
1529 @samp{gdb --args cc1 @dots{}}.
1531 @item -fplugin=@var{name}.so
1533 Load the plugin code in file @var{name}.so, assumed to be a
1534 shared object to be dlopen'd by the compiler. The base name of
1535 the shared object file is used to identify the plugin for the
1536 purposes of argument parsing (See
1537 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1538 Each plugin should define the callback functions specified in the
1541 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1542 @opindex fplugin-arg
1543 Define an argument called @var{key} with a value of @var{value}
1544 for the plugin called @var{name}.
1546 @item -fdump-ada-spec@r{[}-slim@r{]}
1547 @opindex fdump-ada-spec
1548 For C and C++ source and include files, generate corresponding Ada specs.
1549 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1550 GNAT User's Guide}, which provides detailed documentation on this feature.
1552 @item -fada-spec-parent=@var{unit}
1553 @opindex fada-spec-parent
1554 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1555 Ada specs as child units of parent @var{unit}.
1557 @item -fdump-go-spec=@var{file}
1558 @opindex fdump-go-spec
1559 For input files in any language, generate corresponding Go
1560 declarations in @var{file}. This generates Go @code{const},
1561 @code{type}, @code{var}, and @code{func} declarations which may be a
1562 useful way to start writing a Go interface to code written in some
1565 @include @value{srcdir}/../libiberty/at-file.texi
1569 @section Compiling C++ Programs
1571 @cindex suffixes for C++ source
1572 @cindex C++ source file suffixes
1573 C++ source files conventionally use one of the suffixes @samp{.C},
1574 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1575 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1576 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1577 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1578 files with these names and compiles them as C++ programs even if you
1579 call the compiler the same way as for compiling C programs (usually
1580 with the name @command{gcc}).
1584 However, the use of @command{gcc} does not add the C++ library.
1585 @command{g++} is a program that calls GCC and automatically specifies linking
1586 against the C++ library. It treats @samp{.c},
1587 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1588 files unless @option{-x} is used. This program is also useful when
1589 precompiling a C header file with a @samp{.h} extension for use in C++
1590 compilations. On many systems, @command{g++} is also installed with
1591 the name @command{c++}.
1593 @cindex invoking @command{g++}
1594 When you compile C++ programs, you may specify many of the same
1595 command-line options that you use for compiling programs in any
1596 language; or command-line options meaningful for C and related
1597 languages; or options that are meaningful only for C++ programs.
1598 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1599 explanations of options for languages related to C@.
1600 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1601 explanations of options that are meaningful only for C++ programs.
1603 @node C Dialect Options
1604 @section Options Controlling C Dialect
1605 @cindex dialect options
1606 @cindex language dialect options
1607 @cindex options, dialect
1609 The following options control the dialect of C (or languages derived
1610 from C, such as C++, Objective-C and Objective-C++) that the compiler
1614 @cindex ANSI support
1618 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1619 equivalent to @option{-std=c++98}.
1621 This turns off certain features of GCC that are incompatible with ISO
1622 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1623 such as the @code{asm} and @code{typeof} keywords, and
1624 predefined macros such as @code{unix} and @code{vax} that identify the
1625 type of system you are using. It also enables the undesirable and
1626 rarely used ISO trigraph feature. For the C compiler,
1627 it disables recognition of C++ style @samp{//} comments as well as
1628 the @code{inline} keyword.
1630 The alternate keywords @code{__asm__}, @code{__extension__},
1631 @code{__inline__} and @code{__typeof__} continue to work despite
1632 @option{-ansi}. You would not want to use them in an ISO C program, of
1633 course, but it is useful to put them in header files that might be included
1634 in compilations done with @option{-ansi}. Alternate predefined macros
1635 such as @code{__unix__} and @code{__vax__} are also available, with or
1636 without @option{-ansi}.
1638 The @option{-ansi} option does not cause non-ISO programs to be
1639 rejected gratuitously. For that, @option{-Wpedantic} is required in
1640 addition to @option{-ansi}. @xref{Warning Options}.
1642 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1643 option is used. Some header files may notice this macro and refrain
1644 from declaring certain functions or defining certain macros that the
1645 ISO standard doesn't call for; this is to avoid interfering with any
1646 programs that might use these names for other things.
1648 Functions that are normally built in but do not have semantics
1649 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1650 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1651 built-in functions provided by GCC}, for details of the functions
1656 Determine the language standard. @xref{Standards,,Language Standards
1657 Supported by GCC}, for details of these standard versions. This option
1658 is currently only supported when compiling C or C++.
1660 The compiler can accept several base standards, such as @samp{c90} or
1661 @samp{c++98}, and GNU dialects of those standards, such as
1662 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1663 compiler accepts all programs following that standard plus those
1664 using GNU extensions that do not contradict it. For example,
1665 @option{-std=c90} turns off certain features of GCC that are
1666 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1667 keywords, but not other GNU extensions that do not have a meaning in
1668 ISO C90, such as omitting the middle term of a @code{?:}
1669 expression. On the other hand, when a GNU dialect of a standard is
1670 specified, all features supported by the compiler are enabled, even when
1671 those features change the meaning of the base standard. As a result, some
1672 strict-conforming programs may be rejected. The particular standard
1673 is used by @option{-Wpedantic} to identify which features are GNU
1674 extensions given that version of the standard. For example
1675 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1676 comments, while @option{-std=gnu99 -Wpedantic} does not.
1678 A value for this option must be provided; possible values are
1684 Support all ISO C90 programs (certain GNU extensions that conflict
1685 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1687 @item iso9899:199409
1688 ISO C90 as modified in amendment 1.
1694 ISO C99. This standard is substantially completely supported, modulo
1695 bugs and floating-point issues
1696 (mainly but not entirely relating to optional C99 features from
1697 Annexes F and G). See
1698 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1699 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1704 ISO C11, the 2011 revision of the ISO C standard. This standard is
1705 substantially completely supported, modulo bugs, floating-point issues
1706 (mainly but not entirely relating to optional C11 features from
1707 Annexes F and G) and the optional Annexes K (Bounds-checking
1708 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1712 GNU dialect of ISO C90 (including some C99 features).
1716 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1720 GNU dialect of ISO C11. This is the default for C code.
1721 The name @samp{gnu1x} is deprecated.
1725 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1726 additional defect reports. Same as @option{-ansi} for C++ code.
1730 GNU dialect of @option{-std=c++98}. This is the default for
1735 The 2011 ISO C++ standard plus amendments.
1736 The name @samp{c++0x} is deprecated.
1740 GNU dialect of @option{-std=c++11}.
1741 The name @samp{gnu++0x} is deprecated.
1745 The 2014 ISO C++ standard plus amendments.
1746 The name @samp{c++1y} is deprecated.
1750 GNU dialect of @option{-std=c++14}.
1751 The name @samp{gnu++1y} is deprecated.
1754 The next revision of the ISO C++ standard, tentatively planned for
1755 2017. Support is highly experimental, and will almost certainly
1756 change in incompatible ways in future releases.
1759 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1760 and will almost certainly change in incompatible ways in future
1764 @item -fgnu89-inline
1765 @opindex fgnu89-inline
1766 The option @option{-fgnu89-inline} tells GCC to use the traditional
1767 GNU semantics for @code{inline} functions when in C99 mode.
1768 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1769 is accepted and ignored by GCC versions 4.1.3 up to but not including
1770 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1771 C99 mode. Using this option is roughly equivalent to adding the
1772 @code{gnu_inline} function attribute to all inline functions
1773 (@pxref{Function Attributes}).
1775 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1776 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1777 specifies the default behavior). This option was first supported in
1778 GCC 4.3. This option is not supported in @option{-std=c90} or
1779 @option{-std=gnu90} mode.
1781 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1782 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1783 in effect for @code{inline} functions. @xref{Common Predefined
1784 Macros,,,cpp,The C Preprocessor}.
1786 @item -aux-info @var{filename}
1788 Output to the given filename prototyped declarations for all functions
1789 declared and/or defined in a translation unit, including those in header
1790 files. This option is silently ignored in any language other than C@.
1792 Besides declarations, the file indicates, in comments, the origin of
1793 each declaration (source file and line), whether the declaration was
1794 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1795 @samp{O} for old, respectively, in the first character after the line
1796 number and the colon), and whether it came from a declaration or a
1797 definition (@samp{C} or @samp{F}, respectively, in the following
1798 character). In the case of function definitions, a K&R-style list of
1799 arguments followed by their declarations is also provided, inside
1800 comments, after the declaration.
1802 @item -fallow-parameterless-variadic-functions
1803 @opindex fallow-parameterless-variadic-functions
1804 Accept variadic functions without named parameters.
1806 Although it is possible to define such a function, this is not very
1807 useful as it is not possible to read the arguments. This is only
1808 supported for C as this construct is allowed by C++.
1812 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1813 keyword, so that code can use these words as identifiers. You can use
1814 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1815 instead. @option{-ansi} implies @option{-fno-asm}.
1817 In C++, this switch only affects the @code{typeof} keyword, since
1818 @code{asm} and @code{inline} are standard keywords. You may want to
1819 use the @option{-fno-gnu-keywords} flag instead, which has the same
1820 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1821 switch only affects the @code{asm} and @code{typeof} keywords, since
1822 @code{inline} is a standard keyword in ISO C99.
1825 @itemx -fno-builtin-@var{function}
1826 @opindex fno-builtin
1827 @cindex built-in functions
1828 Don't recognize built-in functions that do not begin with
1829 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1830 functions provided by GCC}, for details of the functions affected,
1831 including those which are not built-in functions when @option{-ansi} or
1832 @option{-std} options for strict ISO C conformance are used because they
1833 do not have an ISO standard meaning.
1835 GCC normally generates special code to handle certain built-in functions
1836 more efficiently; for instance, calls to @code{alloca} may become single
1837 instructions which adjust the stack directly, and calls to @code{memcpy}
1838 may become inline copy loops. The resulting code is often both smaller
1839 and faster, but since the function calls no longer appear as such, you
1840 cannot set a breakpoint on those calls, nor can you change the behavior
1841 of the functions by linking with a different library. In addition,
1842 when a function is recognized as a built-in function, GCC may use
1843 information about that function to warn about problems with calls to
1844 that function, or to generate more efficient code, even if the
1845 resulting code still contains calls to that function. For example,
1846 warnings are given with @option{-Wformat} for bad calls to
1847 @code{printf} when @code{printf} is built in and @code{strlen} is
1848 known not to modify global memory.
1850 With the @option{-fno-builtin-@var{function}} option
1851 only the built-in function @var{function} is
1852 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1853 function is named that is not built-in in this version of GCC, this
1854 option is ignored. There is no corresponding
1855 @option{-fbuiltin-@var{function}} option; if you wish to enable
1856 built-in functions selectively when using @option{-fno-builtin} or
1857 @option{-ffreestanding}, you may define macros such as:
1860 #define abs(n) __builtin_abs ((n))
1861 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1866 @cindex hosted environment
1868 Assert that compilation targets a hosted environment. This implies
1869 @option{-fbuiltin}. A hosted environment is one in which the
1870 entire standard library is available, and in which @code{main} has a return
1871 type of @code{int}. Examples are nearly everything except a kernel.
1872 This is equivalent to @option{-fno-freestanding}.
1874 @item -ffreestanding
1875 @opindex ffreestanding
1876 @cindex hosted environment
1878 Assert that compilation targets a freestanding environment. This
1879 implies @option{-fno-builtin}. A freestanding environment
1880 is one in which the standard library may not exist, and program startup may
1881 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1882 This is equivalent to @option{-fno-hosted}.
1884 @xref{Standards,,Language Standards Supported by GCC}, for details of
1885 freestanding and hosted environments.
1889 @cindex OpenMP parallel
1890 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1891 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1892 compiler generates parallel code according to the OpenMP Application
1893 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1894 implies @option{-pthread}, and thus is only supported on targets that
1895 have support for @option{-pthread}. @option{-fopenmp} implies
1896 @option{-fopenmp-simd}.
1899 @opindex fopenmp-simd
1902 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1903 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1908 @cindex Enable Cilk Plus
1909 Enable the usage of Cilk Plus language extension features for C/C++.
1910 When the option @option{-fcilkplus} is specified, enable the usage of
1911 the Cilk Plus Language extension features for C/C++. The present
1912 implementation follows ABI version 1.2. This is an experimental
1913 feature that is only partially complete, and whose interface may
1914 change in future versions of GCC as the official specification
1915 changes. Currently, all features but @code{_Cilk_for} have been
1920 When the option @option{-fgnu-tm} is specified, the compiler
1921 generates code for the Linux variant of Intel's current Transactional
1922 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1923 an experimental feature whose interface may change in future versions
1924 of GCC, as the official specification changes. Please note that not
1925 all architectures are supported for this feature.
1927 For more information on GCC's support for transactional memory,
1928 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1929 Transactional Memory Library}.
1931 Note that the transactional memory feature is not supported with
1932 non-call exceptions (@option{-fnon-call-exceptions}).
1934 @item -fms-extensions
1935 @opindex fms-extensions
1936 Accept some non-standard constructs used in Microsoft header files.
1938 In C++ code, this allows member names in structures to be similar
1939 to previous types declarations.
1948 Some cases of unnamed fields in structures and unions are only
1949 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1950 fields within structs/unions}, for details.
1952 Note that this option is off for all targets but i?86 and x86_64
1953 targets using ms-abi.
1955 @item -fplan9-extensions
1956 @opindex fplan9-extensions
1957 Accept some non-standard constructs used in Plan 9 code.
1959 This enables @option{-fms-extensions}, permits passing pointers to
1960 structures with anonymous fields to functions that expect pointers to
1961 elements of the type of the field, and permits referring to anonymous
1962 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1963 struct/union fields within structs/unions}, for details. This is only
1964 supported for C, not C++.
1968 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1969 options for strict ISO C conformance) implies @option{-trigraphs}.
1971 @cindex traditional C language
1972 @cindex C language, traditional
1974 @itemx -traditional-cpp
1975 @opindex traditional-cpp
1976 @opindex traditional
1977 Formerly, these options caused GCC to attempt to emulate a pre-standard
1978 C compiler. They are now only supported with the @option{-E} switch.
1979 The preprocessor continues to support a pre-standard mode. See the GNU
1980 CPP manual for details.
1982 @item -fcond-mismatch
1983 @opindex fcond-mismatch
1984 Allow conditional expressions with mismatched types in the second and
1985 third arguments. The value of such an expression is void. This option
1986 is not supported for C++.
1988 @item -flax-vector-conversions
1989 @opindex flax-vector-conversions
1990 Allow implicit conversions between vectors with differing numbers of
1991 elements and/or incompatible element types. This option should not be
1994 @item -funsigned-char
1995 @opindex funsigned-char
1996 Let the type @code{char} be unsigned, like @code{unsigned char}.
1998 Each kind of machine has a default for what @code{char} should
1999 be. It is either like @code{unsigned char} by default or like
2000 @code{signed char} by default.
2002 Ideally, a portable program should always use @code{signed char} or
2003 @code{unsigned char} when it depends on the signedness of an object.
2004 But many programs have been written to use plain @code{char} and
2005 expect it to be signed, or expect it to be unsigned, depending on the
2006 machines they were written for. This option, and its inverse, let you
2007 make such a program work with the opposite default.
2009 The type @code{char} is always a distinct type from each of
2010 @code{signed char} or @code{unsigned char}, even though its behavior
2011 is always just like one of those two.
2014 @opindex fsigned-char
2015 Let the type @code{char} be signed, like @code{signed char}.
2017 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2018 the negative form of @option{-funsigned-char}. Likewise, the option
2019 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2021 @item -fsigned-bitfields
2022 @itemx -funsigned-bitfields
2023 @itemx -fno-signed-bitfields
2024 @itemx -fno-unsigned-bitfields
2025 @opindex fsigned-bitfields
2026 @opindex funsigned-bitfields
2027 @opindex fno-signed-bitfields
2028 @opindex fno-unsigned-bitfields
2029 These options control whether a bit-field is signed or unsigned, when the
2030 declaration does not use either @code{signed} or @code{unsigned}. By
2031 default, such a bit-field is signed, because this is consistent: the
2032 basic integer types such as @code{int} are signed types.
2035 @node C++ Dialect Options
2036 @section Options Controlling C++ Dialect
2038 @cindex compiler options, C++
2039 @cindex C++ options, command-line
2040 @cindex options, C++
2041 This section describes the command-line options that are only meaningful
2042 for C++ programs. You can also use most of the GNU compiler options
2043 regardless of what language your program is in. For example, you
2044 might compile a file @file{firstClass.C} like this:
2047 g++ -g -frepo -O -c firstClass.C
2051 In this example, only @option{-frepo} is an option meant
2052 only for C++ programs; you can use the other options with any
2053 language supported by GCC@.
2055 Here is a list of options that are @emph{only} for compiling C++ programs:
2059 @item -fabi-version=@var{n}
2060 @opindex fabi-version
2061 Use version @var{n} of the C++ ABI@. The default is version 0.
2063 Version 0 refers to the version conforming most closely to
2064 the C++ ABI specification. Therefore, the ABI obtained using version 0
2065 will change in different versions of G++ as ABI bugs are fixed.
2067 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2069 Version 2 is the version of the C++ ABI that first appeared in G++
2070 3.4, and was the default through G++ 4.9.
2072 Version 3 corrects an error in mangling a constant address as a
2075 Version 4, which first appeared in G++ 4.5, implements a standard
2076 mangling for vector types.
2078 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2079 attribute const/volatile on function pointer types, decltype of a
2080 plain decl, and use of a function parameter in the declaration of
2083 Version 6, which first appeared in G++ 4.7, corrects the promotion
2084 behavior of C++11 scoped enums and the mangling of template argument
2085 packs, const/static_cast, prefix ++ and --, and a class scope function
2086 used as a template argument.
2088 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2089 builtin type and corrects the mangling of lambdas in default argument
2092 Version 8, which first appeared in G++ 4.9, corrects the substitution
2093 behavior of function types with function-cv-qualifiers.
2095 See also @option{-Wabi}.
2097 @item -fabi-compat-version=@var{n}
2098 @opindex fabi-compat-version
2099 Starting with GCC 4.5, on targets that support strong aliases, G++
2100 works around mangling changes by creating an alias with the correct
2101 mangled name when defining a symbol with an incorrect mangled name.
2102 This switch specifies which ABI version to use for the alias.
2104 With @option{-fabi-version=0} (the default), this defaults to 2. If
2105 another ABI version is explicitly selected, this defaults to 0.
2107 The compatibility version is also set by @option{-Wabi=@var{n}}.
2109 @item -fno-access-control
2110 @opindex fno-access-control
2111 Turn off all access checking. This switch is mainly useful for working
2112 around bugs in the access control code.
2116 Check that the pointer returned by @code{operator new} is non-null
2117 before attempting to modify the storage allocated. This check is
2118 normally unnecessary because the C++ standard specifies that
2119 @code{operator new} only returns @code{0} if it is declared
2120 @code{throw()}, in which case the compiler always checks the
2121 return value even without this option. In all other cases, when
2122 @code{operator new} has a non-empty exception specification, memory
2123 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2124 @samp{new (nothrow)}.
2126 @item -fconstexpr-depth=@var{n}
2127 @opindex fconstexpr-depth
2128 Set the maximum nested evaluation depth for C++11 constexpr functions
2129 to @var{n}. A limit is needed to detect endless recursion during
2130 constant expression evaluation. The minimum specified by the standard
2133 @item -fdeduce-init-list
2134 @opindex fdeduce-init-list
2135 Enable deduction of a template type parameter as
2136 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2139 template <class T> auto forward(T t) -> decltype (realfn (t))
2146 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2150 This deduction was implemented as a possible extension to the
2151 originally proposed semantics for the C++11 standard, but was not part
2152 of the final standard, so it is disabled by default. This option is
2153 deprecated, and may be removed in a future version of G++.
2155 @item -ffriend-injection
2156 @opindex ffriend-injection
2157 Inject friend functions into the enclosing namespace, so that they are
2158 visible outside the scope of the class in which they are declared.
2159 Friend functions were documented to work this way in the old Annotated
2160 C++ Reference Manual, and versions of G++ before 4.1 always worked
2161 that way. However, in ISO C++ a friend function that is not declared
2162 in an enclosing scope can only be found using argument dependent
2163 lookup. This option causes friends to be injected as they were in
2166 This option is for compatibility, and may be removed in a future
2169 @item -fno-elide-constructors
2170 @opindex fno-elide-constructors
2171 The C++ standard allows an implementation to omit creating a temporary
2172 that is only used to initialize another object of the same type.
2173 Specifying this option disables that optimization, and forces G++ to
2174 call the copy constructor in all cases.
2176 @item -fno-enforce-eh-specs
2177 @opindex fno-enforce-eh-specs
2178 Don't generate code to check for violation of exception specifications
2179 at run time. This option violates the C++ standard, but may be useful
2180 for reducing code size in production builds, much like defining
2181 @code{NDEBUG}. This does not give user code permission to throw
2182 exceptions in violation of the exception specifications; the compiler
2183 still optimizes based on the specifications, so throwing an
2184 unexpected exception results in undefined behavior at run time.
2186 @item -fextern-tls-init
2187 @itemx -fno-extern-tls-init
2188 @opindex fextern-tls-init
2189 @opindex fno-extern-tls-init
2190 The C++11 and OpenMP standards allow @code{thread_local} and
2191 @code{threadprivate} variables to have dynamic (runtime)
2192 initialization. To support this, any use of such a variable goes
2193 through a wrapper function that performs any necessary initialization.
2194 When the use and definition of the variable are in the same
2195 translation unit, this overhead can be optimized away, but when the
2196 use is in a different translation unit there is significant overhead
2197 even if the variable doesn't actually need dynamic initialization. If
2198 the programmer can be sure that no use of the variable in a
2199 non-defining TU needs to trigger dynamic initialization (either
2200 because the variable is statically initialized, or a use of the
2201 variable in the defining TU will be executed before any uses in
2202 another TU), they can avoid this overhead with the
2203 @option{-fno-extern-tls-init} option.
2205 On targets that support symbol aliases, the default is
2206 @option{-fextern-tls-init}. On targets that do not support symbol
2207 aliases, the default is @option{-fno-extern-tls-init}.
2210 @itemx -fno-for-scope
2212 @opindex fno-for-scope
2213 If @option{-ffor-scope} is specified, the scope of variables declared in
2214 a @i{for-init-statement} is limited to the @code{for} loop itself,
2215 as specified by the C++ standard.
2216 If @option{-fno-for-scope} is specified, the scope of variables declared in
2217 a @i{for-init-statement} extends to the end of the enclosing scope,
2218 as was the case in old versions of G++, and other (traditional)
2219 implementations of C++.
2221 If neither flag is given, the default is to follow the standard,
2222 but to allow and give a warning for old-style code that would
2223 otherwise be invalid, or have different behavior.
2225 @item -fno-gnu-keywords
2226 @opindex fno-gnu-keywords
2227 Do not recognize @code{typeof} as a keyword, so that code can use this
2228 word as an identifier. You can use the keyword @code{__typeof__} instead.
2229 @option{-ansi} implies @option{-fno-gnu-keywords}.
2231 @item -fno-implicit-templates
2232 @opindex fno-implicit-templates
2233 Never emit code for non-inline templates that are instantiated
2234 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2235 @xref{Template Instantiation}, for more information.
2237 @item -fno-implicit-inline-templates
2238 @opindex fno-implicit-inline-templates
2239 Don't emit code for implicit instantiations of inline templates, either.
2240 The default is to handle inlines differently so that compiles with and
2241 without optimization need the same set of explicit instantiations.
2243 @item -fno-implement-inlines
2244 @opindex fno-implement-inlines
2245 To save space, do not emit out-of-line copies of inline functions
2246 controlled by @code{#pragma implementation}. This causes linker
2247 errors if these functions are not inlined everywhere they are called.
2249 @item -fms-extensions
2250 @opindex fms-extensions
2251 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2252 int and getting a pointer to member function via non-standard syntax.
2254 @item -fno-nonansi-builtins
2255 @opindex fno-nonansi-builtins
2256 Disable built-in declarations of functions that are not mandated by
2257 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2258 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2261 @opindex fnothrow-opt
2262 Treat a @code{throw()} exception specification as if it were a
2263 @code{noexcept} specification to reduce or eliminate the text size
2264 overhead relative to a function with no exception specification. If
2265 the function has local variables of types with non-trivial
2266 destructors, the exception specification actually makes the
2267 function smaller because the EH cleanups for those variables can be
2268 optimized away. The semantic effect is that an exception thrown out of
2269 a function with such an exception specification results in a call
2270 to @code{terminate} rather than @code{unexpected}.
2272 @item -fno-operator-names
2273 @opindex fno-operator-names
2274 Do not treat the operator name keywords @code{and}, @code{bitand},
2275 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2276 synonyms as keywords.
2278 @item -fno-optional-diags
2279 @opindex fno-optional-diags
2280 Disable diagnostics that the standard says a compiler does not need to
2281 issue. Currently, the only such diagnostic issued by G++ is the one for
2282 a name having multiple meanings within a class.
2285 @opindex fpermissive
2286 Downgrade some diagnostics about nonconformant code from errors to
2287 warnings. Thus, using @option{-fpermissive} allows some
2288 nonconforming code to compile.
2290 @item -fno-pretty-templates
2291 @opindex fno-pretty-templates
2292 When an error message refers to a specialization of a function
2293 template, the compiler normally prints the signature of the
2294 template followed by the template arguments and any typedefs or
2295 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2296 rather than @code{void f(int)}) so that it's clear which template is
2297 involved. When an error message refers to a specialization of a class
2298 template, the compiler omits any template arguments that match
2299 the default template arguments for that template. If either of these
2300 behaviors make it harder to understand the error message rather than
2301 easier, you can use @option{-fno-pretty-templates} to disable them.
2305 Enable automatic template instantiation at link time. This option also
2306 implies @option{-fno-implicit-templates}. @xref{Template
2307 Instantiation}, for more information.
2311 Disable generation of information about every class with virtual
2312 functions for use by the C++ run-time type identification features
2313 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2314 of the language, you can save some space by using this flag. Note that
2315 exception handling uses the same information, but G++ generates it as
2316 needed. The @code{dynamic_cast} operator can still be used for casts that
2317 do not require run-time type information, i.e.@: casts to @code{void *} or to
2318 unambiguous base classes.
2320 @item -fsized-deallocation
2321 @opindex fsized-deallocation
2322 Enable the built-in global declarations
2324 void operator delete (void *, std::size_t) noexcept;
2325 void operator delete[] (void *, std::size_t) noexcept;
2327 as introduced in C++14. This is useful for user-defined replacement
2328 deallocation functions that, for example, use the size of the object
2329 to make deallocation faster. Enabled by default under
2330 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2331 warns about places that might want to add a definition.
2335 Emit statistics about front-end processing at the end of the compilation.
2336 This information is generally only useful to the G++ development team.
2338 @item -fstrict-enums
2339 @opindex fstrict-enums
2340 Allow the compiler to optimize using the assumption that a value of
2341 enumerated type can only be one of the values of the enumeration (as
2342 defined in the C++ standard; basically, a value that can be
2343 represented in the minimum number of bits needed to represent all the
2344 enumerators). This assumption may not be valid if the program uses a
2345 cast to convert an arbitrary integer value to the enumerated type.
2347 @item -ftemplate-backtrace-limit=@var{n}
2348 @opindex ftemplate-backtrace-limit
2349 Set the maximum number of template instantiation notes for a single
2350 warning or error to @var{n}. The default value is 10.
2352 @item -ftemplate-depth=@var{n}
2353 @opindex ftemplate-depth
2354 Set the maximum instantiation depth for template classes to @var{n}.
2355 A limit on the template instantiation depth is needed to detect
2356 endless recursions during template class instantiation. ANSI/ISO C++
2357 conforming programs must not rely on a maximum depth greater than 17
2358 (changed to 1024 in C++11). The default value is 900, as the compiler
2359 can run out of stack space before hitting 1024 in some situations.
2361 @item -fno-threadsafe-statics
2362 @opindex fno-threadsafe-statics
2363 Do not emit the extra code to use the routines specified in the C++
2364 ABI for thread-safe initialization of local statics. You can use this
2365 option to reduce code size slightly in code that doesn't need to be
2368 @item -fuse-cxa-atexit
2369 @opindex fuse-cxa-atexit
2370 Register destructors for objects with static storage duration with the
2371 @code{__cxa_atexit} function rather than the @code{atexit} function.
2372 This option is required for fully standards-compliant handling of static
2373 destructors, but only works if your C library supports
2374 @code{__cxa_atexit}.
2376 @item -fno-use-cxa-get-exception-ptr
2377 @opindex fno-use-cxa-get-exception-ptr
2378 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2379 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2380 if the runtime routine is not available.
2382 @item -fvisibility-inlines-hidden
2383 @opindex fvisibility-inlines-hidden
2384 This switch declares that the user does not attempt to compare
2385 pointers to inline functions or methods where the addresses of the two functions
2386 are taken in different shared objects.
2388 The effect of this is that GCC may, effectively, mark inline methods with
2389 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2390 appear in the export table of a DSO and do not require a PLT indirection
2391 when used within the DSO@. Enabling this option can have a dramatic effect
2392 on load and link times of a DSO as it massively reduces the size of the
2393 dynamic export table when the library makes heavy use of templates.
2395 The behavior of this switch is not quite the same as marking the
2396 methods as hidden directly, because it does not affect static variables
2397 local to the function or cause the compiler to deduce that
2398 the function is defined in only one shared object.
2400 You may mark a method as having a visibility explicitly to negate the
2401 effect of the switch for that method. For example, if you do want to
2402 compare pointers to a particular inline method, you might mark it as
2403 having default visibility. Marking the enclosing class with explicit
2404 visibility has no effect.
2406 Explicitly instantiated inline methods are unaffected by this option
2407 as their linkage might otherwise cross a shared library boundary.
2408 @xref{Template Instantiation}.
2410 @item -fvisibility-ms-compat
2411 @opindex fvisibility-ms-compat
2412 This flag attempts to use visibility settings to make GCC's C++
2413 linkage model compatible with that of Microsoft Visual Studio.
2415 The flag makes these changes to GCC's linkage model:
2419 It sets the default visibility to @code{hidden}, like
2420 @option{-fvisibility=hidden}.
2423 Types, but not their members, are not hidden by default.
2426 The One Definition Rule is relaxed for types without explicit
2427 visibility specifications that are defined in more than one
2428 shared object: those declarations are permitted if they are
2429 permitted when this option is not used.
2432 In new code it is better to use @option{-fvisibility=hidden} and
2433 export those classes that are intended to be externally visible.
2434 Unfortunately it is possible for code to rely, perhaps accidentally,
2435 on the Visual Studio behavior.
2437 Among the consequences of these changes are that static data members
2438 of the same type with the same name but defined in different shared
2439 objects are different, so changing one does not change the other;
2440 and that pointers to function members defined in different shared
2441 objects may not compare equal. When this flag is given, it is a
2442 violation of the ODR to define types with the same name differently.
2444 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2445 @opindex fvtable-verify
2446 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2447 feature that verifies at run time, for every virtual call, that
2448 the vtable pointer through which the call is made is valid for the type of
2449 the object, and has not been corrupted or overwritten. If an invalid vtable
2450 pointer is detected at run time, an error is reported and execution of the
2451 program is immediately halted.
2453 This option causes run-time data structures to be built at program startup,
2454 which are used for verifying the vtable pointers.
2455 The options @samp{std} and @samp{preinit}
2456 control the timing of when these data structures are built. In both cases the
2457 data structures are built before execution reaches @code{main}. Using
2458 @option{-fvtable-verify=std} causes the data structures to be built after
2459 shared libraries have been loaded and initialized.
2460 @option{-fvtable-verify=preinit} causes them to be built before shared
2461 libraries have been loaded and initialized.
2463 If this option appears multiple times in the command line with different
2464 values specified, @samp{none} takes highest priority over both @samp{std} and
2465 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2469 When used in conjunction with @option{-fvtable-verify=std} or
2470 @option{-fvtable-verify=preinit}, causes debug versions of the
2471 runtime functions for the vtable verification feature to be called.
2472 This flag also causes the compiler to log information about which
2473 vtable pointers it finds for each class.
2474 This information is written to a file named @file{vtv_set_ptr_data.log}
2475 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2476 if that is defined or the current working directory otherwise.
2478 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2479 file, be sure to delete any existing one.
2482 @opindex fvtv-counts
2483 This is a debugging flag. When used in conjunction with
2484 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2485 causes the compiler to keep track of the total number of virtual calls
2486 it encounters and the number of verifications it inserts. It also
2487 counts the number of calls to certain run-time library functions
2488 that it inserts and logs this information for each compilation unit.
2489 The compiler writes this information to a file named
2490 @file{vtv_count_data.log} in the directory named by the environment
2491 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2492 directory otherwise. It also counts the size of the vtable pointer sets
2493 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2494 in the same directory.
2496 Note: This feature @emph{appends} data to the log files. To get fresh log
2497 files, be sure to delete any existing ones.
2501 Do not use weak symbol support, even if it is provided by the linker.
2502 By default, G++ uses weak symbols if they are available. This
2503 option exists only for testing, and should not be used by end-users;
2504 it results in inferior code and has no benefits. This option may
2505 be removed in a future release of G++.
2509 Do not search for header files in the standard directories specific to
2510 C++, but do still search the other standard directories. (This option
2511 is used when building the C++ library.)
2514 In addition, these optimization, warning, and code generation options
2515 have meanings only for C++ programs:
2518 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2521 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2522 G++ to warn when it generates code that is probably not compatible with the
2523 vendor-neutral C++ ABI@. Since G++ now defaults to
2524 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2525 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2526 or an older compatibility version is selected (with
2527 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2529 Although an effort has been made to warn about
2530 all such cases, there are probably some cases that are not warned about,
2531 even though G++ is generating incompatible code. There may also be
2532 cases where warnings are emitted even though the code that is generated
2535 You should rewrite your code to avoid these warnings if you are
2536 concerned about the fact that code generated by G++ may not be binary
2537 compatible with code generated by other compilers.
2539 @option{-Wabi} can also be used with an explicit version number to
2540 warn about compatibility with a particular @option{-fabi-version}
2541 level, e.g. @option{-Wabi=2} to warn about changes relative to
2542 @option{-fabi-version=2}. Specifying a version number also sets
2543 @option{-fabi-compat-version=@var{n}}.
2545 The known incompatibilities in @option{-fabi-version=2} (which was the
2546 default from GCC 3.4 to 4.9) include:
2551 A template with a non-type template parameter of reference type was
2552 mangled incorrectly:
2555 template <int &> struct S @{@};
2559 This was fixed in @option{-fabi-version=3}.
2562 SIMD vector types declared using @code{__attribute ((vector_size))} were
2563 mangled in a non-standard way that does not allow for overloading of
2564 functions taking vectors of different sizes.
2566 The mangling was changed in @option{-fabi-version=4}.
2569 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2570 qualifiers, and @code{decltype} of a plain declaration was folded away.
2572 These mangling issues were fixed in @option{-fabi-version=5}.
2575 Scoped enumerators passed as arguments to a variadic function are
2576 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2577 On most targets this does not actually affect the parameter passing
2578 ABI, as there is no way to pass an argument smaller than @code{int}.
2580 Also, the ABI changed the mangling of template argument packs,
2581 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2582 a class scope function used as a template argument.
2584 These issues were corrected in @option{-fabi-version=6}.
2587 Lambdas in default argument scope were mangled incorrectly, and the
2588 ABI changed the mangling of @code{nullptr_t}.
2590 These issues were corrected in @option{-fabi-version=7}.
2593 When mangling a function type with function-cv-qualifiers, the
2594 un-qualified function type was incorrectly treated as a substitution
2597 This was fixed in @option{-fabi-version=8}.
2600 It also warns about psABI-related changes. The known psABI changes at this
2606 For SysV/x86-64, unions with @code{long double} members are
2607 passed in memory as specified in psABI. For example:
2617 @code{union U} is always passed in memory.
2621 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2622 @opindex Wctor-dtor-privacy
2623 @opindex Wno-ctor-dtor-privacy
2624 Warn when a class seems unusable because all the constructors or
2625 destructors in that class are private, and it has neither friends nor
2626 public static member functions. Also warn if there are no non-private
2627 methods, and there's at least one private member function that isn't
2628 a constructor or destructor.
2630 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2631 @opindex Wdelete-non-virtual-dtor
2632 @opindex Wno-delete-non-virtual-dtor
2633 Warn when @code{delete} is used to destroy an instance of a class that
2634 has virtual functions and non-virtual destructor. It is unsafe to delete
2635 an instance of a derived class through a pointer to a base class if the
2636 base class does not have a virtual destructor. This warning is enabled
2639 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2640 @opindex Wliteral-suffix
2641 @opindex Wno-literal-suffix
2642 Warn when a string or character literal is followed by a ud-suffix which does
2643 not begin with an underscore. As a conforming extension, GCC treats such
2644 suffixes as separate preprocessing tokens in order to maintain backwards
2645 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2649 #define __STDC_FORMAT_MACROS
2650 #include <inttypes.h>
2655 printf("My int64: %"PRId64"\n", i64);
2659 In this case, @code{PRId64} is treated as a separate preprocessing token.
2661 This warning is enabled by default.
2663 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2665 @opindex Wno-narrowing
2666 Warn when a narrowing conversion prohibited by C++11 occurs within
2670 int i = @{ 2.2 @}; // error: narrowing from double to int
2673 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2675 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses for
2676 non-constants the diagnostic required by the standard. Note that this
2677 does not affect the meaning of well-formed code; narrowing conversions
2678 are still considered ill-formed in SFINAE context.
2680 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2682 @opindex Wno-noexcept
2683 Warn when a noexcept-expression evaluates to false because of a call
2684 to a function that does not have a non-throwing exception
2685 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2686 the compiler to never throw an exception.
2688 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2689 @opindex Wnon-virtual-dtor
2690 @opindex Wno-non-virtual-dtor
2691 Warn when a class has virtual functions and an accessible non-virtual
2692 destructor itself or in an accessible polymorphic base class, in which
2693 case it is possible but unsafe to delete an instance of a derived
2694 class through a pointer to the class itself or base class. This
2695 warning is automatically enabled if @option{-Weffc++} is specified.
2697 @item -Wreorder @r{(C++ and Objective-C++ only)}
2699 @opindex Wno-reorder
2700 @cindex reordering, warning
2701 @cindex warning for reordering of member initializers
2702 Warn when the order of member initializers given in the code does not
2703 match the order in which they must be executed. For instance:
2709 A(): j (0), i (1) @{ @}
2714 The compiler rearranges the member initializers for @code{i}
2715 and @code{j} to match the declaration order of the members, emitting
2716 a warning to that effect. This warning is enabled by @option{-Wall}.
2718 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2719 @opindex fext-numeric-literals
2720 @opindex fno-ext-numeric-literals
2721 Accept imaginary, fixed-point, or machine-defined
2722 literal number suffixes as GNU extensions.
2723 When this option is turned off these suffixes are treated
2724 as C++11 user-defined literal numeric suffixes.
2725 This is on by default for all pre-C++11 dialects and all GNU dialects:
2726 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2727 @option{-std=gnu++14}.
2728 This option is off by default
2729 for ISO C++11 onwards (@option{-std=c++11}, ...).
2732 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2735 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2738 Warn about violations of the following style guidelines from Scott Meyers'
2739 @cite{Effective C++} series of books:
2743 Define a copy constructor and an assignment operator for classes
2744 with dynamically-allocated memory.
2747 Prefer initialization to assignment in constructors.
2750 Have @code{operator=} return a reference to @code{*this}.
2753 Don't try to return a reference when you must return an object.
2756 Distinguish between prefix and postfix forms of increment and
2757 decrement operators.
2760 Never overload @code{&&}, @code{||}, or @code{,}.
2764 This option also enables @option{-Wnon-virtual-dtor}, which is also
2765 one of the effective C++ recommendations. However, the check is
2766 extended to warn about the lack of virtual destructor in accessible
2767 non-polymorphic bases classes too.
2769 When selecting this option, be aware that the standard library
2770 headers do not obey all of these guidelines; use @samp{grep -v}
2771 to filter out those warnings.
2773 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2774 @opindex Wstrict-null-sentinel
2775 @opindex Wno-strict-null-sentinel
2776 Warn about the use of an uncasted @code{NULL} as sentinel. When
2777 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2778 to @code{__null}. Although it is a null pointer constant rather than a
2779 null pointer, it is guaranteed to be of the same size as a pointer.
2780 But this use is not portable across different compilers.
2782 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2783 @opindex Wno-non-template-friend
2784 @opindex Wnon-template-friend
2785 Disable warnings when non-templatized friend functions are declared
2786 within a template. Since the advent of explicit template specification
2787 support in G++, if the name of the friend is an unqualified-id (i.e.,
2788 @samp{friend foo(int)}), the C++ language specification demands that the
2789 friend declare or define an ordinary, nontemplate function. (Section
2790 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2791 could be interpreted as a particular specialization of a templatized
2792 function. Because this non-conforming behavior is no longer the default
2793 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2794 check existing code for potential trouble spots and is on by default.
2795 This new compiler behavior can be turned off with
2796 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2797 but disables the helpful warning.
2799 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2800 @opindex Wold-style-cast
2801 @opindex Wno-old-style-cast
2802 Warn if an old-style (C-style) cast to a non-void type is used within
2803 a C++ program. The new-style casts (@code{dynamic_cast},
2804 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2805 less vulnerable to unintended effects and much easier to search for.
2807 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2808 @opindex Woverloaded-virtual
2809 @opindex Wno-overloaded-virtual
2810 @cindex overloaded virtual function, warning
2811 @cindex warning for overloaded virtual function
2812 Warn when a function declaration hides virtual functions from a
2813 base class. For example, in:
2820 struct B: public A @{
2825 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2836 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2837 @opindex Wno-pmf-conversions
2838 @opindex Wpmf-conversions
2839 Disable the diagnostic for converting a bound pointer to member function
2842 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2843 @opindex Wsign-promo
2844 @opindex Wno-sign-promo
2845 Warn when overload resolution chooses a promotion from unsigned or
2846 enumerated type to a signed type, over a conversion to an unsigned type of
2847 the same size. Previous versions of G++ tried to preserve
2848 unsignedness, but the standard mandates the current behavior.
2851 @node Objective-C and Objective-C++ Dialect Options
2852 @section Options Controlling Objective-C and Objective-C++ Dialects
2854 @cindex compiler options, Objective-C and Objective-C++
2855 @cindex Objective-C and Objective-C++ options, command-line
2856 @cindex options, Objective-C and Objective-C++
2857 (NOTE: This manual does not describe the Objective-C and Objective-C++
2858 languages themselves. @xref{Standards,,Language Standards
2859 Supported by GCC}, for references.)
2861 This section describes the command-line options that are only meaningful
2862 for Objective-C and Objective-C++ programs. You can also use most of
2863 the language-independent GNU compiler options.
2864 For example, you might compile a file @file{some_class.m} like this:
2867 gcc -g -fgnu-runtime -O -c some_class.m
2871 In this example, @option{-fgnu-runtime} is an option meant only for
2872 Objective-C and Objective-C++ programs; you can use the other options with
2873 any language supported by GCC@.
2875 Note that since Objective-C is an extension of the C language, Objective-C
2876 compilations may also use options specific to the C front-end (e.g.,
2877 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2878 C++-specific options (e.g., @option{-Wabi}).
2880 Here is a list of options that are @emph{only} for compiling Objective-C
2881 and Objective-C++ programs:
2884 @item -fconstant-string-class=@var{class-name}
2885 @opindex fconstant-string-class
2886 Use @var{class-name} as the name of the class to instantiate for each
2887 literal string specified with the syntax @code{@@"@dots{}"}. The default
2888 class name is @code{NXConstantString} if the GNU runtime is being used, and
2889 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2890 @option{-fconstant-cfstrings} option, if also present, overrides the
2891 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2892 to be laid out as constant CoreFoundation strings.
2895 @opindex fgnu-runtime
2896 Generate object code compatible with the standard GNU Objective-C
2897 runtime. This is the default for most types of systems.
2899 @item -fnext-runtime
2900 @opindex fnext-runtime
2901 Generate output compatible with the NeXT runtime. This is the default
2902 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2903 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2906 @item -fno-nil-receivers
2907 @opindex fno-nil-receivers
2908 Assume that all Objective-C message dispatches (@code{[receiver
2909 message:arg]}) in this translation unit ensure that the receiver is
2910 not @code{nil}. This allows for more efficient entry points in the
2911 runtime to be used. This option is only available in conjunction with
2912 the NeXT runtime and ABI version 0 or 1.
2914 @item -fobjc-abi-version=@var{n}
2915 @opindex fobjc-abi-version
2916 Use version @var{n} of the Objective-C ABI for the selected runtime.
2917 This option is currently supported only for the NeXT runtime. In that
2918 case, Version 0 is the traditional (32-bit) ABI without support for
2919 properties and other Objective-C 2.0 additions. Version 1 is the
2920 traditional (32-bit) ABI with support for properties and other
2921 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2922 nothing is specified, the default is Version 0 on 32-bit target
2923 machines, and Version 2 on 64-bit target machines.
2925 @item -fobjc-call-cxx-cdtors
2926 @opindex fobjc-call-cxx-cdtors
2927 For each Objective-C class, check if any of its instance variables is a
2928 C++ object with a non-trivial default constructor. If so, synthesize a
2929 special @code{- (id) .cxx_construct} instance method which runs
2930 non-trivial default constructors on any such instance variables, in order,
2931 and then return @code{self}. Similarly, check if any instance variable
2932 is a C++ object with a non-trivial destructor, and if so, synthesize a
2933 special @code{- (void) .cxx_destruct} method which runs
2934 all such default destructors, in reverse order.
2936 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2937 methods thusly generated only operate on instance variables
2938 declared in the current Objective-C class, and not those inherited
2939 from superclasses. It is the responsibility of the Objective-C
2940 runtime to invoke all such methods in an object's inheritance
2941 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2942 by the runtime immediately after a new object instance is allocated;
2943 the @code{- (void) .cxx_destruct} methods are invoked immediately
2944 before the runtime deallocates an object instance.
2946 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2947 support for invoking the @code{- (id) .cxx_construct} and
2948 @code{- (void) .cxx_destruct} methods.
2950 @item -fobjc-direct-dispatch
2951 @opindex fobjc-direct-dispatch
2952 Allow fast jumps to the message dispatcher. On Darwin this is
2953 accomplished via the comm page.
2955 @item -fobjc-exceptions
2956 @opindex fobjc-exceptions
2957 Enable syntactic support for structured exception handling in
2958 Objective-C, similar to what is offered by C++ and Java. This option
2959 is required to use the Objective-C keywords @code{@@try},
2960 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2961 @code{@@synchronized}. This option is available with both the GNU
2962 runtime and the NeXT runtime (but not available in conjunction with
2963 the NeXT runtime on Mac OS X 10.2 and earlier).
2967 Enable garbage collection (GC) in Objective-C and Objective-C++
2968 programs. This option is only available with the NeXT runtime; the
2969 GNU runtime has a different garbage collection implementation that
2970 does not require special compiler flags.
2972 @item -fobjc-nilcheck
2973 @opindex fobjc-nilcheck
2974 For the NeXT runtime with version 2 of the ABI, check for a nil
2975 receiver in method invocations before doing the actual method call.
2976 This is the default and can be disabled using
2977 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2978 checked for nil in this way no matter what this flag is set to.
2979 Currently this flag does nothing when the GNU runtime, or an older
2980 version of the NeXT runtime ABI, is used.
2982 @item -fobjc-std=objc1
2984 Conform to the language syntax of Objective-C 1.0, the language
2985 recognized by GCC 4.0. This only affects the Objective-C additions to
2986 the C/C++ language; it does not affect conformance to C/C++ standards,
2987 which is controlled by the separate C/C++ dialect option flags. When
2988 this option is used with the Objective-C or Objective-C++ compiler,
2989 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2990 This is useful if you need to make sure that your Objective-C code can
2991 be compiled with older versions of GCC@.
2993 @item -freplace-objc-classes
2994 @opindex freplace-objc-classes
2995 Emit a special marker instructing @command{ld(1)} not to statically link in
2996 the resulting object file, and allow @command{dyld(1)} to load it in at
2997 run time instead. This is used in conjunction with the Fix-and-Continue
2998 debugging mode, where the object file in question may be recompiled and
2999 dynamically reloaded in the course of program execution, without the need
3000 to restart the program itself. Currently, Fix-and-Continue functionality
3001 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3006 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3007 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3008 compile time) with static class references that get initialized at load time,
3009 which improves run-time performance. Specifying the @option{-fzero-link} flag
3010 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3011 to be retained. This is useful in Zero-Link debugging mode, since it allows
3012 for individual class implementations to be modified during program execution.
3013 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3014 regardless of command-line options.
3016 @item -fno-local-ivars
3017 @opindex fno-local-ivars
3018 @opindex flocal-ivars
3019 By default instance variables in Objective-C can be accessed as if
3020 they were local variables from within the methods of the class they're
3021 declared in. This can lead to shadowing between instance variables
3022 and other variables declared either locally inside a class method or
3023 globally with the same name. Specifying the @option{-fno-local-ivars}
3024 flag disables this behavior thus avoiding variable shadowing issues.
3026 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3027 @opindex fivar-visibility
3028 Set the default instance variable visibility to the specified option
3029 so that instance variables declared outside the scope of any access
3030 modifier directives default to the specified visibility.
3034 Dump interface declarations for all classes seen in the source file to a
3035 file named @file{@var{sourcename}.decl}.
3037 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3038 @opindex Wassign-intercept
3039 @opindex Wno-assign-intercept
3040 Warn whenever an Objective-C assignment is being intercepted by the
3043 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3044 @opindex Wno-protocol
3046 If a class is declared to implement a protocol, a warning is issued for
3047 every method in the protocol that is not implemented by the class. The
3048 default behavior is to issue a warning for every method not explicitly
3049 implemented in the class, even if a method implementation is inherited
3050 from the superclass. If you use the @option{-Wno-protocol} option, then
3051 methods inherited from the superclass are considered to be implemented,
3052 and no warning is issued for them.
3054 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3056 @opindex Wno-selector
3057 Warn if multiple methods of different types for the same selector are
3058 found during compilation. The check is performed on the list of methods
3059 in the final stage of compilation. Additionally, a check is performed
3060 for each selector appearing in a @code{@@selector(@dots{})}
3061 expression, and a corresponding method for that selector has been found
3062 during compilation. Because these checks scan the method table only at
3063 the end of compilation, these warnings are not produced if the final
3064 stage of compilation is not reached, for example because an error is
3065 found during compilation, or because the @option{-fsyntax-only} option is
3068 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3069 @opindex Wstrict-selector-match
3070 @opindex Wno-strict-selector-match
3071 Warn if multiple methods with differing argument and/or return types are
3072 found for a given selector when attempting to send a message using this
3073 selector to a receiver of type @code{id} or @code{Class}. When this flag
3074 is off (which is the default behavior), the compiler omits such warnings
3075 if any differences found are confined to types that share the same size
3078 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3079 @opindex Wundeclared-selector
3080 @opindex Wno-undeclared-selector
3081 Warn if a @code{@@selector(@dots{})} expression referring to an
3082 undeclared selector is found. A selector is considered undeclared if no
3083 method with that name has been declared before the
3084 @code{@@selector(@dots{})} expression, either explicitly in an
3085 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3086 an @code{@@implementation} section. This option always performs its
3087 checks as soon as a @code{@@selector(@dots{})} expression is found,
3088 while @option{-Wselector} only performs its checks in the final stage of
3089 compilation. This also enforces the coding style convention
3090 that methods and selectors must be declared before being used.
3092 @item -print-objc-runtime-info
3093 @opindex print-objc-runtime-info
3094 Generate C header describing the largest structure that is passed by
3099 @node Language Independent Options
3100 @section Options to Control Diagnostic Messages Formatting
3101 @cindex options to control diagnostics formatting
3102 @cindex diagnostic messages
3103 @cindex message formatting
3105 Traditionally, diagnostic messages have been formatted irrespective of
3106 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3107 options described below
3108 to control the formatting algorithm for diagnostic messages,
3109 e.g.@: how many characters per line, how often source location
3110 information should be reported. Note that some language front ends may not
3111 honor these options.
3114 @item -fmessage-length=@var{n}
3115 @opindex fmessage-length
3116 Try to format error messages so that they fit on lines of about
3117 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3118 done; each error message appears on a single line. This is the
3119 default for all front ends.
3121 @item -fdiagnostics-show-location=once
3122 @opindex fdiagnostics-show-location
3123 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3124 reporter to emit source location information @emph{once}; that is, in
3125 case the message is too long to fit on a single physical line and has to
3126 be wrapped, the source location won't be emitted (as prefix) again,
3127 over and over, in subsequent continuation lines. This is the default
3130 @item -fdiagnostics-show-location=every-line
3131 Only meaningful in line-wrapping mode. Instructs the diagnostic
3132 messages reporter to emit the same source location information (as
3133 prefix) for physical lines that result from the process of breaking
3134 a message which is too long to fit on a single line.
3136 @item -fdiagnostics-color[=@var{WHEN}]
3137 @itemx -fno-diagnostics-color
3138 @opindex fdiagnostics-color
3139 @cindex highlight, color, colour
3140 @vindex GCC_COLORS @r{environment variable}
3141 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3142 or @samp{auto}. The default depends on how the compiler has been configured,
3143 it can be any of the above @var{WHEN} options or also @samp{never}
3144 if @env{GCC_COLORS} environment variable isn't present in the environment,
3145 and @samp{auto} otherwise.
3146 @samp{auto} means to use color only when the standard error is a terminal.
3147 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3148 aliases for @option{-fdiagnostics-color=always} and
3149 @option{-fdiagnostics-color=never}, respectively.
3151 The colors are defined by the environment variable @env{GCC_COLORS}.
3152 Its value is a colon-separated list of capabilities and Select Graphic
3153 Rendition (SGR) substrings. SGR commands are interpreted by the
3154 terminal or terminal emulator. (See the section in the documentation
3155 of your text terminal for permitted values and their meanings as
3156 character attributes.) These substring values are integers in decimal
3157 representation and can be concatenated with semicolons.
3158 Common values to concatenate include
3160 @samp{4} for underline,
3162 @samp{7} for inverse,
3163 @samp{39} for default foreground color,
3164 @samp{30} to @samp{37} for foreground colors,
3165 @samp{90} to @samp{97} for 16-color mode foreground colors,
3166 @samp{38;5;0} to @samp{38;5;255}
3167 for 88-color and 256-color modes foreground colors,
3168 @samp{49} for default background color,
3169 @samp{40} to @samp{47} for background colors,
3170 @samp{100} to @samp{107} for 16-color mode background colors,
3171 and @samp{48;5;0} to @samp{48;5;255}
3172 for 88-color and 256-color modes background colors.
3174 The default @env{GCC_COLORS} is
3176 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3179 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3180 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3181 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3182 string disables colors.
3183 Supported capabilities are as follows.
3187 @vindex error GCC_COLORS @r{capability}
3188 SGR substring for error: markers.
3191 @vindex warning GCC_COLORS @r{capability}
3192 SGR substring for warning: markers.
3195 @vindex note GCC_COLORS @r{capability}
3196 SGR substring for note: markers.
3199 @vindex caret GCC_COLORS @r{capability}
3200 SGR substring for caret line.
3203 @vindex locus GCC_COLORS @r{capability}
3204 SGR substring for location information, @samp{file:line} or
3205 @samp{file:line:column} etc.
3208 @vindex quote GCC_COLORS @r{capability}
3209 SGR substring for information printed within quotes.
3212 @item -fno-diagnostics-show-option
3213 @opindex fno-diagnostics-show-option
3214 @opindex fdiagnostics-show-option
3215 By default, each diagnostic emitted includes text indicating the
3216 command-line option that directly controls the diagnostic (if such an
3217 option is known to the diagnostic machinery). Specifying the
3218 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3220 @item -fno-diagnostics-show-caret
3221 @opindex fno-diagnostics-show-caret
3222 @opindex fdiagnostics-show-caret
3223 By default, each diagnostic emitted includes the original source line
3224 and a caret '^' indicating the column. This option suppresses this
3225 information. The source line is truncated to @var{n} characters, if
3226 the @option{-fmessage-length=n} is given. When the output is done
3227 to the terminal, the width is limited to the width given by the
3228 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3232 @node Warning Options
3233 @section Options to Request or Suppress Warnings
3234 @cindex options to control warnings
3235 @cindex warning messages
3236 @cindex messages, warning
3237 @cindex suppressing warnings
3239 Warnings are diagnostic messages that report constructions that
3240 are not inherently erroneous but that are risky or suggest there
3241 may have been an error.
3243 The following language-independent options do not enable specific
3244 warnings but control the kinds of diagnostics produced by GCC@.
3247 @cindex syntax checking
3249 @opindex fsyntax-only
3250 Check the code for syntax errors, but don't do anything beyond that.
3252 @item -fmax-errors=@var{n}
3253 @opindex fmax-errors
3254 Limits the maximum number of error messages to @var{n}, at which point
3255 GCC bails out rather than attempting to continue processing the source
3256 code. If @var{n} is 0 (the default), there is no limit on the number
3257 of error messages produced. If @option{-Wfatal-errors} is also
3258 specified, then @option{-Wfatal-errors} takes precedence over this
3263 Inhibit all warning messages.
3268 Make all warnings into errors.
3273 Make the specified warning into an error. The specifier for a warning
3274 is appended; for example @option{-Werror=switch} turns the warnings
3275 controlled by @option{-Wswitch} into errors. This switch takes a
3276 negative form, to be used to negate @option{-Werror} for specific
3277 warnings; for example @option{-Wno-error=switch} makes
3278 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3281 The warning message for each controllable warning includes the
3282 option that controls the warning. That option can then be used with
3283 @option{-Werror=} and @option{-Wno-error=} as described above.
3284 (Printing of the option in the warning message can be disabled using the
3285 @option{-fno-diagnostics-show-option} flag.)
3287 Note that specifying @option{-Werror=}@var{foo} automatically implies
3288 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3291 @item -Wfatal-errors
3292 @opindex Wfatal-errors
3293 @opindex Wno-fatal-errors
3294 This option causes the compiler to abort compilation on the first error
3295 occurred rather than trying to keep going and printing further error
3300 You can request many specific warnings with options beginning with
3301 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3302 implicit declarations. Each of these specific warning options also
3303 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3304 example, @option{-Wno-implicit}. This manual lists only one of the
3305 two forms, whichever is not the default. For further
3306 language-specific options also refer to @ref{C++ Dialect Options} and
3307 @ref{Objective-C and Objective-C++ Dialect Options}.
3309 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3310 options, such as @option{-Wunused}, which may turn on further options,
3311 such as @option{-Wunused-value}. The combined effect of positive and
3312 negative forms is that more specific options have priority over less
3313 specific ones, independently of their position in the command-line. For
3314 options of the same specificity, the last one takes effect. Options
3315 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3316 as if they appeared at the end of the command-line.
3318 When an unrecognized warning option is requested (e.g.,
3319 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3320 that the option is not recognized. However, if the @option{-Wno-} form
3321 is used, the behavior is slightly different: no diagnostic is
3322 produced for @option{-Wno-unknown-warning} unless other diagnostics
3323 are being produced. This allows the use of new @option{-Wno-} options
3324 with old compilers, but if something goes wrong, the compiler
3325 warns that an unrecognized option is present.
3332 Issue all the warnings demanded by strict ISO C and ISO C++;
3333 reject all programs that use forbidden extensions, and some other
3334 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3335 version of the ISO C standard specified by any @option{-std} option used.
3337 Valid ISO C and ISO C++ programs should compile properly with or without
3338 this option (though a rare few require @option{-ansi} or a
3339 @option{-std} option specifying the required version of ISO C)@. However,
3340 without this option, certain GNU extensions and traditional C and C++
3341 features are supported as well. With this option, they are rejected.
3343 @option{-Wpedantic} does not cause warning messages for use of the
3344 alternate keywords whose names begin and end with @samp{__}. Pedantic
3345 warnings are also disabled in the expression that follows
3346 @code{__extension__}. However, only system header files should use
3347 these escape routes; application programs should avoid them.
3348 @xref{Alternate Keywords}.
3350 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3351 C conformance. They soon find that it does not do quite what they want:
3352 it finds some non-ISO practices, but not all---only those for which
3353 ISO C @emph{requires} a diagnostic, and some others for which
3354 diagnostics have been added.
3356 A feature to report any failure to conform to ISO C might be useful in
3357 some instances, but would require considerable additional work and would
3358 be quite different from @option{-Wpedantic}. We don't have plans to
3359 support such a feature in the near future.
3361 Where the standard specified with @option{-std} represents a GNU
3362 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3363 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3364 extended dialect is based. Warnings from @option{-Wpedantic} are given
3365 where they are required by the base standard. (It does not make sense
3366 for such warnings to be given only for features not in the specified GNU
3367 C dialect, since by definition the GNU dialects of C include all
3368 features the compiler supports with the given option, and there would be
3369 nothing to warn about.)
3371 @item -pedantic-errors
3372 @opindex pedantic-errors
3373 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3374 requires a diagnostic, in some cases where there is undefined behavior
3375 at compile-time and in some other cases that do not prevent compilation
3376 of programs that are valid according to the standard. This is not
3377 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3378 by this option and not enabled by the latter and vice versa.
3383 This enables all the warnings about constructions that some users
3384 consider questionable, and that are easy to avoid (or modify to
3385 prevent the warning), even in conjunction with macros. This also
3386 enables some language-specific warnings described in @ref{C++ Dialect
3387 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3389 @option{-Wall} turns on the following warning flags:
3391 @gccoptlist{-Waddress @gol
3392 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3393 -Wc++11-compat -Wc++14-compat@gol
3394 -Wchar-subscripts @gol
3395 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3396 -Wimplicit-int @r{(C and Objective-C only)} @gol
3397 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3400 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3401 -Wmaybe-uninitialized @gol
3402 -Wmissing-braces @r{(only for C/ObjC)} @gol
3409 -Wsequence-point @gol
3410 -Wsign-compare @r{(only in C++)} @gol
3411 -Wstrict-aliasing @gol
3412 -Wstrict-overflow=1 @gol
3415 -Wuninitialized @gol
3416 -Wunknown-pragmas @gol
3417 -Wunused-function @gol
3420 -Wunused-variable @gol
3421 -Wvolatile-register-var @gol
3424 Note that some warning flags are not implied by @option{-Wall}. Some of
3425 them warn about constructions that users generally do not consider
3426 questionable, but which occasionally you might wish to check for;
3427 others warn about constructions that are necessary or hard to avoid in
3428 some cases, and there is no simple way to modify the code to suppress
3429 the warning. Some of them are enabled by @option{-Wextra} but many of
3430 them must be enabled individually.
3436 This enables some extra warning flags that are not enabled by
3437 @option{-Wall}. (This option used to be called @option{-W}. The older
3438 name is still supported, but the newer name is more descriptive.)
3440 @gccoptlist{-Wclobbered @gol
3442 -Wignored-qualifiers @gol
3443 -Wmissing-field-initializers @gol
3444 -Wmissing-parameter-type @r{(C only)} @gol
3445 -Wold-style-declaration @r{(C only)} @gol
3446 -Woverride-init @gol
3449 -Wuninitialized @gol
3450 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3451 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3454 The option @option{-Wextra} also prints warning messages for the
3460 A pointer is compared against integer zero with @code{<}, @code{<=},
3461 @code{>}, or @code{>=}.
3464 (C++ only) An enumerator and a non-enumerator both appear in a
3465 conditional expression.
3468 (C++ only) Ambiguous virtual bases.
3471 (C++ only) Subscripting an array that has been declared @code{register}.
3474 (C++ only) Taking the address of a variable that has been declared
3478 (C++ only) A base class is not initialized in a derived class's copy
3483 @item -Wchar-subscripts
3484 @opindex Wchar-subscripts
3485 @opindex Wno-char-subscripts
3486 Warn if an array subscript has type @code{char}. This is a common cause
3487 of error, as programmers often forget that this type is signed on some
3489 This warning is enabled by @option{-Wall}.
3493 @opindex Wno-comment
3494 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3495 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3496 This warning is enabled by @option{-Wall}.
3498 @item -Wno-coverage-mismatch
3499 @opindex Wno-coverage-mismatch
3500 Warn if feedback profiles do not match when using the
3501 @option{-fprofile-use} option.
3502 If a source file is changed between compiling with @option{-fprofile-gen} and
3503 with @option{-fprofile-use}, the files with the profile feedback can fail
3504 to match the source file and GCC cannot use the profile feedback
3505 information. By default, this warning is enabled and is treated as an
3506 error. @option{-Wno-coverage-mismatch} can be used to disable the
3507 warning or @option{-Wno-error=coverage-mismatch} can be used to
3508 disable the error. Disabling the error for this warning can result in
3509 poorly optimized code and is useful only in the
3510 case of very minor changes such as bug fixes to an existing code-base.
3511 Completely disabling the warning is not recommended.
3514 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3516 Suppress warning messages emitted by @code{#warning} directives.
3518 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3519 @opindex Wdouble-promotion
3520 @opindex Wno-double-promotion
3521 Give a warning when a value of type @code{float} is implicitly
3522 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3523 floating-point unit implement @code{float} in hardware, but emulate
3524 @code{double} in software. On such a machine, doing computations
3525 using @code{double} values is much more expensive because of the
3526 overhead required for software emulation.
3528 It is easy to accidentally do computations with @code{double} because
3529 floating-point literals are implicitly of type @code{double}. For
3533 float area(float radius)
3535 return 3.14159 * radius * radius;
3539 the compiler performs the entire computation with @code{double}
3540 because the floating-point literal is a @code{double}.
3543 @itemx -Wformat=@var{n}
3546 @opindex ffreestanding
3547 @opindex fno-builtin
3549 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3550 the arguments supplied have types appropriate to the format string
3551 specified, and that the conversions specified in the format string make
3552 sense. This includes standard functions, and others specified by format
3553 attributes (@pxref{Function Attributes}), in the @code{printf},
3554 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3555 not in the C standard) families (or other target-specific families).
3556 Which functions are checked without format attributes having been
3557 specified depends on the standard version selected, and such checks of
3558 functions without the attribute specified are disabled by
3559 @option{-ffreestanding} or @option{-fno-builtin}.
3561 The formats are checked against the format features supported by GNU
3562 libc version 2.2. These include all ISO C90 and C99 features, as well
3563 as features from the Single Unix Specification and some BSD and GNU
3564 extensions. Other library implementations may not support all these
3565 features; GCC does not support warning about features that go beyond a
3566 particular library's limitations. However, if @option{-Wpedantic} is used
3567 with @option{-Wformat}, warnings are given about format features not
3568 in the selected standard version (but not for @code{strfmon} formats,
3569 since those are not in any version of the C standard). @xref{C Dialect
3570 Options,,Options Controlling C Dialect}.
3577 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3578 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3579 @option{-Wformat} also checks for null format arguments for several
3580 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3581 aspects of this level of format checking can be disabled by the
3582 options: @option{-Wno-format-contains-nul},
3583 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3584 @option{-Wformat} is enabled by @option{-Wall}.
3586 @item -Wno-format-contains-nul
3587 @opindex Wno-format-contains-nul
3588 @opindex Wformat-contains-nul
3589 If @option{-Wformat} is specified, do not warn about format strings that
3592 @item -Wno-format-extra-args
3593 @opindex Wno-format-extra-args
3594 @opindex Wformat-extra-args
3595 If @option{-Wformat} is specified, do not warn about excess arguments to a
3596 @code{printf} or @code{scanf} format function. The C standard specifies
3597 that such arguments are ignored.
3599 Where the unused arguments lie between used arguments that are
3600 specified with @samp{$} operand number specifications, normally
3601 warnings are still given, since the implementation could not know what
3602 type to pass to @code{va_arg} to skip the unused arguments. However,
3603 in the case of @code{scanf} formats, this option suppresses the
3604 warning if the unused arguments are all pointers, since the Single
3605 Unix Specification says that such unused arguments are allowed.
3607 @item -Wno-format-zero-length
3608 @opindex Wno-format-zero-length
3609 @opindex Wformat-zero-length
3610 If @option{-Wformat} is specified, do not warn about zero-length formats.
3611 The C standard specifies that zero-length formats are allowed.
3616 Enable @option{-Wformat} plus additional format checks. Currently
3617 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3618 -Wformat-signedness -Wformat-y2k}.
3620 @item -Wformat-nonliteral
3621 @opindex Wformat-nonliteral
3622 @opindex Wno-format-nonliteral
3623 If @option{-Wformat} is specified, also warn if the format string is not a
3624 string literal and so cannot be checked, unless the format function
3625 takes its format arguments as a @code{va_list}.
3627 @item -Wformat-security
3628 @opindex Wformat-security
3629 @opindex Wno-format-security
3630 If @option{-Wformat} is specified, also warn about uses of format
3631 functions that represent possible security problems. At present, this
3632 warns about calls to @code{printf} and @code{scanf} functions where the
3633 format string is not a string literal and there are no format arguments,
3634 as in @code{printf (foo);}. This may be a security hole if the format
3635 string came from untrusted input and contains @samp{%n}. (This is
3636 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3637 in future warnings may be added to @option{-Wformat-security} that are not
3638 included in @option{-Wformat-nonliteral}.)
3640 @item -Wformat-signedness
3641 @opindex Wformat-signedness
3642 @opindex Wno-format-signedness
3643 If @option{-Wformat} is specified, also warn if the format string
3644 requires an unsigned argument and the argument is signed and vice versa.
3647 @opindex Wformat-y2k
3648 @opindex Wno-format-y2k
3649 If @option{-Wformat} is specified, also warn about @code{strftime}
3650 formats that may yield only a two-digit year.
3655 @opindex Wno-nonnull
3656 Warn about passing a null pointer for arguments marked as
3657 requiring a non-null value by the @code{nonnull} function attribute.
3659 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3660 can be disabled with the @option{-Wno-nonnull} option.
3662 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3664 @opindex Wno-init-self
3665 Warn about uninitialized variables that are initialized with themselves.
3666 Note this option can only be used with the @option{-Wuninitialized} option.
3668 For example, GCC warns about @code{i} being uninitialized in the
3669 following snippet only when @option{-Winit-self} has been specified:
3680 This warning is enabled by @option{-Wall} in C++.
3682 @item -Wimplicit-int @r{(C and Objective-C only)}
3683 @opindex Wimplicit-int
3684 @opindex Wno-implicit-int
3685 Warn when a declaration does not specify a type.
3686 This warning is enabled by @option{-Wall}.
3688 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3689 @opindex Wimplicit-function-declaration
3690 @opindex Wno-implicit-function-declaration
3691 Give a warning whenever a function is used before being declared. In
3692 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3693 enabled by default and it is made into an error by
3694 @option{-pedantic-errors}. This warning is also enabled by
3697 @item -Wimplicit @r{(C and Objective-C only)}
3699 @opindex Wno-implicit
3700 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3701 This warning is enabled by @option{-Wall}.
3703 @item -Wignored-qualifiers @r{(C and C++ only)}
3704 @opindex Wignored-qualifiers
3705 @opindex Wno-ignored-qualifiers
3706 Warn if the return type of a function has a type qualifier
3707 such as @code{const}. For ISO C such a type qualifier has no effect,
3708 since the value returned by a function is not an lvalue.
3709 For C++, the warning is only emitted for scalar types or @code{void}.
3710 ISO C prohibits qualified @code{void} return types on function
3711 definitions, so such return types always receive a warning
3712 even without this option.
3714 This warning is also enabled by @option{-Wextra}.
3719 Warn if the type of @code{main} is suspicious. @code{main} should be
3720 a function with external linkage, returning int, taking either zero
3721 arguments, two, or three arguments of appropriate types. This warning
3722 is enabled by default in C++ and is enabled by either @option{-Wall}
3723 or @option{-Wpedantic}.
3725 @item -Wmissing-braces
3726 @opindex Wmissing-braces
3727 @opindex Wno-missing-braces
3728 Warn if an aggregate or union initializer is not fully bracketed. In
3729 the following example, the initializer for @code{a} is not fully
3730 bracketed, but that for @code{b} is fully bracketed. This warning is
3731 enabled by @option{-Wall} in C.
3734 int a[2][2] = @{ 0, 1, 2, 3 @};
3735 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3738 This warning is enabled by @option{-Wall}.
3740 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3741 @opindex Wmissing-include-dirs
3742 @opindex Wno-missing-include-dirs
3743 Warn if a user-supplied include directory does not exist.
3746 @opindex Wparentheses
3747 @opindex Wno-parentheses
3748 Warn if parentheses are omitted in certain contexts, such
3749 as when there is an assignment in a context where a truth value
3750 is expected, or when operators are nested whose precedence people
3751 often get confused about.
3753 Also warn if a comparison like @code{x<=y<=z} appears; this is
3754 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3755 interpretation from that of ordinary mathematical notation.
3757 Also warn about constructions where there may be confusion to which
3758 @code{if} statement an @code{else} branch belongs. Here is an example of
3773 In C/C++, every @code{else} branch belongs to the innermost possible
3774 @code{if} statement, which in this example is @code{if (b)}. This is
3775 often not what the programmer expected, as illustrated in the above
3776 example by indentation the programmer chose. When there is the
3777 potential for this confusion, GCC issues a warning when this flag
3778 is specified. To eliminate the warning, add explicit braces around
3779 the innermost @code{if} statement so there is no way the @code{else}
3780 can belong to the enclosing @code{if}. The resulting code
3797 Also warn for dangerous uses of the GNU extension to
3798 @code{?:} with omitted middle operand. When the condition
3799 in the @code{?}: operator is a boolean expression, the omitted value is
3800 always 1. Often programmers expect it to be a value computed
3801 inside the conditional expression instead.
3803 This warning is enabled by @option{-Wall}.
3805 @item -Wsequence-point
3806 @opindex Wsequence-point
3807 @opindex Wno-sequence-point
3808 Warn about code that may have undefined semantics because of violations
3809 of sequence point rules in the C and C++ standards.
3811 The C and C++ standards define the order in which expressions in a C/C++
3812 program are evaluated in terms of @dfn{sequence points}, which represent
3813 a partial ordering between the execution of parts of the program: those
3814 executed before the sequence point, and those executed after it. These
3815 occur after the evaluation of a full expression (one which is not part
3816 of a larger expression), after the evaluation of the first operand of a
3817 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3818 function is called (but after the evaluation of its arguments and the
3819 expression denoting the called function), and in certain other places.
3820 Other than as expressed by the sequence point rules, the order of
3821 evaluation of subexpressions of an expression is not specified. All
3822 these rules describe only a partial order rather than a total order,
3823 since, for example, if two functions are called within one expression
3824 with no sequence point between them, the order in which the functions
3825 are called is not specified. However, the standards committee have
3826 ruled that function calls do not overlap.
3828 It is not specified when between sequence points modifications to the
3829 values of objects take effect. Programs whose behavior depends on this
3830 have undefined behavior; the C and C++ standards specify that ``Between
3831 the previous and next sequence point an object shall have its stored
3832 value modified at most once by the evaluation of an expression.
3833 Furthermore, the prior value shall be read only to determine the value
3834 to be stored.''. If a program breaks these rules, the results on any
3835 particular implementation are entirely unpredictable.
3837 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3838 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3839 diagnosed by this option, and it may give an occasional false positive
3840 result, but in general it has been found fairly effective at detecting
3841 this sort of problem in programs.
3843 The standard is worded confusingly, therefore there is some debate
3844 over the precise meaning of the sequence point rules in subtle cases.
3845 Links to discussions of the problem, including proposed formal
3846 definitions, may be found on the GCC readings page, at
3847 @uref{http://gcc.gnu.org/@/readings.html}.
3849 This warning is enabled by @option{-Wall} for C and C++.
3851 @item -Wno-return-local-addr
3852 @opindex Wno-return-local-addr
3853 @opindex Wreturn-local-addr
3854 Do not warn about returning a pointer (or in C++, a reference) to a
3855 variable that goes out of scope after the function returns.
3858 @opindex Wreturn-type
3859 @opindex Wno-return-type
3860 Warn whenever a function is defined with a return type that defaults
3861 to @code{int}. Also warn about any @code{return} statement with no
3862 return value in a function whose return type is not @code{void}
3863 (falling off the end of the function body is considered returning
3864 without a value), and about a @code{return} statement with an
3865 expression in a function whose return type is @code{void}.
3867 For C++, a function without return type always produces a diagnostic
3868 message, even when @option{-Wno-return-type} is specified. The only
3869 exceptions are @code{main} and functions defined in system headers.
3871 This warning is enabled by @option{-Wall}.
3873 @item -Wshift-count-negative
3874 @opindex Wshift-count-negative
3875 @opindex Wno-shift-count-negative
3876 Warn if shift count is negative. This warning is enabled by default.
3878 @item -Wshift-count-overflow
3879 @opindex Wshift-count-overflow
3880 @opindex Wno-shift-count-overflow
3881 Warn if shift count >= width of type. This warning is enabled by default.
3886 Warn whenever a @code{switch} statement has an index of enumerated type
3887 and lacks a @code{case} for one or more of the named codes of that
3888 enumeration. (The presence of a @code{default} label prevents this
3889 warning.) @code{case} labels outside the enumeration range also
3890 provoke warnings when this option is used (even if there is a
3891 @code{default} label).
3892 This warning is enabled by @option{-Wall}.
3894 @item -Wswitch-default
3895 @opindex Wswitch-default
3896 @opindex Wno-switch-default
3897 Warn whenever a @code{switch} statement does not have a @code{default}
3901 @opindex Wswitch-enum
3902 @opindex Wno-switch-enum
3903 Warn whenever a @code{switch} statement has an index of enumerated type
3904 and lacks a @code{case} for one or more of the named codes of that
3905 enumeration. @code{case} labels outside the enumeration range also
3906 provoke warnings when this option is used. The only difference
3907 between @option{-Wswitch} and this option is that this option gives a
3908 warning about an omitted enumeration code even if there is a
3909 @code{default} label.
3912 @opindex Wswitch-bool
3913 @opindex Wno-switch-bool
3914 Warn whenever a @code{switch} statement has an index of boolean type.
3915 It is possible to suppress this warning by casting the controlling
3916 expression to a type other than @code{bool}. For example:
3919 switch ((int) (a == 4))
3925 This warning is enabled by default for C and C++ programs.
3927 @item -Wsync-nand @r{(C and C++ only)}
3929 @opindex Wno-sync-nand
3930 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3931 built-in functions are used. These functions changed semantics in GCC 4.4.
3935 @opindex Wno-trigraphs
3936 Warn if any trigraphs are encountered that might change the meaning of
3937 the program (trigraphs within comments are not warned about).
3938 This warning is enabled by @option{-Wall}.
3940 @item -Wunused-but-set-parameter
3941 @opindex Wunused-but-set-parameter
3942 @opindex Wno-unused-but-set-parameter
3943 Warn whenever a function parameter is assigned to, but otherwise unused
3944 (aside from its declaration).
3946 To suppress this warning use the @code{unused} attribute
3947 (@pxref{Variable Attributes}).
3949 This warning is also enabled by @option{-Wunused} together with
3952 @item -Wunused-but-set-variable
3953 @opindex Wunused-but-set-variable
3954 @opindex Wno-unused-but-set-variable
3955 Warn whenever a local variable is assigned to, but otherwise unused
3956 (aside from its declaration).
3957 This warning is enabled by @option{-Wall}.
3959 To suppress this warning use the @code{unused} attribute
3960 (@pxref{Variable Attributes}).
3962 This warning is also enabled by @option{-Wunused}, which is enabled
3965 @item -Wunused-function
3966 @opindex Wunused-function
3967 @opindex Wno-unused-function
3968 Warn whenever a static function is declared but not defined or a
3969 non-inline static function is unused.
3970 This warning is enabled by @option{-Wall}.
3972 @item -Wunused-label
3973 @opindex Wunused-label
3974 @opindex Wno-unused-label
3975 Warn whenever a label is declared but not used.
3976 This warning is enabled by @option{-Wall}.
3978 To suppress this warning use the @code{unused} attribute
3979 (@pxref{Variable Attributes}).
3981 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3982 @opindex Wunused-local-typedefs
3983 Warn when a typedef locally defined in a function is not used.
3984 This warning is enabled by @option{-Wall}.
3986 @item -Wunused-parameter
3987 @opindex Wunused-parameter
3988 @opindex Wno-unused-parameter
3989 Warn whenever a function parameter is unused aside from its declaration.
3991 To suppress this warning use the @code{unused} attribute
3992 (@pxref{Variable Attributes}).
3994 @item -Wno-unused-result
3995 @opindex Wunused-result
3996 @opindex Wno-unused-result
3997 Do not warn if a caller of a function marked with attribute
3998 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3999 its return value. The default is @option{-Wunused-result}.
4001 @item -Wunused-variable
4002 @opindex Wunused-variable
4003 @opindex Wno-unused-variable
4004 Warn whenever a local variable or non-constant static variable is unused
4005 aside from its declaration.
4006 This warning is enabled by @option{-Wall}.
4008 To suppress this warning use the @code{unused} attribute
4009 (@pxref{Variable Attributes}).
4011 @item -Wunused-value
4012 @opindex Wunused-value
4013 @opindex Wno-unused-value
4014 Warn whenever a statement computes a result that is explicitly not
4015 used. To suppress this warning cast the unused expression to
4016 @code{void}. This includes an expression-statement or the left-hand
4017 side of a comma expression that contains no side effects. For example,
4018 an expression such as @code{x[i,j]} causes a warning, while
4019 @code{x[(void)i,j]} does not.
4021 This warning is enabled by @option{-Wall}.
4026 All the above @option{-Wunused} options combined.
4028 In order to get a warning about an unused function parameter, you must
4029 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4030 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4032 @item -Wuninitialized
4033 @opindex Wuninitialized
4034 @opindex Wno-uninitialized
4035 Warn if an automatic variable is used without first being initialized
4036 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4037 warn if a non-static reference or non-static @code{const} member
4038 appears in a class without constructors.
4040 If you want to warn about code that uses the uninitialized value of the
4041 variable in its own initializer, use the @option{-Winit-self} option.
4043 These warnings occur for individual uninitialized or clobbered
4044 elements of structure, union or array variables as well as for
4045 variables that are uninitialized or clobbered as a whole. They do
4046 not occur for variables or elements declared @code{volatile}. Because
4047 these warnings depend on optimization, the exact variables or elements
4048 for which there are warnings depends on the precise optimization
4049 options and version of GCC used.
4051 Note that there may be no warning about a variable that is used only
4052 to compute a value that itself is never used, because such
4053 computations may be deleted by data flow analysis before the warnings
4056 @item -Wmaybe-uninitialized
4057 @opindex Wmaybe-uninitialized
4058 @opindex Wno-maybe-uninitialized
4059 For an automatic variable, if there exists a path from the function
4060 entry to a use of the variable that is initialized, but there exist
4061 some other paths for which the variable is not initialized, the compiler
4062 emits a warning if it cannot prove the uninitialized paths are not
4063 executed at run time. These warnings are made optional because GCC is
4064 not smart enough to see all the reasons why the code might be correct
4065 in spite of appearing to have an error. Here is one example of how
4086 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4087 always initialized, but GCC doesn't know this. To suppress the
4088 warning, you need to provide a default case with assert(0) or
4091 @cindex @code{longjmp} warnings
4092 This option also warns when a non-volatile automatic variable might be
4093 changed by a call to @code{longjmp}. These warnings as well are possible
4094 only in optimizing compilation.
4096 The compiler sees only the calls to @code{setjmp}. It cannot know
4097 where @code{longjmp} will be called; in fact, a signal handler could
4098 call it at any point in the code. As a result, you may get a warning
4099 even when there is in fact no problem because @code{longjmp} cannot
4100 in fact be called at the place that would cause a problem.
4102 Some spurious warnings can be avoided if you declare all the functions
4103 you use that never return as @code{noreturn}. @xref{Function
4106 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4108 @item -Wunknown-pragmas
4109 @opindex Wunknown-pragmas
4110 @opindex Wno-unknown-pragmas
4111 @cindex warning for unknown pragmas
4112 @cindex unknown pragmas, warning
4113 @cindex pragmas, warning of unknown
4114 Warn when a @code{#pragma} directive is encountered that is not understood by
4115 GCC@. If this command-line option is used, warnings are even issued
4116 for unknown pragmas in system header files. This is not the case if
4117 the warnings are only enabled by the @option{-Wall} command-line option.
4120 @opindex Wno-pragmas
4122 Do not warn about misuses of pragmas, such as incorrect parameters,
4123 invalid syntax, or conflicts between pragmas. See also
4124 @option{-Wunknown-pragmas}.
4126 @item -Wstrict-aliasing
4127 @opindex Wstrict-aliasing
4128 @opindex Wno-strict-aliasing
4129 This option is only active when @option{-fstrict-aliasing} is active.
4130 It warns about code that might break the strict aliasing rules that the
4131 compiler is using for optimization. The warning does not catch all
4132 cases, but does attempt to catch the more common pitfalls. It is
4133 included in @option{-Wall}.
4134 It is equivalent to @option{-Wstrict-aliasing=3}
4136 @item -Wstrict-aliasing=n
4137 @opindex Wstrict-aliasing=n
4138 This option is only active when @option{-fstrict-aliasing} is active.
4139 It warns about code that might break the strict aliasing rules that the
4140 compiler is using for optimization.
4141 Higher levels correspond to higher accuracy (fewer false positives).
4142 Higher levels also correspond to more effort, similar to the way @option{-O}
4144 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4146 Level 1: Most aggressive, quick, least accurate.
4147 Possibly useful when higher levels
4148 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4149 false negatives. However, it has many false positives.
4150 Warns for all pointer conversions between possibly incompatible types,
4151 even if never dereferenced. Runs in the front end only.
4153 Level 2: Aggressive, quick, not too precise.
4154 May still have many false positives (not as many as level 1 though),
4155 and few false negatives (but possibly more than level 1).
4156 Unlike level 1, it only warns when an address is taken. Warns about
4157 incomplete types. Runs in the front end only.
4159 Level 3 (default for @option{-Wstrict-aliasing}):
4160 Should have very few false positives and few false
4161 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4162 Takes care of the common pun+dereference pattern in the front end:
4163 @code{*(int*)&some_float}.
4164 If optimization is enabled, it also runs in the back end, where it deals
4165 with multiple statement cases using flow-sensitive points-to information.
4166 Only warns when the converted pointer is dereferenced.
4167 Does not warn about incomplete types.
4169 @item -Wstrict-overflow
4170 @itemx -Wstrict-overflow=@var{n}
4171 @opindex Wstrict-overflow
4172 @opindex Wno-strict-overflow
4173 This option is only active when @option{-fstrict-overflow} is active.
4174 It warns about cases where the compiler optimizes based on the
4175 assumption that signed overflow does not occur. Note that it does not
4176 warn about all cases where the code might overflow: it only warns
4177 about cases where the compiler implements some optimization. Thus
4178 this warning depends on the optimization level.
4180 An optimization that assumes that signed overflow does not occur is
4181 perfectly safe if the values of the variables involved are such that
4182 overflow never does, in fact, occur. Therefore this warning can
4183 easily give a false positive: a warning about code that is not
4184 actually a problem. To help focus on important issues, several
4185 warning levels are defined. No warnings are issued for the use of
4186 undefined signed overflow when estimating how many iterations a loop
4187 requires, in particular when determining whether a loop will be
4191 @item -Wstrict-overflow=1
4192 Warn about cases that are both questionable and easy to avoid. For
4193 example, with @option{-fstrict-overflow}, the compiler simplifies
4194 @code{x + 1 > x} to @code{1}. This level of
4195 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4196 are not, and must be explicitly requested.
4198 @item -Wstrict-overflow=2
4199 Also warn about other cases where a comparison is simplified to a
4200 constant. For example: @code{abs (x) >= 0}. This can only be
4201 simplified when @option{-fstrict-overflow} is in effect, because
4202 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4203 zero. @option{-Wstrict-overflow} (with no level) is the same as
4204 @option{-Wstrict-overflow=2}.
4206 @item -Wstrict-overflow=3
4207 Also warn about other cases where a comparison is simplified. For
4208 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4210 @item -Wstrict-overflow=4
4211 Also warn about other simplifications not covered by the above cases.
4212 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4214 @item -Wstrict-overflow=5
4215 Also warn about cases where the compiler reduces the magnitude of a
4216 constant involved in a comparison. For example: @code{x + 2 > y} is
4217 simplified to @code{x + 1 >= y}. This is reported only at the
4218 highest warning level because this simplification applies to many
4219 comparisons, so this warning level gives a very large number of
4223 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4224 @opindex Wsuggest-attribute=
4225 @opindex Wno-suggest-attribute=
4226 Warn for cases where adding an attribute may be beneficial. The
4227 attributes currently supported are listed below.
4230 @item -Wsuggest-attribute=pure
4231 @itemx -Wsuggest-attribute=const
4232 @itemx -Wsuggest-attribute=noreturn
4233 @opindex Wsuggest-attribute=pure
4234 @opindex Wno-suggest-attribute=pure
4235 @opindex Wsuggest-attribute=const
4236 @opindex Wno-suggest-attribute=const
4237 @opindex Wsuggest-attribute=noreturn
4238 @opindex Wno-suggest-attribute=noreturn
4240 Warn about functions that might be candidates for attributes
4241 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4242 functions visible in other compilation units or (in the case of @code{pure} and
4243 @code{const}) if it cannot prove that the function returns normally. A function
4244 returns normally if it doesn't contain an infinite loop or return abnormally
4245 by throwing, calling @code{abort} or trapping. This analysis requires option
4246 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4247 higher. Higher optimization levels improve the accuracy of the analysis.
4249 @item -Wsuggest-attribute=format
4250 @itemx -Wmissing-format-attribute
4251 @opindex Wsuggest-attribute=format
4252 @opindex Wmissing-format-attribute
4253 @opindex Wno-suggest-attribute=format
4254 @opindex Wno-missing-format-attribute
4258 Warn about function pointers that might be candidates for @code{format}
4259 attributes. Note these are only possible candidates, not absolute ones.
4260 GCC guesses that function pointers with @code{format} attributes that
4261 are used in assignment, initialization, parameter passing or return
4262 statements should have a corresponding @code{format} attribute in the
4263 resulting type. I.e.@: the left-hand side of the assignment or
4264 initialization, the type of the parameter variable, or the return type
4265 of the containing function respectively should also have a @code{format}
4266 attribute to avoid the warning.
4268 GCC also warns about function definitions that might be
4269 candidates for @code{format} attributes. Again, these are only
4270 possible candidates. GCC guesses that @code{format} attributes
4271 might be appropriate for any function that calls a function like
4272 @code{vprintf} or @code{vscanf}, but this might not always be the
4273 case, and some functions for which @code{format} attributes are
4274 appropriate may not be detected.
4277 @item -Wsuggest-final-types
4278 @opindex Wno-suggest-final-types
4279 @opindex Wsuggest-final-types
4280 Warn about types with virtual methods where code quality would be improved
4281 if the type were declared with the C++11 @code{final} specifier,
4283 declared in an anonymous namespace. This allows GCC to more aggressively
4284 devirtualize the polymorphic calls. This warning is more effective with link
4285 time optimization, where the information about the class hierarchy graph is
4288 @item -Wsuggest-final-methods
4289 @opindex Wno-suggest-final-methods
4290 @opindex Wsuggest-final-methods
4291 Warn about virtual methods where code quality would be improved if the method
4292 were declared with the C++11 @code{final} specifier,
4293 or, if possible, its type were
4294 declared in an anonymous namespace or with the @code{final} specifier.
4296 more effective with link time optimization, where the information about the
4297 class hierarchy graph is more complete. It is recommended to first consider
4298 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4301 @item -Wsuggest-override
4302 Warn about overriding virtual functions that are not marked with the override
4305 @item -Warray-bounds
4306 @itemx -Warray-bounds=@var{n}
4307 @opindex Wno-array-bounds
4308 @opindex Warray-bounds
4309 This option is only active when @option{-ftree-vrp} is active
4310 (default for @option{-O2} and above). It warns about subscripts to arrays
4311 that are always out of bounds. This warning is enabled by @option{-Wall}.
4314 @item -Warray-bounds=1
4315 This is the warning level of @option{-Warray-bounds} and is enabled
4316 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4318 @item -Warray-bounds=2
4319 This warning level also warns about out of bounds access for
4320 arrays at the end of a struct and for arrays accessed through
4321 pointers. This warning level may give a larger number of
4322 false positives and is deactivated by default.
4326 @item -Wbool-compare
4327 @opindex Wno-bool-compare
4328 @opindex Wbool-compare
4329 Warn about boolean expression compared with an integer value different from
4330 @code{true}/@code{false}. For instance, the following comparison is
4335 if ((n > 1) == 2) @{ @dots{} @}
4337 This warning is enabled by @option{-Wall}.
4339 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4340 @opindex Wno-discarded-qualifiers
4341 @opindex Wdiscarded-qualifiers
4342 Do not warn if type qualifiers on pointers are being discarded.
4343 Typically, the compiler warns if a @code{const char *} variable is
4344 passed to a function that takes a @code{char *} parameter. This option
4345 can be used to suppress such a warning.
4347 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4348 @opindex Wno-discarded-array-qualifiers
4349 @opindex Wdiscarded-array-qualifiers
4350 Do not warn if type qualifiers on arrays which are pointer targets
4351 are being discarded. Typically, the compiler warns if a
4352 @code{const int (*)[]} variable is passed to a function that
4353 takes a @code{int (*)[]} parameter. This option can be used to
4354 suppress such a warning.
4356 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4357 @opindex Wno-incompatible-pointer-types
4358 @opindex Wincompatible-pointer-types
4359 Do not warn when there is a conversion between pointers that have incompatible
4360 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4361 which warns for pointer argument passing or assignment with different
4364 @item -Wno-int-conversion @r{(C and Objective-C only)}
4365 @opindex Wno-int-conversion
4366 @opindex Wint-conversion
4367 Do not warn about incompatible integer to pointer and pointer to integer
4368 conversions. This warning is about implicit conversions; for explicit
4369 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4370 @option{-Wno-pointer-to-int-cast} may be used.
4372 @item -Wno-div-by-zero
4373 @opindex Wno-div-by-zero
4374 @opindex Wdiv-by-zero
4375 Do not warn about compile-time integer division by zero. Floating-point
4376 division by zero is not warned about, as it can be a legitimate way of
4377 obtaining infinities and NaNs.
4379 @item -Wsystem-headers
4380 @opindex Wsystem-headers
4381 @opindex Wno-system-headers
4382 @cindex warnings from system headers
4383 @cindex system headers, warnings from
4384 Print warning messages for constructs found in system header files.
4385 Warnings from system headers are normally suppressed, on the assumption
4386 that they usually do not indicate real problems and would only make the
4387 compiler output harder to read. Using this command-line option tells
4388 GCC to emit warnings from system headers as if they occurred in user
4389 code. However, note that using @option{-Wall} in conjunction with this
4390 option does @emph{not} warn about unknown pragmas in system
4391 headers---for that, @option{-Wunknown-pragmas} must also be used.
4394 @opindex Wtrampolines
4395 @opindex Wno-trampolines
4396 Warn about trampolines generated for pointers to nested functions.
4397 A trampoline is a small piece of data or code that is created at run
4398 time on the stack when the address of a nested function is taken, and is
4399 used to call the nested function indirectly. For some targets, it is
4400 made up of data only and thus requires no special treatment. But, for
4401 most targets, it is made up of code and thus requires the stack to be
4402 made executable in order for the program to work properly.
4405 @opindex Wfloat-equal
4406 @opindex Wno-float-equal
4407 Warn if floating-point values are used in equality comparisons.
4409 The idea behind this is that sometimes it is convenient (for the
4410 programmer) to consider floating-point values as approximations to
4411 infinitely precise real numbers. If you are doing this, then you need
4412 to compute (by analyzing the code, or in some other way) the maximum or
4413 likely maximum error that the computation introduces, and allow for it
4414 when performing comparisons (and when producing output, but that's a
4415 different problem). In particular, instead of testing for equality, you
4416 should check to see whether the two values have ranges that overlap; and
4417 this is done with the relational operators, so equality comparisons are
4420 @item -Wtraditional @r{(C and Objective-C only)}
4421 @opindex Wtraditional
4422 @opindex Wno-traditional
4423 Warn about certain constructs that behave differently in traditional and
4424 ISO C@. Also warn about ISO C constructs that have no traditional C
4425 equivalent, and/or problematic constructs that should be avoided.
4429 Macro parameters that appear within string literals in the macro body.
4430 In traditional C macro replacement takes place within string literals,
4431 but in ISO C it does not.
4434 In traditional C, some preprocessor directives did not exist.
4435 Traditional preprocessors only considered a line to be a directive
4436 if the @samp{#} appeared in column 1 on the line. Therefore
4437 @option{-Wtraditional} warns about directives that traditional C
4438 understands but ignores because the @samp{#} does not appear as the
4439 first character on the line. It also suggests you hide directives like
4440 @code{#pragma} not understood by traditional C by indenting them. Some
4441 traditional implementations do not recognize @code{#elif}, so this option
4442 suggests avoiding it altogether.
4445 A function-like macro that appears without arguments.
4448 The unary plus operator.
4451 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4452 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4453 constants.) Note, these suffixes appear in macros defined in the system
4454 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4455 Use of these macros in user code might normally lead to spurious
4456 warnings, however GCC's integrated preprocessor has enough context to
4457 avoid warning in these cases.
4460 A function declared external in one block and then used after the end of
4464 A @code{switch} statement has an operand of type @code{long}.
4467 A non-@code{static} function declaration follows a @code{static} one.
4468 This construct is not accepted by some traditional C compilers.
4471 The ISO type of an integer constant has a different width or
4472 signedness from its traditional type. This warning is only issued if
4473 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4474 typically represent bit patterns, are not warned about.
4477 Usage of ISO string concatenation is detected.
4480 Initialization of automatic aggregates.
4483 Identifier conflicts with labels. Traditional C lacks a separate
4484 namespace for labels.
4487 Initialization of unions. If the initializer is zero, the warning is
4488 omitted. This is done under the assumption that the zero initializer in
4489 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4490 initializer warnings and relies on default initialization to zero in the
4494 Conversions by prototypes between fixed/floating-point values and vice
4495 versa. The absence of these prototypes when compiling with traditional
4496 C causes serious problems. This is a subset of the possible
4497 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4500 Use of ISO C style function definitions. This warning intentionally is
4501 @emph{not} issued for prototype declarations or variadic functions
4502 because these ISO C features appear in your code when using
4503 libiberty's traditional C compatibility macros, @code{PARAMS} and
4504 @code{VPARAMS}. This warning is also bypassed for nested functions
4505 because that feature is already a GCC extension and thus not relevant to
4506 traditional C compatibility.
4509 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4510 @opindex Wtraditional-conversion
4511 @opindex Wno-traditional-conversion
4512 Warn if a prototype causes a type conversion that is different from what
4513 would happen to the same argument in the absence of a prototype. This
4514 includes conversions of fixed point to floating and vice versa, and
4515 conversions changing the width or signedness of a fixed-point argument
4516 except when the same as the default promotion.
4518 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4519 @opindex Wdeclaration-after-statement
4520 @opindex Wno-declaration-after-statement
4521 Warn when a declaration is found after a statement in a block. This
4522 construct, known from C++, was introduced with ISO C99 and is by default
4523 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4524 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4529 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4531 @item -Wno-endif-labels
4532 @opindex Wno-endif-labels
4533 @opindex Wendif-labels
4534 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4539 Warn whenever a local variable or type declaration shadows another
4540 variable, parameter, type, class member (in C++), or instance variable
4541 (in Objective-C) or whenever a built-in function is shadowed. Note
4542 that in C++, the compiler warns if a local variable shadows an
4543 explicit typedef, but not if it shadows a struct/class/enum.
4545 @item -Wno-shadow-ivar @r{(Objective-C only)}
4546 @opindex Wno-shadow-ivar
4547 @opindex Wshadow-ivar
4548 Do not warn whenever a local variable shadows an instance variable in an
4551 @item -Wlarger-than=@var{len}
4552 @opindex Wlarger-than=@var{len}
4553 @opindex Wlarger-than-@var{len}
4554 Warn whenever an object of larger than @var{len} bytes is defined.
4556 @item -Wframe-larger-than=@var{len}
4557 @opindex Wframe-larger-than
4558 Warn if the size of a function frame is larger than @var{len} bytes.
4559 The computation done to determine the stack frame size is approximate
4560 and not conservative.
4561 The actual requirements may be somewhat greater than @var{len}
4562 even if you do not get a warning. In addition, any space allocated
4563 via @code{alloca}, variable-length arrays, or related constructs
4564 is not included by the compiler when determining
4565 whether or not to issue a warning.
4567 @item -Wno-free-nonheap-object
4568 @opindex Wno-free-nonheap-object
4569 @opindex Wfree-nonheap-object
4570 Do not warn when attempting to free an object that was not allocated
4573 @item -Wstack-usage=@var{len}
4574 @opindex Wstack-usage
4575 Warn if the stack usage of a function might be larger than @var{len} bytes.
4576 The computation done to determine the stack usage is conservative.
4577 Any space allocated via @code{alloca}, variable-length arrays, or related
4578 constructs is included by the compiler when determining whether or not to
4581 The message is in keeping with the output of @option{-fstack-usage}.
4585 If the stack usage is fully static but exceeds the specified amount, it's:
4588 warning: stack usage is 1120 bytes
4591 If the stack usage is (partly) dynamic but bounded, it's:
4594 warning: stack usage might be 1648 bytes
4597 If the stack usage is (partly) dynamic and not bounded, it's:
4600 warning: stack usage might be unbounded
4604 @item -Wunsafe-loop-optimizations
4605 @opindex Wunsafe-loop-optimizations
4606 @opindex Wno-unsafe-loop-optimizations
4607 Warn if the loop cannot be optimized because the compiler cannot
4608 assume anything on the bounds of the loop indices. With
4609 @option{-funsafe-loop-optimizations} warn if the compiler makes
4612 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4613 @opindex Wno-pedantic-ms-format
4614 @opindex Wpedantic-ms-format
4615 When used in combination with @option{-Wformat}
4616 and @option{-pedantic} without GNU extensions, this option
4617 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4618 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4619 which depend on the MS runtime.
4621 @item -Wpointer-arith
4622 @opindex Wpointer-arith
4623 @opindex Wno-pointer-arith
4624 Warn about anything that depends on the ``size of'' a function type or
4625 of @code{void}. GNU C assigns these types a size of 1, for
4626 convenience in calculations with @code{void *} pointers and pointers
4627 to functions. In C++, warn also when an arithmetic operation involves
4628 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4631 @opindex Wtype-limits
4632 @opindex Wno-type-limits
4633 Warn if a comparison is always true or always false due to the limited
4634 range of the data type, but do not warn for constant expressions. For
4635 example, warn if an unsigned variable is compared against zero with
4636 @code{<} or @code{>=}. This warning is also enabled by
4639 @item -Wbad-function-cast @r{(C and Objective-C only)}
4640 @opindex Wbad-function-cast
4641 @opindex Wno-bad-function-cast
4642 Warn when a function call is cast to a non-matching type.
4643 For example, warn if a call to a function returning an integer type
4644 is cast to a pointer type.
4646 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4647 @opindex Wc90-c99-compat
4648 @opindex Wno-c90-c99-compat
4649 Warn about features not present in ISO C90, but present in ISO C99.
4650 For instance, warn about use of variable length arrays, @code{long long}
4651 type, @code{bool} type, compound literals, designated initializers, and so
4652 on. This option is independent of the standards mode. Warnings are disabled
4653 in the expression that follows @code{__extension__}.
4655 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4656 @opindex Wc99-c11-compat
4657 @opindex Wno-c99-c11-compat
4658 Warn about features not present in ISO C99, but present in ISO C11.
4659 For instance, warn about use of anonymous structures and unions,
4660 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4661 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4662 and so on. This option is independent of the standards mode. Warnings are
4663 disabled in the expression that follows @code{__extension__}.
4665 @item -Wc++-compat @r{(C and Objective-C only)}
4666 @opindex Wc++-compat
4667 Warn about ISO C constructs that are outside of the common subset of
4668 ISO C and ISO C++, e.g.@: request for implicit conversion from
4669 @code{void *} to a pointer to non-@code{void} type.
4671 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4672 @opindex Wc++11-compat
4673 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4674 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4675 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4676 enabled by @option{-Wall}.
4678 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4679 @opindex Wc++14-compat
4680 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4681 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4685 @opindex Wno-cast-qual
4686 Warn whenever a pointer is cast so as to remove a type qualifier from
4687 the target type. For example, warn if a @code{const char *} is cast
4688 to an ordinary @code{char *}.
4690 Also warn when making a cast that introduces a type qualifier in an
4691 unsafe way. For example, casting @code{char **} to @code{const char **}
4692 is unsafe, as in this example:
4695 /* p is char ** value. */
4696 const char **q = (const char **) p;
4697 /* Assignment of readonly string to const char * is OK. */
4699 /* Now char** pointer points to read-only memory. */
4704 @opindex Wcast-align
4705 @opindex Wno-cast-align
4706 Warn whenever a pointer is cast such that the required alignment of the
4707 target is increased. For example, warn if a @code{char *} is cast to
4708 an @code{int *} on machines where integers can only be accessed at
4709 two- or four-byte boundaries.
4711 @item -Wwrite-strings
4712 @opindex Wwrite-strings
4713 @opindex Wno-write-strings
4714 When compiling C, give string constants the type @code{const
4715 char[@var{length}]} so that copying the address of one into a
4716 non-@code{const} @code{char *} pointer produces a warning. These
4717 warnings help you find at compile time code that can try to write
4718 into a string constant, but only if you have been very careful about
4719 using @code{const} in declarations and prototypes. Otherwise, it is
4720 just a nuisance. This is why we did not make @option{-Wall} request
4723 When compiling C++, warn about the deprecated conversion from string
4724 literals to @code{char *}. This warning is enabled by default for C++
4729 @opindex Wno-clobbered
4730 Warn for variables that might be changed by @code{longjmp} or
4731 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4733 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4734 @opindex Wconditionally-supported
4735 @opindex Wno-conditionally-supported
4736 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4739 @opindex Wconversion
4740 @opindex Wno-conversion
4741 Warn for implicit conversions that may alter a value. This includes
4742 conversions between real and integer, like @code{abs (x)} when
4743 @code{x} is @code{double}; conversions between signed and unsigned,
4744 like @code{unsigned ui = -1}; and conversions to smaller types, like
4745 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4746 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4747 changed by the conversion like in @code{abs (2.0)}. Warnings about
4748 conversions between signed and unsigned integers can be disabled by
4749 using @option{-Wno-sign-conversion}.
4751 For C++, also warn for confusing overload resolution for user-defined
4752 conversions; and conversions that never use a type conversion
4753 operator: conversions to @code{void}, the same type, a base class or a
4754 reference to them. Warnings about conversions between signed and
4755 unsigned integers are disabled by default in C++ unless
4756 @option{-Wsign-conversion} is explicitly enabled.
4758 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4759 @opindex Wconversion-null
4760 @opindex Wno-conversion-null
4761 Do not warn for conversions between @code{NULL} and non-pointer
4762 types. @option{-Wconversion-null} is enabled by default.
4764 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4765 @opindex Wzero-as-null-pointer-constant
4766 @opindex Wno-zero-as-null-pointer-constant
4767 Warn when a literal '0' is used as null pointer constant. This can
4768 be useful to facilitate the conversion to @code{nullptr} in C++11.
4772 @opindex Wno-date-time
4773 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4774 are encountered as they might prevent bit-wise-identical reproducible
4777 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4778 @opindex Wdelete-incomplete
4779 @opindex Wno-delete-incomplete
4780 Warn when deleting a pointer to incomplete type, which may cause
4781 undefined behavior at runtime. This warning is enabled by default.
4783 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4784 @opindex Wuseless-cast
4785 @opindex Wno-useless-cast
4786 Warn when an expression is casted to its own type.
4789 @opindex Wempty-body
4790 @opindex Wno-empty-body
4791 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
4792 while} statement. This warning is also enabled by @option{-Wextra}.
4794 @item -Wenum-compare
4795 @opindex Wenum-compare
4796 @opindex Wno-enum-compare
4797 Warn about a comparison between values of different enumerated types.
4798 In C++ enumeral mismatches in conditional expressions are also
4799 diagnosed and the warning is enabled by default. In C this warning is
4800 enabled by @option{-Wall}.
4802 @item -Wjump-misses-init @r{(C, Objective-C only)}
4803 @opindex Wjump-misses-init
4804 @opindex Wno-jump-misses-init
4805 Warn if a @code{goto} statement or a @code{switch} statement jumps
4806 forward across the initialization of a variable, or jumps backward to a
4807 label after the variable has been initialized. This only warns about
4808 variables that are initialized when they are declared. This warning is
4809 only supported for C and Objective-C; in C++ this sort of branch is an
4812 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4813 can be disabled with the @option{-Wno-jump-misses-init} option.
4815 @item -Wsign-compare
4816 @opindex Wsign-compare
4817 @opindex Wno-sign-compare
4818 @cindex warning for comparison of signed and unsigned values
4819 @cindex comparison of signed and unsigned values, warning
4820 @cindex signed and unsigned values, comparison warning
4821 Warn when a comparison between signed and unsigned values could produce
4822 an incorrect result when the signed value is converted to unsigned.
4823 This warning is also enabled by @option{-Wextra}; to get the other warnings
4824 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4826 @item -Wsign-conversion
4827 @opindex Wsign-conversion
4828 @opindex Wno-sign-conversion
4829 Warn for implicit conversions that may change the sign of an integer
4830 value, like assigning a signed integer expression to an unsigned
4831 integer variable. An explicit cast silences the warning. In C, this
4832 option is enabled also by @option{-Wconversion}.
4834 @item -Wfloat-conversion
4835 @opindex Wfloat-conversion
4836 @opindex Wno-float-conversion
4837 Warn for implicit conversions that reduce the precision of a real value.
4838 This includes conversions from real to integer, and from higher precision
4839 real to lower precision real values. This option is also enabled by
4840 @option{-Wconversion}.
4842 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
4843 @opindex Wsized-deallocation
4844 @opindex Wno-sized-deallocation
4845 Warn about a definition of an unsized deallocation function
4847 void operator delete (void *) noexcept;
4848 void operator delete[] (void *) noexcept;
4850 without a definition of the corresponding sized deallocation function
4852 void operator delete (void *, std::size_t) noexcept;
4853 void operator delete[] (void *, std::size_t) noexcept;
4855 or vice versa. Enabled by @option{-Wextra} along with
4856 @option{-fsized-deallocation}.
4858 @item -Wsizeof-pointer-memaccess
4859 @opindex Wsizeof-pointer-memaccess
4860 @opindex Wno-sizeof-pointer-memaccess
4861 Warn for suspicious length parameters to certain string and memory built-in
4862 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4863 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4864 but a pointer, and suggests a possible fix, or about
4865 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4868 @item -Wsizeof-array-argument
4869 @opindex Wsizeof-array-argument
4870 @opindex Wno-sizeof-array-argument
4871 Warn when the @code{sizeof} operator is applied to a parameter that is
4872 declared as an array in a function definition. This warning is enabled by
4873 default for C and C++ programs.
4875 @item -Wmemset-transposed-args
4876 @opindex Wmemset-transposed-args
4877 @opindex Wno-memset-transposed-args
4878 Warn for suspicious calls to the @code{memset} built-in function, if the
4879 second argument is not zero and the third argument is zero. This warns e.g.@
4880 about @code{memset (buf, sizeof buf, 0)} where most probably
4881 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4882 is only emitted if the third argument is literal zero, if it is some expression
4883 that is folded to zero, or e.g. a cast of zero to some type etc., it
4884 is far less likely that user has mistakenly exchanged the arguments and
4885 no warning is emitted. This warning is enabled by @option{-Wall}.
4889 @opindex Wno-address
4890 Warn about suspicious uses of memory addresses. These include using
4891 the address of a function in a conditional expression, such as
4892 @code{void func(void); if (func)}, and comparisons against the memory
4893 address of a string literal, such as @code{if (x == "abc")}. Such
4894 uses typically indicate a programmer error: the address of a function
4895 always evaluates to true, so their use in a conditional usually
4896 indicate that the programmer forgot the parentheses in a function
4897 call; and comparisons against string literals result in unspecified
4898 behavior and are not portable in C, so they usually indicate that the
4899 programmer intended to use @code{strcmp}. This warning is enabled by
4903 @opindex Wlogical-op
4904 @opindex Wno-logical-op
4905 Warn about suspicious uses of logical operators in expressions.
4906 This includes using logical operators in contexts where a
4907 bit-wise operator is likely to be expected.
4909 @item -Wlogical-not-parentheses
4910 @opindex Wlogical-not-parentheses
4911 @opindex Wno-logical-not-parentheses
4912 Warn about logical not used on the left hand side operand of a comparison.
4913 This option does not warn if the RHS operand is of a boolean type. Its
4914 purpose is to detect suspicious code like the following:
4918 if (!a > 1) @{ @dots{} @}
4921 It is possible to suppress the warning by wrapping the LHS into
4924 if ((!a) > 1) @{ @dots{} @}
4927 This warning is enabled by @option{-Wall}.
4929 @item -Waggregate-return
4930 @opindex Waggregate-return
4931 @opindex Wno-aggregate-return
4932 Warn if any functions that return structures or unions are defined or
4933 called. (In languages where you can return an array, this also elicits
4936 @item -Wno-aggressive-loop-optimizations
4937 @opindex Wno-aggressive-loop-optimizations
4938 @opindex Waggressive-loop-optimizations
4939 Warn if in a loop with constant number of iterations the compiler detects
4940 undefined behavior in some statement during one or more of the iterations.
4942 @item -Wno-attributes
4943 @opindex Wno-attributes
4944 @opindex Wattributes
4945 Do not warn if an unexpected @code{__attribute__} is used, such as
4946 unrecognized attributes, function attributes applied to variables,
4947 etc. This does not stop errors for incorrect use of supported
4950 @item -Wno-builtin-macro-redefined
4951 @opindex Wno-builtin-macro-redefined
4952 @opindex Wbuiltin-macro-redefined
4953 Do not warn if certain built-in macros are redefined. This suppresses
4954 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4955 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4957 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4958 @opindex Wstrict-prototypes
4959 @opindex Wno-strict-prototypes
4960 Warn if a function is declared or defined without specifying the
4961 argument types. (An old-style function definition is permitted without
4962 a warning if preceded by a declaration that specifies the argument
4965 @item -Wold-style-declaration @r{(C and Objective-C only)}
4966 @opindex Wold-style-declaration
4967 @opindex Wno-old-style-declaration
4968 Warn for obsolescent usages, according to the C Standard, in a
4969 declaration. For example, warn if storage-class specifiers like
4970 @code{static} are not the first things in a declaration. This warning
4971 is also enabled by @option{-Wextra}.
4973 @item -Wold-style-definition @r{(C and Objective-C only)}
4974 @opindex Wold-style-definition
4975 @opindex Wno-old-style-definition
4976 Warn if an old-style function definition is used. A warning is given
4977 even if there is a previous prototype.
4979 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4980 @opindex Wmissing-parameter-type
4981 @opindex Wno-missing-parameter-type
4982 A function parameter is declared without a type specifier in K&R-style
4989 This warning is also enabled by @option{-Wextra}.
4991 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4992 @opindex Wmissing-prototypes
4993 @opindex Wno-missing-prototypes
4994 Warn if a global function is defined without a previous prototype
4995 declaration. This warning is issued even if the definition itself
4996 provides a prototype. Use this option to detect global functions
4997 that do not have a matching prototype declaration in a header file.
4998 This option is not valid for C++ because all function declarations
4999 provide prototypes and a non-matching declaration declares an
5000 overload rather than conflict with an earlier declaration.
5001 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5003 @item -Wmissing-declarations
5004 @opindex Wmissing-declarations
5005 @opindex Wno-missing-declarations
5006 Warn if a global function is defined without a previous declaration.
5007 Do so even if the definition itself provides a prototype.
5008 Use this option to detect global functions that are not declared in
5009 header files. In C, no warnings are issued for functions with previous
5010 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5011 missing prototypes. In C++, no warnings are issued for function templates,
5012 or for inline functions, or for functions in anonymous namespaces.
5014 @item -Wmissing-field-initializers
5015 @opindex Wmissing-field-initializers
5016 @opindex Wno-missing-field-initializers
5020 Warn if a structure's initializer has some fields missing. For
5021 example, the following code causes such a warning, because
5022 @code{x.h} is implicitly zero:
5025 struct s @{ int f, g, h; @};
5026 struct s x = @{ 3, 4 @};
5029 This option does not warn about designated initializers, so the following
5030 modification does not trigger a warning:
5033 struct s @{ int f, g, h; @};
5034 struct s x = @{ .f = 3, .g = 4 @};
5037 In C++ this option does not warn either about the empty @{ @}
5038 initializer, for example:
5041 struct s @{ int f, g, h; @};
5045 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5046 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5048 @item -Wno-multichar
5049 @opindex Wno-multichar
5051 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5052 Usually they indicate a typo in the user's code, as they have
5053 implementation-defined values, and should not be used in portable code.
5055 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5056 @opindex Wnormalized=
5057 @opindex Wnormalized
5058 @opindex Wno-normalized
5061 @cindex character set, input normalization
5062 In ISO C and ISO C++, two identifiers are different if they are
5063 different sequences of characters. However, sometimes when characters
5064 outside the basic ASCII character set are used, you can have two
5065 different character sequences that look the same. To avoid confusion,
5066 the ISO 10646 standard sets out some @dfn{normalization rules} which
5067 when applied ensure that two sequences that look the same are turned into
5068 the same sequence. GCC can warn you if you are using identifiers that
5069 have not been normalized; this option controls that warning.
5071 There are four levels of warning supported by GCC@. The default is
5072 @option{-Wnormalized=nfc}, which warns about any identifier that is
5073 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5074 recommended form for most uses. It is equivalent to
5075 @option{-Wnormalized}.
5077 Unfortunately, there are some characters allowed in identifiers by
5078 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5079 identifiers. That is, there's no way to use these symbols in portable
5080 ISO C or C++ and have all your identifiers in NFC@.
5081 @option{-Wnormalized=id} suppresses the warning for these characters.
5082 It is hoped that future versions of the standards involved will correct
5083 this, which is why this option is not the default.
5085 You can switch the warning off for all characters by writing
5086 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5087 only do this if you are using some other normalization scheme (like
5088 ``D''), because otherwise you can easily create bugs that are
5089 literally impossible to see.
5091 Some characters in ISO 10646 have distinct meanings but look identical
5092 in some fonts or display methodologies, especially once formatting has
5093 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5094 LETTER N'', displays just like a regular @code{n} that has been
5095 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5096 normalization scheme to convert all these into a standard form as
5097 well, and GCC warns if your code is not in NFKC if you use
5098 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5099 about every identifier that contains the letter O because it might be
5100 confused with the digit 0, and so is not the default, but may be
5101 useful as a local coding convention if the programming environment
5102 cannot be fixed to display these characters distinctly.
5104 @item -Wno-deprecated
5105 @opindex Wno-deprecated
5106 @opindex Wdeprecated
5107 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5109 @item -Wno-deprecated-declarations
5110 @opindex Wno-deprecated-declarations
5111 @opindex Wdeprecated-declarations
5112 Do not warn about uses of functions (@pxref{Function Attributes}),
5113 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5114 Attributes}) marked as deprecated by using the @code{deprecated}
5118 @opindex Wno-overflow
5120 Do not warn about compile-time overflow in constant expressions.
5125 Warn about One Definition Rule violations during link-time optimization.
5126 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5129 @opindex Wopenm-simd
5130 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5131 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
5132 be used to relax the cost model.
5134 @item -Woverride-init @r{(C and Objective-C only)}
5135 @opindex Woverride-init
5136 @opindex Wno-override-init
5140 Warn if an initialized field without side effects is overridden when
5141 using designated initializers (@pxref{Designated Inits, , Designated
5144 This warning is included in @option{-Wextra}. To get other
5145 @option{-Wextra} warnings without this one, use @option{-Wextra
5146 -Wno-override-init}.
5151 Warn if a structure is given the packed attribute, but the packed
5152 attribute has no effect on the layout or size of the structure.
5153 Such structures may be mis-aligned for little benefit. For
5154 instance, in this code, the variable @code{f.x} in @code{struct bar}
5155 is misaligned even though @code{struct bar} does not itself
5156 have the packed attribute:
5163 @} __attribute__((packed));
5171 @item -Wpacked-bitfield-compat
5172 @opindex Wpacked-bitfield-compat
5173 @opindex Wno-packed-bitfield-compat
5174 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5175 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5176 the change can lead to differences in the structure layout. GCC
5177 informs you when the offset of such a field has changed in GCC 4.4.
5178 For example there is no longer a 4-bit padding between field @code{a}
5179 and @code{b} in this structure:
5186 @} __attribute__ ((packed));
5189 This warning is enabled by default. Use
5190 @option{-Wno-packed-bitfield-compat} to disable this warning.
5195 Warn if padding is included in a structure, either to align an element
5196 of the structure or to align the whole structure. Sometimes when this
5197 happens it is possible to rearrange the fields of the structure to
5198 reduce the padding and so make the structure smaller.
5200 @item -Wredundant-decls
5201 @opindex Wredundant-decls
5202 @opindex Wno-redundant-decls
5203 Warn if anything is declared more than once in the same scope, even in
5204 cases where multiple declaration is valid and changes nothing.
5206 @item -Wnested-externs @r{(C and Objective-C only)}
5207 @opindex Wnested-externs
5208 @opindex Wno-nested-externs
5209 Warn if an @code{extern} declaration is encountered within a function.
5211 @item -Wno-inherited-variadic-ctor
5212 @opindex Winherited-variadic-ctor
5213 @opindex Wno-inherited-variadic-ctor
5214 Suppress warnings about use of C++11 inheriting constructors when the
5215 base class inherited from has a C variadic constructor; the warning is
5216 on by default because the ellipsis is not inherited.
5221 Warn if a function that is declared as inline cannot be inlined.
5222 Even with this option, the compiler does not warn about failures to
5223 inline functions declared in system headers.
5225 The compiler uses a variety of heuristics to determine whether or not
5226 to inline a function. For example, the compiler takes into account
5227 the size of the function being inlined and the amount of inlining
5228 that has already been done in the current function. Therefore,
5229 seemingly insignificant changes in the source program can cause the
5230 warnings produced by @option{-Winline} to appear or disappear.
5232 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5233 @opindex Wno-invalid-offsetof
5234 @opindex Winvalid-offsetof
5235 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5236 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5237 to a non-standard-layout type is undefined. In existing C++ implementations,
5238 however, @code{offsetof} typically gives meaningful results.
5239 This flag is for users who are aware that they are
5240 writing nonportable code and who have deliberately chosen to ignore the
5243 The restrictions on @code{offsetof} may be relaxed in a future version
5244 of the C++ standard.
5246 @item -Wno-int-to-pointer-cast
5247 @opindex Wno-int-to-pointer-cast
5248 @opindex Wint-to-pointer-cast
5249 Suppress warnings from casts to pointer type of an integer of a
5250 different size. In C++, casting to a pointer type of smaller size is
5251 an error. @option{Wint-to-pointer-cast} is enabled by default.
5254 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5255 @opindex Wno-pointer-to-int-cast
5256 @opindex Wpointer-to-int-cast
5257 Suppress warnings from casts from a pointer to an integer type of a
5261 @opindex Winvalid-pch
5262 @opindex Wno-invalid-pch
5263 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5264 the search path but can't be used.
5268 @opindex Wno-long-long
5269 Warn if @code{long long} type is used. This is enabled by either
5270 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5271 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5273 @item -Wvariadic-macros
5274 @opindex Wvariadic-macros
5275 @opindex Wno-variadic-macros
5276 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5277 alternate syntax is used in ISO C99 mode. This is enabled by either
5278 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5279 messages, use @option{-Wno-variadic-macros}.
5283 @opindex Wno-varargs
5284 Warn upon questionable usage of the macros used to handle variable
5285 arguments like @code{va_start}. This is default. To inhibit the
5286 warning messages, use @option{-Wno-varargs}.
5288 @item -Wvector-operation-performance
5289 @opindex Wvector-operation-performance
5290 @opindex Wno-vector-operation-performance
5291 Warn if vector operation is not implemented via SIMD capabilities of the
5292 architecture. Mainly useful for the performance tuning.
5293 Vector operation can be implemented @code{piecewise}, which means that the
5294 scalar operation is performed on every vector element;
5295 @code{in parallel}, which means that the vector operation is implemented
5296 using scalars of wider type, which normally is more performance efficient;
5297 and @code{as a single scalar}, which means that vector fits into a
5300 @item -Wno-virtual-move-assign
5301 @opindex Wvirtual-move-assign
5302 @opindex Wno-virtual-move-assign
5303 Suppress warnings about inheriting from a virtual base with a
5304 non-trivial C++11 move assignment operator. This is dangerous because
5305 if the virtual base is reachable along more than one path, it is
5306 moved multiple times, which can mean both objects end up in the
5307 moved-from state. If the move assignment operator is written to avoid
5308 moving from a moved-from object, this warning can be disabled.
5313 Warn if variable length array is used in the code.
5314 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5315 the variable length array.
5317 @item -Wvolatile-register-var
5318 @opindex Wvolatile-register-var
5319 @opindex Wno-volatile-register-var
5320 Warn if a register variable is declared volatile. The volatile
5321 modifier does not inhibit all optimizations that may eliminate reads
5322 and/or writes to register variables. This warning is enabled by
5325 @item -Wdisabled-optimization
5326 @opindex Wdisabled-optimization
5327 @opindex Wno-disabled-optimization
5328 Warn if a requested optimization pass is disabled. This warning does
5329 not generally indicate that there is anything wrong with your code; it
5330 merely indicates that GCC's optimizers are unable to handle the code
5331 effectively. Often, the problem is that your code is too big or too
5332 complex; GCC refuses to optimize programs when the optimization
5333 itself is likely to take inordinate amounts of time.
5335 @item -Wpointer-sign @r{(C and Objective-C only)}
5336 @opindex Wpointer-sign
5337 @opindex Wno-pointer-sign
5338 Warn for pointer argument passing or assignment with different signedness.
5339 This option is only supported for C and Objective-C@. It is implied by
5340 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5341 @option{-Wno-pointer-sign}.
5343 @item -Wstack-protector
5344 @opindex Wstack-protector
5345 @opindex Wno-stack-protector
5346 This option is only active when @option{-fstack-protector} is active. It
5347 warns about functions that are not protected against stack smashing.
5349 @item -Woverlength-strings
5350 @opindex Woverlength-strings
5351 @opindex Wno-overlength-strings
5352 Warn about string constants that are longer than the ``minimum
5353 maximum'' length specified in the C standard. Modern compilers
5354 generally allow string constants that are much longer than the
5355 standard's minimum limit, but very portable programs should avoid
5356 using longer strings.
5358 The limit applies @emph{after} string constant concatenation, and does
5359 not count the trailing NUL@. In C90, the limit was 509 characters; in
5360 C99, it was raised to 4095. C++98 does not specify a normative
5361 minimum maximum, so we do not diagnose overlength strings in C++@.
5363 This option is implied by @option{-Wpedantic}, and can be disabled with
5364 @option{-Wno-overlength-strings}.
5366 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5367 @opindex Wunsuffixed-float-constants
5369 Issue a warning for any floating constant that does not have
5370 a suffix. When used together with @option{-Wsystem-headers} it
5371 warns about such constants in system header files. This can be useful
5372 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5373 from the decimal floating-point extension to C99.
5375 @item -Wno-designated-init @r{(C and Objective-C only)}
5376 Suppress warnings when a positional initializer is used to initialize
5377 a structure that has been marked with the @code{designated_init}
5382 @node Debugging Options
5383 @section Options for Debugging Your Program or GCC
5384 @cindex options, debugging
5385 @cindex debugging information options
5387 GCC has various special options that are used for debugging
5388 either your program or GCC:
5393 Produce debugging information in the operating system's native format
5394 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5397 On most systems that use stabs format, @option{-g} enables use of extra
5398 debugging information that only GDB can use; this extra information
5399 makes debugging work better in GDB but probably makes other debuggers
5401 refuse to read the program. If you want to control for certain whether
5402 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5403 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5405 GCC allows you to use @option{-g} with
5406 @option{-O}. The shortcuts taken by optimized code may occasionally
5407 produce surprising results: some variables you declared may not exist
5408 at all; flow of control may briefly move where you did not expect it;
5409 some statements may not be executed because they compute constant
5410 results or their values are already at hand; some statements may
5411 execute in different places because they have been moved out of loops.
5413 Nevertheless it proves possible to debug optimized output. This makes
5414 it reasonable to use the optimizer for programs that might have bugs.
5416 The following options are useful when GCC is generated with the
5417 capability for more than one debugging format.
5420 @opindex gsplit-dwarf
5421 Separate as much dwarf debugging information as possible into a
5422 separate output file with the extension .dwo. This option allows
5423 the build system to avoid linking files with debug information. To
5424 be useful, this option requires a debugger capable of reading .dwo
5429 Produce debugging information for use by GDB@. This means to use the
5430 most expressive format available (DWARF 2, stabs, or the native format
5431 if neither of those are supported), including GDB extensions if at all
5436 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5438 @item -ggnu-pubnames
5439 @opindex ggnu-pubnames
5440 Generate .debug_pubnames and .debug_pubtypes sections in a format
5441 suitable for conversion into a GDB@ index. This option is only useful
5442 with a linker that can produce GDB@ index version 7.
5446 Produce debugging information in stabs format (if that is supported),
5447 without GDB extensions. This is the format used by DBX on most BSD
5448 systems. On MIPS, Alpha and System V Release 4 systems this option
5449 produces stabs debugging output that is not understood by DBX or SDB@.
5450 On System V Release 4 systems this option requires the GNU assembler.
5452 @item -feliminate-unused-debug-symbols
5453 @opindex feliminate-unused-debug-symbols
5454 Produce debugging information in stabs format (if that is supported),
5455 for only symbols that are actually used.
5457 @item -femit-class-debug-always
5458 @opindex femit-class-debug-always
5459 Instead of emitting debugging information for a C++ class in only one
5460 object file, emit it in all object files using the class. This option
5461 should be used only with debuggers that are unable to handle the way GCC
5462 normally emits debugging information for classes because using this
5463 option increases the size of debugging information by as much as a
5466 @item -fdebug-types-section
5467 @opindex fdebug-types-section
5468 @opindex fno-debug-types-section
5469 When using DWARF Version 4 or higher, type DIEs can be put into
5470 their own @code{.debug_types} section instead of making them part of the
5471 @code{.debug_info} section. It is more efficient to put them in a separate
5472 comdat sections since the linker can then remove duplicates.
5473 But not all DWARF consumers support @code{.debug_types} sections yet
5474 and on some objects @code{.debug_types} produces larger instead of smaller
5475 debugging information.
5479 Produce debugging information in stabs format (if that is supported),
5480 using GNU extensions understood only by the GNU debugger (GDB)@. The
5481 use of these extensions is likely to make other debuggers crash or
5482 refuse to read the program.
5486 Produce debugging information in COFF format (if that is supported).
5487 This is the format used by SDB on most System V systems prior to
5492 Produce debugging information in XCOFF format (if that is supported).
5493 This is the format used by the DBX debugger on IBM RS/6000 systems.
5497 Produce debugging information in XCOFF format (if that is supported),
5498 using GNU extensions understood only by the GNU debugger (GDB)@. The
5499 use of these extensions is likely to make other debuggers crash or
5500 refuse to read the program, and may cause assemblers other than the GNU
5501 assembler (GAS) to fail with an error.
5503 @item -gdwarf-@var{version}
5504 @opindex gdwarf-@var{version}
5505 Produce debugging information in DWARF format (if that is supported).
5506 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5507 for most targets is 4. DWARF Version 5 is only experimental.
5509 Note that with DWARF Version 2, some ports require and always
5510 use some non-conflicting DWARF 3 extensions in the unwind tables.
5512 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5513 for maximum benefit.
5515 @item -grecord-gcc-switches
5516 @opindex grecord-gcc-switches
5517 This switch causes the command-line options used to invoke the
5518 compiler that may affect code generation to be appended to the
5519 DW_AT_producer attribute in DWARF debugging information. The options
5520 are concatenated with spaces separating them from each other and from
5521 the compiler version. See also @option{-frecord-gcc-switches} for another
5522 way of storing compiler options into the object file. This is the default.
5524 @item -gno-record-gcc-switches
5525 @opindex gno-record-gcc-switches
5526 Disallow appending command-line options to the DW_AT_producer attribute
5527 in DWARF debugging information.
5529 @item -gstrict-dwarf
5530 @opindex gstrict-dwarf
5531 Disallow using extensions of later DWARF standard version than selected
5532 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5533 DWARF extensions from later standard versions is allowed.
5535 @item -gno-strict-dwarf
5536 @opindex gno-strict-dwarf
5537 Allow using extensions of later DWARF standard version than selected with
5538 @option{-gdwarf-@var{version}}.
5540 @item -gz@r{[}=@var{type}@r{]}
5542 Produce compressed debug sections in DWARF format, if that is supported.
5543 If @var{type} is not given, the default type depends on the capabilities
5544 of the assembler and linker used. @var{type} may be one of
5545 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5546 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5547 compression in traditional GNU format). If the linker doesn't support
5548 writing compressed debug sections, the option is rejected. Otherwise,
5549 if the assembler does not support them, @option{-gz} is silently ignored
5550 when producing object files.
5554 Produce debugging information in Alpha/VMS debug format (if that is
5555 supported). This is the format used by DEBUG on Alpha/VMS systems.
5558 @itemx -ggdb@var{level}
5559 @itemx -gstabs@var{level}
5560 @itemx -gcoff@var{level}
5561 @itemx -gxcoff@var{level}
5562 @itemx -gvms@var{level}
5563 Request debugging information and also use @var{level} to specify how
5564 much information. The default level is 2.
5566 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5569 Level 1 produces minimal information, enough for making backtraces in
5570 parts of the program that you don't plan to debug. This includes
5571 descriptions of functions and external variables, and line number
5572 tables, but no information about local variables.
5574 Level 3 includes extra information, such as all the macro definitions
5575 present in the program. Some debuggers support macro expansion when
5576 you use @option{-g3}.
5578 @option{-gdwarf-2} does not accept a concatenated debug level, because
5579 GCC used to support an option @option{-gdwarf} that meant to generate
5580 debug information in version 1 of the DWARF format (which is very
5581 different from version 2), and it would have been too confusing. That
5582 debug format is long obsolete, but the option cannot be changed now.
5583 Instead use an additional @option{-g@var{level}} option to change the
5584 debug level for DWARF.
5588 Turn off generation of debug info, if leaving out this option
5589 generates it, or turn it on at level 2 otherwise. The position of this
5590 argument in the command line does not matter; it takes effect after all
5591 other options are processed, and it does so only once, no matter how
5592 many times it is given. This is mainly intended to be used with
5593 @option{-fcompare-debug}.
5595 @item -fsanitize=address
5596 @opindex fsanitize=address
5597 Enable AddressSanitizer, a fast memory error detector.
5598 Memory access instructions are instrumented to detect
5599 out-of-bounds and use-after-free bugs.
5600 See @uref{http://code.google.com/p/address-sanitizer/} for
5601 more details. The run-time behavior can be influenced using the
5602 @env{ASAN_OPTIONS} environment variable; see
5603 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5604 a list of supported options.
5606 @item -fsanitize=kernel-address
5607 @opindex fsanitize=kernel-address
5608 Enable AddressSanitizer for Linux kernel.
5609 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5611 @item -fsanitize=thread
5612 @opindex fsanitize=thread
5613 Enable ThreadSanitizer, a fast data race detector.
5614 Memory access instructions are instrumented to detect
5615 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5616 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5617 environment variable; see
5618 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5621 @item -fsanitize=leak
5622 @opindex fsanitize=leak
5623 Enable LeakSanitizer, a memory leak detector.
5624 This option only matters for linking of executables and if neither
5625 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5626 case the executable is linked against a library that overrides @code{malloc}
5627 and other allocator functions. See
5628 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5629 details. The run-time behavior can be influenced using the
5630 @env{LSAN_OPTIONS} environment variable.
5632 @item -fsanitize=undefined
5633 @opindex fsanitize=undefined
5634 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5635 Various computations are instrumented to detect undefined behavior
5636 at runtime. Current suboptions are:
5640 @item -fsanitize=shift
5641 @opindex fsanitize=shift
5642 This option enables checking that the result of a shift operation is
5643 not undefined. Note that what exactly is considered undefined differs
5644 slightly between C and C++, as well as between ISO C90 and C99, etc.
5646 @item -fsanitize=integer-divide-by-zero
5647 @opindex fsanitize=integer-divide-by-zero
5648 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5650 @item -fsanitize=unreachable
5651 @opindex fsanitize=unreachable
5652 With this option, the compiler turns the @code{__builtin_unreachable}
5653 call into a diagnostics message call instead. When reaching the
5654 @code{__builtin_unreachable} call, the behavior is undefined.
5656 @item -fsanitize=vla-bound
5657 @opindex fsanitize=vla-bound
5658 This option instructs the compiler to check that the size of a variable
5659 length array is positive.
5661 @item -fsanitize=null
5662 @opindex fsanitize=null
5663 This option enables pointer checking. Particularly, the application
5664 built with this option turned on will issue an error message when it
5665 tries to dereference a NULL pointer, or if a reference (possibly an
5666 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5667 on an object pointed by a NULL pointer.
5669 @item -fsanitize=return
5670 @opindex fsanitize=return
5671 This option enables return statement checking. Programs
5672 built with this option turned on will issue an error message
5673 when the end of a non-void function is reached without actually
5674 returning a value. This option works in C++ only.
5676 @item -fsanitize=signed-integer-overflow
5677 @opindex fsanitize=signed-integer-overflow
5678 This option enables signed integer overflow checking. We check that
5679 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5680 does not overflow in the signed arithmetics. Note, integer promotion
5681 rules must be taken into account. That is, the following is not an
5684 signed char a = SCHAR_MAX;
5688 @item -fsanitize=bounds
5689 @opindex fsanitize=bounds
5690 This option enables instrumentation of array bounds. Various out of bounds
5691 accesses are detected. Flexible array members and initializers of variables
5692 with static storage are not instrumented.
5694 @item -fsanitize=alignment
5695 @opindex fsanitize=alignment
5697 This option enables checking of alignment of pointers when they are
5698 dereferenced, or when a reference is bound to insufficiently aligned target,
5699 or when a method or constructor is invoked on insufficiently aligned object.
5701 @item -fsanitize=object-size
5702 @opindex fsanitize=object-size
5703 This option enables instrumentation of memory references using the
5704 @code{__builtin_object_size} function. Various out of bounds pointer
5705 accesses are detected.
5707 @item -fsanitize=float-divide-by-zero
5708 @opindex fsanitize=float-divide-by-zero
5709 Detect floating-point division by zero. Unlike other similar options,
5710 @option{-fsanitize=float-divide-by-zero} is not enabled by
5711 @option{-fsanitize=undefined}, since floating-point division by zero can
5712 be a legitimate way of obtaining infinities and NaNs.
5714 @item -fsanitize=float-cast-overflow
5715 @opindex fsanitize=float-cast-overflow
5716 This option enables floating-point type to integer conversion checking.
5717 We check that the result of the conversion does not overflow.
5718 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
5719 not enabled by @option{-fsanitize=undefined}.
5720 This option does not work well with @code{FE_INVALID} exceptions enabled.
5722 @item -fsanitize=nonnull-attribute
5723 @opindex fsanitize=nonnull-attribute
5725 This option enables instrumentation of calls, checking whether null values
5726 are not passed to arguments marked as requiring a non-null value by the
5727 @code{nonnull} function attribute.
5729 @item -fsanitize=returns-nonnull-attribute
5730 @opindex fsanitize=returns-nonnull-attribute
5732 This option enables instrumentation of return statements in functions
5733 marked with @code{returns_nonnull} function attribute, to detect returning
5734 of null values from such functions.
5736 @item -fsanitize=bool
5737 @opindex fsanitize=bool
5739 This option enables instrumentation of loads from bool. If a value other
5740 than 0/1 is loaded, a run-time error is issued.
5742 @item -fsanitize=enum
5743 @opindex fsanitize=enum
5745 This option enables instrumentation of loads from an enum type. If
5746 a value outside the range of values for the enum type is loaded,
5747 a run-time error is issued.
5751 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5752 @option{-fsanitize=undefined} gives a diagnostic message.
5753 This currently works only for the C family of languages.
5755 @item -fno-sanitize=all
5756 @opindex fno-sanitize=all
5758 This option disables all previously enabled sanitizers.
5759 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
5762 @item -fasan-shadow-offset=@var{number}
5763 @opindex fasan-shadow-offset
5764 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5765 It is useful for experimenting with different shadow memory layouts in
5766 Kernel AddressSanitizer.
5768 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5769 @opindex fsanitize-recover
5770 @opindex fno-sanitize-recover
5771 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5772 mentioned in comma-separated list of @var{opts}. Enabling this option
5773 for a sanitizer component causes it to attempt to continue
5774 running the program as if no error happened. This means multiple
5775 runtime errors can be reported in a single program run, and the exit
5776 code of the program may indicate success even when errors
5777 have been reported. The @option{-fno-sanitize-recover=} can be used to alter
5778 this behavior: only the first detected error is reported
5779 and program then exits with a non-zero exit code.
5781 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5782 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5783 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5784 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5785 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
5786 accepted, the former enables recovery for all sanitizers that support it,
5787 the latter disables recovery for all sanitizers that support it.
5789 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5791 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5794 Similarly @option{-fno-sanitize-recover} is equivalent to
5796 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5799 @item -fsanitize-undefined-trap-on-error
5800 @opindex fsanitize-undefined-trap-on-error
5801 The @option{-fsanitize-undefined-trap-on-error} instructs the compiler to
5802 report undefined behavior using @code{__builtin_trap} rather than
5803 a @code{libubsan} library routine. The advantage of this is that the
5804 @code{libubsan} library is not needed and is not linked in, so this
5805 is usable even in freestanding environments.
5807 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5808 @opindex fdump-final-insns
5809 Dump the final internal representation (RTL) to @var{file}. If the
5810 optional argument is omitted (or if @var{file} is @code{.}), the name
5811 of the dump file is determined by appending @code{.gkd} to the
5812 compilation output file name.
5814 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5815 @opindex fcompare-debug
5816 @opindex fno-compare-debug
5817 If no error occurs during compilation, run the compiler a second time,
5818 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5819 passed to the second compilation. Dump the final internal
5820 representation in both compilations, and print an error if they differ.
5822 If the equal sign is omitted, the default @option{-gtoggle} is used.
5824 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5825 and nonzero, implicitly enables @option{-fcompare-debug}. If
5826 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5827 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5830 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5831 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5832 of the final representation and the second compilation, preventing even
5833 @env{GCC_COMPARE_DEBUG} from taking effect.
5835 To verify full coverage during @option{-fcompare-debug} testing, set
5836 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
5837 which GCC rejects as an invalid option in any actual compilation
5838 (rather than preprocessing, assembly or linking). To get just a
5839 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5840 not overridden} will do.
5842 @item -fcompare-debug-second
5843 @opindex fcompare-debug-second
5844 This option is implicitly passed to the compiler for the second
5845 compilation requested by @option{-fcompare-debug}, along with options to
5846 silence warnings, and omitting other options that would cause
5847 side-effect compiler outputs to files or to the standard output. Dump
5848 files and preserved temporary files are renamed so as to contain the
5849 @code{.gk} additional extension during the second compilation, to avoid
5850 overwriting those generated by the first.
5852 When this option is passed to the compiler driver, it causes the
5853 @emph{first} compilation to be skipped, which makes it useful for little
5854 other than debugging the compiler proper.
5856 @item -feliminate-dwarf2-dups
5857 @opindex feliminate-dwarf2-dups
5858 Compress DWARF 2 debugging information by eliminating duplicated
5859 information about each symbol. This option only makes sense when
5860 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5862 @item -femit-struct-debug-baseonly
5863 @opindex femit-struct-debug-baseonly
5864 Emit debug information for struct-like types
5865 only when the base name of the compilation source file
5866 matches the base name of file in which the struct is defined.
5868 This option substantially reduces the size of debugging information,
5869 but at significant potential loss in type information to the debugger.
5870 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5871 See @option{-femit-struct-debug-detailed} for more detailed control.
5873 This option works only with DWARF 2.
5875 @item -femit-struct-debug-reduced
5876 @opindex femit-struct-debug-reduced
5877 Emit debug information for struct-like types
5878 only when the base name of the compilation source file
5879 matches the base name of file in which the type is defined,
5880 unless the struct is a template or defined in a system header.
5882 This option significantly reduces the size of debugging information,
5883 with some potential loss in type information to the debugger.
5884 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5885 See @option{-femit-struct-debug-detailed} for more detailed control.
5887 This option works only with DWARF 2.
5889 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5890 @opindex femit-struct-debug-detailed
5891 Specify the struct-like types
5892 for which the compiler generates debug information.
5893 The intent is to reduce duplicate struct debug information
5894 between different object files within the same program.
5896 This option is a detailed version of
5897 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5898 which serves for most needs.
5900 A specification has the syntax@*
5901 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5903 The optional first word limits the specification to
5904 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5905 A struct type is used directly when it is the type of a variable, member.
5906 Indirect uses arise through pointers to structs.
5907 That is, when use of an incomplete struct is valid, the use is indirect.
5909 @samp{struct one direct; struct two * indirect;}.
5911 The optional second word limits the specification to
5912 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5913 Generic structs are a bit complicated to explain.
5914 For C++, these are non-explicit specializations of template classes,
5915 or non-template classes within the above.
5916 Other programming languages have generics,
5917 but @option{-femit-struct-debug-detailed} does not yet implement them.
5919 The third word specifies the source files for those
5920 structs for which the compiler should emit debug information.
5921 The values @samp{none} and @samp{any} have the normal meaning.
5922 The value @samp{base} means that
5923 the base of name of the file in which the type declaration appears
5924 must match the base of the name of the main compilation file.
5925 In practice, this means that when compiling @file{foo.c}, debug information
5926 is generated for types declared in that file and @file{foo.h},
5927 but not other header files.
5928 The value @samp{sys} means those types satisfying @samp{base}
5929 or declared in system or compiler headers.
5931 You may need to experiment to determine the best settings for your application.
5933 The default is @option{-femit-struct-debug-detailed=all}.
5935 This option works only with DWARF 2.
5937 @item -fno-merge-debug-strings
5938 @opindex fmerge-debug-strings
5939 @opindex fno-merge-debug-strings
5940 Direct the linker to not merge together strings in the debugging
5941 information that are identical in different object files. Merging is
5942 not supported by all assemblers or linkers. Merging decreases the size
5943 of the debug information in the output file at the cost of increasing
5944 link processing time. Merging is enabled by default.
5946 @item -fdebug-prefix-map=@var{old}=@var{new}
5947 @opindex fdebug-prefix-map
5948 When compiling files in directory @file{@var{old}}, record debugging
5949 information describing them as in @file{@var{new}} instead.
5951 @item -fno-dwarf2-cfi-asm
5952 @opindex fdwarf2-cfi-asm
5953 @opindex fno-dwarf2-cfi-asm
5954 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5955 instead of using GAS @code{.cfi_*} directives.
5957 @cindex @command{prof}
5960 Generate extra code to write profile information suitable for the
5961 analysis program @command{prof}. You must use this option when compiling
5962 the source files you want data about, and you must also use it when
5965 @cindex @command{gprof}
5968 Generate extra code to write profile information suitable for the
5969 analysis program @command{gprof}. You must use this option when compiling
5970 the source files you want data about, and you must also use it when
5975 Makes the compiler print out each function name as it is compiled, and
5976 print some statistics about each pass when it finishes.
5979 @opindex ftime-report
5980 Makes the compiler print some statistics about the time consumed by each
5981 pass when it finishes.
5984 @opindex fmem-report
5985 Makes the compiler print some statistics about permanent memory
5986 allocation when it finishes.
5988 @item -fmem-report-wpa
5989 @opindex fmem-report-wpa
5990 Makes the compiler print some statistics about permanent memory
5991 allocation for the WPA phase only.
5993 @item -fpre-ipa-mem-report
5994 @opindex fpre-ipa-mem-report
5995 @item -fpost-ipa-mem-report
5996 @opindex fpost-ipa-mem-report
5997 Makes the compiler print some statistics about permanent memory
5998 allocation before or after interprocedural optimization.
6000 @item -fprofile-report
6001 @opindex fprofile-report
6002 Makes the compiler print some statistics about consistency of the
6003 (estimated) profile and effect of individual passes.
6006 @opindex fstack-usage
6007 Makes the compiler output stack usage information for the program, on a
6008 per-function basis. The filename for the dump is made by appending
6009 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6010 the output file, if explicitly specified and it is not an executable,
6011 otherwise it is the basename of the source file. An entry is made up
6016 The name of the function.
6020 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6023 The qualifier @code{static} means that the function manipulates the stack
6024 statically: a fixed number of bytes are allocated for the frame on function
6025 entry and released on function exit; no stack adjustments are otherwise made
6026 in the function. The second field is this fixed number of bytes.
6028 The qualifier @code{dynamic} means that the function manipulates the stack
6029 dynamically: in addition to the static allocation described above, stack
6030 adjustments are made in the body of the function, for example to push/pop
6031 arguments around function calls. If the qualifier @code{bounded} is also
6032 present, the amount of these adjustments is bounded at compile time and
6033 the second field is an upper bound of the total amount of stack used by
6034 the function. If it is not present, the amount of these adjustments is
6035 not bounded at compile time and the second field only represents the
6038 @item -fprofile-arcs
6039 @opindex fprofile-arcs
6040 Add code so that program flow @dfn{arcs} are instrumented. During
6041 execution the program records how many times each branch and call is
6042 executed and how many times it is taken or returns. When the compiled
6043 program exits it saves this data to a file called
6044 @file{@var{auxname}.gcda} for each source file. The data may be used for
6045 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6046 test coverage analysis (@option{-ftest-coverage}). Each object file's
6047 @var{auxname} is generated from the name of the output file, if
6048 explicitly specified and it is not the final executable, otherwise it is
6049 the basename of the source file. In both cases any suffix is removed
6050 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6051 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6052 @xref{Cross-profiling}.
6054 @cindex @command{gcov}
6058 This option is used to compile and link code instrumented for coverage
6059 analysis. The option is a synonym for @option{-fprofile-arcs}
6060 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6061 linking). See the documentation for those options for more details.
6066 Compile the source files with @option{-fprofile-arcs} plus optimization
6067 and code generation options. For test coverage analysis, use the
6068 additional @option{-ftest-coverage} option. You do not need to profile
6069 every source file in a program.
6072 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6073 (the latter implies the former).
6076 Run the program on a representative workload to generate the arc profile
6077 information. This may be repeated any number of times. You can run
6078 concurrent instances of your program, and provided that the file system
6079 supports locking, the data files will be correctly updated. Also
6080 @code{fork} calls are detected and correctly handled (double counting
6084 For profile-directed optimizations, compile the source files again with
6085 the same optimization and code generation options plus
6086 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6087 Control Optimization}).
6090 For test coverage analysis, use @command{gcov} to produce human readable
6091 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6092 @command{gcov} documentation for further information.
6096 With @option{-fprofile-arcs}, for each function of your program GCC
6097 creates a program flow graph, then finds a spanning tree for the graph.
6098 Only arcs that are not on the spanning tree have to be instrumented: the
6099 compiler adds code to count the number of times that these arcs are
6100 executed. When an arc is the only exit or only entrance to a block, the
6101 instrumentation code can be added to the block; otherwise, a new basic
6102 block must be created to hold the instrumentation code.
6105 @item -ftest-coverage
6106 @opindex ftest-coverage
6107 Produce a notes file that the @command{gcov} code-coverage utility
6108 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6109 show program coverage. Each source file's note file is called
6110 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6111 above for a description of @var{auxname} and instructions on how to
6112 generate test coverage data. Coverage data matches the source files
6113 more closely if you do not optimize.
6115 @item -fdbg-cnt-list
6116 @opindex fdbg-cnt-list
6117 Print the name and the counter upper bound for all debug counters.
6120 @item -fdbg-cnt=@var{counter-value-list}
6122 Set the internal debug counter upper bound. @var{counter-value-list}
6123 is a comma-separated list of @var{name}:@var{value} pairs
6124 which sets the upper bound of each debug counter @var{name} to @var{value}.
6125 All debug counters have the initial upper bound of @code{UINT_MAX};
6126 thus @code{dbg_cnt} returns true always unless the upper bound
6127 is set by this option.
6128 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6129 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6131 @item -fenable-@var{kind}-@var{pass}
6132 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6136 This is a set of options that are used to explicitly disable/enable
6137 optimization passes. These options are intended for use for debugging GCC.
6138 Compiler users should use regular options for enabling/disabling
6143 @item -fdisable-ipa-@var{pass}
6144 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6145 statically invoked in the compiler multiple times, the pass name should be
6146 appended with a sequential number starting from 1.
6148 @item -fdisable-rtl-@var{pass}
6149 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6150 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6151 statically invoked in the compiler multiple times, the pass name should be
6152 appended with a sequential number starting from 1. @var{range-list} is a
6153 comma-separated list of function ranges or assembler names. Each range is a number
6154 pair separated by a colon. The range is inclusive in both ends. If the range
6155 is trivial, the number pair can be simplified as a single number. If the
6156 function's call graph node's @var{uid} falls within one of the specified ranges,
6157 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6158 function header of a dump file, and the pass names can be dumped by using
6159 option @option{-fdump-passes}.
6161 @item -fdisable-tree-@var{pass}
6162 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6163 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6166 @item -fenable-ipa-@var{pass}
6167 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6168 statically invoked in the compiler multiple times, the pass name should be
6169 appended with a sequential number starting from 1.
6171 @item -fenable-rtl-@var{pass}
6172 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6173 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6174 description and examples.
6176 @item -fenable-tree-@var{pass}
6177 @itemx -fenable-tree-@var{pass}=@var{range-list}
6178 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6179 of option arguments.
6183 Here are some examples showing uses of these options.
6187 # disable ccp1 for all functions
6189 # disable complete unroll for function whose cgraph node uid is 1
6190 -fenable-tree-cunroll=1
6191 # disable gcse2 for functions at the following ranges [1,1],
6192 # [300,400], and [400,1000]
6193 # disable gcse2 for functions foo and foo2
6194 -fdisable-rtl-gcse2=foo,foo2
6195 # disable early inlining
6196 -fdisable-tree-einline
6197 # disable ipa inlining
6198 -fdisable-ipa-inline
6199 # enable tree full unroll
6200 -fenable-tree-unroll
6204 @item -d@var{letters}
6205 @itemx -fdump-rtl-@var{pass}
6206 @itemx -fdump-rtl-@var{pass}=@var{filename}
6208 @opindex fdump-rtl-@var{pass}
6209 Says to make debugging dumps during compilation at times specified by
6210 @var{letters}. This is used for debugging the RTL-based passes of the
6211 compiler. The file names for most of the dumps are made by appending
6212 a pass number and a word to the @var{dumpname}, and the files are
6213 created in the directory of the output file. In case of
6214 @option{=@var{filename}} option, the dump is output on the given file
6215 instead of the pass numbered dump files. Note that the pass number is
6216 computed statically as passes get registered into the pass manager.
6217 Thus the numbering is not related to the dynamic order of execution of
6218 passes. In particular, a pass installed by a plugin could have a
6219 number over 200 even if it executed quite early. @var{dumpname} is
6220 generated from the name of the output file, if explicitly specified
6221 and it is not an executable, otherwise it is the basename of the
6222 source file. These switches may have different effects when
6223 @option{-E} is used for preprocessing.
6225 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6226 @option{-d} option @var{letters}. Here are the possible
6227 letters for use in @var{pass} and @var{letters}, and their meanings:
6231 @item -fdump-rtl-alignments
6232 @opindex fdump-rtl-alignments
6233 Dump after branch alignments have been computed.
6235 @item -fdump-rtl-asmcons
6236 @opindex fdump-rtl-asmcons
6237 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6239 @item -fdump-rtl-auto_inc_dec
6240 @opindex fdump-rtl-auto_inc_dec
6241 Dump after auto-inc-dec discovery. This pass is only run on
6242 architectures that have auto inc or auto dec instructions.
6244 @item -fdump-rtl-barriers
6245 @opindex fdump-rtl-barriers
6246 Dump after cleaning up the barrier instructions.
6248 @item -fdump-rtl-bbpart
6249 @opindex fdump-rtl-bbpart
6250 Dump after partitioning hot and cold basic blocks.
6252 @item -fdump-rtl-bbro
6253 @opindex fdump-rtl-bbro
6254 Dump after block reordering.
6256 @item -fdump-rtl-btl1
6257 @itemx -fdump-rtl-btl2
6258 @opindex fdump-rtl-btl2
6259 @opindex fdump-rtl-btl2
6260 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6261 after the two branch
6262 target load optimization passes.
6264 @item -fdump-rtl-bypass
6265 @opindex fdump-rtl-bypass
6266 Dump after jump bypassing and control flow optimizations.
6268 @item -fdump-rtl-combine
6269 @opindex fdump-rtl-combine
6270 Dump after the RTL instruction combination pass.
6272 @item -fdump-rtl-compgotos
6273 @opindex fdump-rtl-compgotos
6274 Dump after duplicating the computed gotos.
6276 @item -fdump-rtl-ce1
6277 @itemx -fdump-rtl-ce2
6278 @itemx -fdump-rtl-ce3
6279 @opindex fdump-rtl-ce1
6280 @opindex fdump-rtl-ce2
6281 @opindex fdump-rtl-ce3
6282 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6283 @option{-fdump-rtl-ce3} enable dumping after the three
6284 if conversion passes.
6286 @item -fdump-rtl-cprop_hardreg
6287 @opindex fdump-rtl-cprop_hardreg
6288 Dump after hard register copy propagation.
6290 @item -fdump-rtl-csa
6291 @opindex fdump-rtl-csa
6292 Dump after combining stack adjustments.
6294 @item -fdump-rtl-cse1
6295 @itemx -fdump-rtl-cse2
6296 @opindex fdump-rtl-cse1
6297 @opindex fdump-rtl-cse2
6298 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6299 the two common subexpression elimination passes.
6301 @item -fdump-rtl-dce
6302 @opindex fdump-rtl-dce
6303 Dump after the standalone dead code elimination passes.
6305 @item -fdump-rtl-dbr
6306 @opindex fdump-rtl-dbr
6307 Dump after delayed branch scheduling.
6309 @item -fdump-rtl-dce1
6310 @itemx -fdump-rtl-dce2
6311 @opindex fdump-rtl-dce1
6312 @opindex fdump-rtl-dce2
6313 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6314 the two dead store elimination passes.
6317 @opindex fdump-rtl-eh
6318 Dump after finalization of EH handling code.
6320 @item -fdump-rtl-eh_ranges
6321 @opindex fdump-rtl-eh_ranges
6322 Dump after conversion of EH handling range regions.
6324 @item -fdump-rtl-expand
6325 @opindex fdump-rtl-expand
6326 Dump after RTL generation.
6328 @item -fdump-rtl-fwprop1
6329 @itemx -fdump-rtl-fwprop2
6330 @opindex fdump-rtl-fwprop1
6331 @opindex fdump-rtl-fwprop2
6332 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6333 dumping after the two forward propagation passes.
6335 @item -fdump-rtl-gcse1
6336 @itemx -fdump-rtl-gcse2
6337 @opindex fdump-rtl-gcse1
6338 @opindex fdump-rtl-gcse2
6339 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6340 after global common subexpression elimination.
6342 @item -fdump-rtl-init-regs
6343 @opindex fdump-rtl-init-regs
6344 Dump after the initialization of the registers.
6346 @item -fdump-rtl-initvals
6347 @opindex fdump-rtl-initvals
6348 Dump after the computation of the initial value sets.
6350 @item -fdump-rtl-into_cfglayout
6351 @opindex fdump-rtl-into_cfglayout
6352 Dump after converting to cfglayout mode.
6354 @item -fdump-rtl-ira
6355 @opindex fdump-rtl-ira
6356 Dump after iterated register allocation.
6358 @item -fdump-rtl-jump
6359 @opindex fdump-rtl-jump
6360 Dump after the second jump optimization.
6362 @item -fdump-rtl-loop2
6363 @opindex fdump-rtl-loop2
6364 @option{-fdump-rtl-loop2} enables dumping after the rtl
6365 loop optimization passes.
6367 @item -fdump-rtl-mach
6368 @opindex fdump-rtl-mach
6369 Dump after performing the machine dependent reorganization pass, if that
6372 @item -fdump-rtl-mode_sw
6373 @opindex fdump-rtl-mode_sw
6374 Dump after removing redundant mode switches.
6376 @item -fdump-rtl-rnreg
6377 @opindex fdump-rtl-rnreg
6378 Dump after register renumbering.
6380 @item -fdump-rtl-outof_cfglayout
6381 @opindex fdump-rtl-outof_cfglayout
6382 Dump after converting from cfglayout mode.
6384 @item -fdump-rtl-peephole2
6385 @opindex fdump-rtl-peephole2
6386 Dump after the peephole pass.
6388 @item -fdump-rtl-postreload
6389 @opindex fdump-rtl-postreload
6390 Dump after post-reload optimizations.
6392 @item -fdump-rtl-pro_and_epilogue
6393 @opindex fdump-rtl-pro_and_epilogue
6394 Dump after generating the function prologues and epilogues.
6396 @item -fdump-rtl-sched1
6397 @itemx -fdump-rtl-sched2
6398 @opindex fdump-rtl-sched1
6399 @opindex fdump-rtl-sched2
6400 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6401 after the basic block scheduling passes.
6403 @item -fdump-rtl-ree
6404 @opindex fdump-rtl-ree
6405 Dump after sign/zero extension elimination.
6407 @item -fdump-rtl-seqabstr
6408 @opindex fdump-rtl-seqabstr
6409 Dump after common sequence discovery.
6411 @item -fdump-rtl-shorten
6412 @opindex fdump-rtl-shorten
6413 Dump after shortening branches.
6415 @item -fdump-rtl-sibling
6416 @opindex fdump-rtl-sibling
6417 Dump after sibling call optimizations.
6419 @item -fdump-rtl-split1
6420 @itemx -fdump-rtl-split2
6421 @itemx -fdump-rtl-split3
6422 @itemx -fdump-rtl-split4
6423 @itemx -fdump-rtl-split5
6424 @opindex fdump-rtl-split1
6425 @opindex fdump-rtl-split2
6426 @opindex fdump-rtl-split3
6427 @opindex fdump-rtl-split4
6428 @opindex fdump-rtl-split5
6429 These options enable dumping after five rounds of
6430 instruction splitting.
6432 @item -fdump-rtl-sms
6433 @opindex fdump-rtl-sms
6434 Dump after modulo scheduling. This pass is only run on some
6437 @item -fdump-rtl-stack
6438 @opindex fdump-rtl-stack
6439 Dump after conversion from GCC's ``flat register file'' registers to the
6440 x87's stack-like registers. This pass is only run on x86 variants.
6442 @item -fdump-rtl-subreg1
6443 @itemx -fdump-rtl-subreg2
6444 @opindex fdump-rtl-subreg1
6445 @opindex fdump-rtl-subreg2
6446 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6447 the two subreg expansion passes.
6449 @item -fdump-rtl-unshare
6450 @opindex fdump-rtl-unshare
6451 Dump after all rtl has been unshared.
6453 @item -fdump-rtl-vartrack
6454 @opindex fdump-rtl-vartrack
6455 Dump after variable tracking.
6457 @item -fdump-rtl-vregs
6458 @opindex fdump-rtl-vregs
6459 Dump after converting virtual registers to hard registers.
6461 @item -fdump-rtl-web
6462 @opindex fdump-rtl-web
6463 Dump after live range splitting.
6465 @item -fdump-rtl-regclass
6466 @itemx -fdump-rtl-subregs_of_mode_init
6467 @itemx -fdump-rtl-subregs_of_mode_finish
6468 @itemx -fdump-rtl-dfinit
6469 @itemx -fdump-rtl-dfinish
6470 @opindex fdump-rtl-regclass
6471 @opindex fdump-rtl-subregs_of_mode_init
6472 @opindex fdump-rtl-subregs_of_mode_finish
6473 @opindex fdump-rtl-dfinit
6474 @opindex fdump-rtl-dfinish
6475 These dumps are defined but always produce empty files.
6478 @itemx -fdump-rtl-all
6480 @opindex fdump-rtl-all
6481 Produce all the dumps listed above.
6485 Annotate the assembler output with miscellaneous debugging information.
6489 Dump all macro definitions, at the end of preprocessing, in addition to
6494 Produce a core dump whenever an error occurs.
6498 Annotate the assembler output with a comment indicating which
6499 pattern and alternative is used. The length of each instruction is
6504 Dump the RTL in the assembler output as a comment before each instruction.
6505 Also turns on @option{-dp} annotation.
6509 Just generate RTL for a function instead of compiling it. Usually used
6510 with @option{-fdump-rtl-expand}.
6514 @opindex fdump-noaddr
6515 When doing debugging dumps, suppress address output. This makes it more
6516 feasible to use diff on debugging dumps for compiler invocations with
6517 different compiler binaries and/or different
6518 text / bss / data / heap / stack / dso start locations.
6521 @opindex freport-bug
6522 Collect and dump debug information into temporary file if ICE in C/C++
6525 @item -fdump-unnumbered
6526 @opindex fdump-unnumbered
6527 When doing debugging dumps, suppress instruction numbers and address output.
6528 This makes it more feasible to use diff on debugging dumps for compiler
6529 invocations with different options, in particular with and without
6532 @item -fdump-unnumbered-links
6533 @opindex fdump-unnumbered-links
6534 When doing debugging dumps (see @option{-d} option above), suppress
6535 instruction numbers for the links to the previous and next instructions
6538 @item -fdump-translation-unit @r{(C++ only)}
6539 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6540 @opindex fdump-translation-unit
6541 Dump a representation of the tree structure for the entire translation
6542 unit to a file. The file name is made by appending @file{.tu} to the
6543 source file name, and the file is created in the same directory as the
6544 output file. If the @samp{-@var{options}} form is used, @var{options}
6545 controls the details of the dump as described for the
6546 @option{-fdump-tree} options.
6548 @item -fdump-class-hierarchy @r{(C++ only)}
6549 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6550 @opindex fdump-class-hierarchy
6551 Dump a representation of each class's hierarchy and virtual function
6552 table layout to a file. The file name is made by appending
6553 @file{.class} to the source file name, and the file is created in the
6554 same directory as the output file. If the @samp{-@var{options}} form
6555 is used, @var{options} controls the details of the dump as described
6556 for the @option{-fdump-tree} options.
6558 @item -fdump-ipa-@var{switch}
6560 Control the dumping at various stages of inter-procedural analysis
6561 language tree to a file. The file name is generated by appending a
6562 switch specific suffix to the source file name, and the file is created
6563 in the same directory as the output file. The following dumps are
6568 Enables all inter-procedural analysis dumps.
6571 Dumps information about call-graph optimization, unused function removal,
6572 and inlining decisions.
6575 Dump after function inlining.
6580 @opindex fdump-passes
6581 Dump the list of optimization passes that are turned on and off by
6582 the current command-line options.
6584 @item -fdump-statistics-@var{option}
6585 @opindex fdump-statistics
6586 Enable and control dumping of pass statistics in a separate file. The
6587 file name is generated by appending a suffix ending in
6588 @samp{.statistics} to the source file name, and the file is created in
6589 the same directory as the output file. If the @samp{-@var{option}}
6590 form is used, @samp{-stats} causes counters to be summed over the
6591 whole compilation unit while @samp{-details} dumps every event as
6592 the passes generate them. The default with no option is to sum
6593 counters for each function compiled.
6595 @item -fdump-tree-@var{switch}
6596 @itemx -fdump-tree-@var{switch}-@var{options}
6597 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6599 Control the dumping at various stages of processing the intermediate
6600 language tree to a file. The file name is generated by appending a
6601 switch-specific suffix to the source file name, and the file is
6602 created in the same directory as the output file. In case of
6603 @option{=@var{filename}} option, the dump is output on the given file
6604 instead of the auto named dump files. If the @samp{-@var{options}}
6605 form is used, @var{options} is a list of @samp{-} separated options
6606 which control the details of the dump. Not all options are applicable
6607 to all dumps; those that are not meaningful are ignored. The
6608 following options are available
6612 Print the address of each node. Usually this is not meaningful as it
6613 changes according to the environment and source file. Its primary use
6614 is for tying up a dump file with a debug environment.
6616 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6617 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6618 use working backward from mangled names in the assembly file.
6620 When dumping front-end intermediate representations, inhibit dumping
6621 of members of a scope or body of a function merely because that scope
6622 has been reached. Only dump such items when they are directly reachable
6625 When dumping pretty-printed trees, this option inhibits dumping the
6626 bodies of control structures.
6628 When dumping RTL, print the RTL in slim (condensed) form instead of
6629 the default LISP-like representation.
6631 Print a raw representation of the tree. By default, trees are
6632 pretty-printed into a C-like representation.
6634 Enable more detailed dumps (not honored by every dump option). Also
6635 include information from the optimization passes.
6637 Enable dumping various statistics about the pass (not honored by every dump
6640 Enable showing basic block boundaries (disabled in raw dumps).
6642 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6643 dump a representation of the control flow graph suitable for viewing with
6644 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6645 the file is pretty-printed as a subgraph, so that GraphViz can render them
6646 all in a single plot.
6648 This option currently only works for RTL dumps, and the RTL is always
6649 dumped in slim form.
6651 Enable showing virtual operands for every statement.
6653 Enable showing line numbers for statements.
6655 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6657 Enable showing the tree dump for each statement.
6659 Enable showing the EH region number holding each statement.
6661 Enable showing scalar evolution analysis details.
6663 Enable showing optimization information (only available in certain
6666 Enable showing missed optimization information (only available in certain
6669 Enable other detailed optimization information (only available in
6671 @item =@var{filename}
6672 Instead of an auto named dump file, output into the given file
6673 name. The file names @file{stdout} and @file{stderr} are treated
6674 specially and are considered already open standard streams. For
6678 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6679 -fdump-tree-pre=stderr file.c
6682 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6683 output on to @file{stderr}. If two conflicting dump filenames are
6684 given for the same pass, then the latter option overrides the earlier
6688 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6689 and @option{lineno}.
6692 Turn on all optimization options, i.e., @option{optimized},
6693 @option{missed}, and @option{note}.
6696 The following tree dumps are possible:
6700 @opindex fdump-tree-original
6701 Dump before any tree based optimization, to @file{@var{file}.original}.
6704 @opindex fdump-tree-optimized
6705 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6708 @opindex fdump-tree-gimple
6709 Dump each function before and after the gimplification pass to a file. The
6710 file name is made by appending @file{.gimple} to the source file name.
6713 @opindex fdump-tree-cfg
6714 Dump the control flow graph of each function to a file. The file name is
6715 made by appending @file{.cfg} to the source file name.
6718 @opindex fdump-tree-ch
6719 Dump each function after copying loop headers. The file name is made by
6720 appending @file{.ch} to the source file name.
6723 @opindex fdump-tree-ssa
6724 Dump SSA related information to a file. The file name is made by appending
6725 @file{.ssa} to the source file name.
6728 @opindex fdump-tree-alias
6729 Dump aliasing information for each function. The file name is made by
6730 appending @file{.alias} to the source file name.
6733 @opindex fdump-tree-ccp
6734 Dump each function after CCP@. The file name is made by appending
6735 @file{.ccp} to the source file name.
6738 @opindex fdump-tree-storeccp
6739 Dump each function after STORE-CCP@. The file name is made by appending
6740 @file{.storeccp} to the source file name.
6743 @opindex fdump-tree-pre
6744 Dump trees after partial redundancy elimination. The file name is made
6745 by appending @file{.pre} to the source file name.
6748 @opindex fdump-tree-fre
6749 Dump trees after full redundancy elimination. The file name is made
6750 by appending @file{.fre} to the source file name.
6753 @opindex fdump-tree-copyprop
6754 Dump trees after copy propagation. The file name is made
6755 by appending @file{.copyprop} to the source file name.
6757 @item store_copyprop
6758 @opindex fdump-tree-store_copyprop
6759 Dump trees after store copy-propagation. The file name is made
6760 by appending @file{.store_copyprop} to the source file name.
6763 @opindex fdump-tree-dce
6764 Dump each function after dead code elimination. The file name is made by
6765 appending @file{.dce} to the source file name.
6768 @opindex fdump-tree-sra
6769 Dump each function after performing scalar replacement of aggregates. The
6770 file name is made by appending @file{.sra} to the source file name.
6773 @opindex fdump-tree-sink
6774 Dump each function after performing code sinking. The file name is made
6775 by appending @file{.sink} to the source file name.
6778 @opindex fdump-tree-dom
6779 Dump each function after applying dominator tree optimizations. The file
6780 name is made by appending @file{.dom} to the source file name.
6783 @opindex fdump-tree-dse
6784 Dump each function after applying dead store elimination. The file
6785 name is made by appending @file{.dse} to the source file name.
6788 @opindex fdump-tree-phiopt
6789 Dump each function after optimizing PHI nodes into straightline code. The file
6790 name is made by appending @file{.phiopt} to the source file name.
6793 @opindex fdump-tree-forwprop
6794 Dump each function after forward propagating single use variables. The file
6795 name is made by appending @file{.forwprop} to the source file name.
6798 @opindex fdump-tree-copyrename
6799 Dump each function after applying the copy rename optimization. The file
6800 name is made by appending @file{.copyrename} to the source file name.
6803 @opindex fdump-tree-nrv
6804 Dump each function after applying the named return value optimization on
6805 generic trees. The file name is made by appending @file{.nrv} to the source
6809 @opindex fdump-tree-vect
6810 Dump each function after applying vectorization of loops. The file name is
6811 made by appending @file{.vect} to the source file name.
6814 @opindex fdump-tree-slp
6815 Dump each function after applying vectorization of basic blocks. The file name
6816 is made by appending @file{.slp} to the source file name.
6819 @opindex fdump-tree-vrp
6820 Dump each function after Value Range Propagation (VRP). The file name
6821 is made by appending @file{.vrp} to the source file name.
6824 @opindex fdump-tree-all
6825 Enable all the available tree dumps with the flags provided in this option.
6829 @itemx -fopt-info-@var{options}
6830 @itemx -fopt-info-@var{options}=@var{filename}
6832 Controls optimization dumps from various optimization passes. If the
6833 @samp{-@var{options}} form is used, @var{options} is a list of
6834 @samp{-} separated option keywords to select the dump details and
6837 The @var{options} can be divided into two groups: options describing the
6838 verbosity of the dump, and options describing which optimizations
6839 should be included. The options from both the groups can be freely
6840 mixed as they are non-overlapping. However, in case of any conflicts,
6841 the later options override the earlier options on the command
6844 The following options control the dump verbosity:
6848 Print information when an optimization is successfully applied. It is
6849 up to a pass to decide which information is relevant. For example, the
6850 vectorizer passes print the source location of loops which are
6851 successfully vectorized.
6853 Print information about missed optimizations. Individual passes
6854 control which information to include in the output.
6856 Print verbose information about optimizations, such as certain
6857 transformations, more detailed messages about decisions etc.
6859 Print detailed optimization information. This includes
6860 @samp{optimized}, @samp{missed}, and @samp{note}.
6863 One or more of the following option keywords can be used to describe a
6864 group of optimizations:
6868 Enable dumps from all interprocedural optimizations.
6870 Enable dumps from all loop optimizations.
6872 Enable dumps from all inlining optimizations.
6874 Enable dumps from all vectorization optimizations.
6876 Enable dumps from all optimizations. This is a superset of
6877 the optimization groups listed above.
6881 omitted, it defaults to @samp{optimized-optall}, which means to dump all
6882 info about successful optimizations from all the passes.
6884 If the @var{filename} is provided, then the dumps from all the
6885 applicable optimizations are concatenated into the @var{filename}.
6886 Otherwise the dump is output onto @file{stderr}. Though multiple
6887 @option{-fopt-info} options are accepted, only one of them can include
6888 a @var{filename}. If other filenames are provided then all but the
6889 first such option are ignored.
6891 Note that the output @var{filename} is overwritten
6892 in case of multiple translation units. If a combined output from
6893 multiple translation units is desired, @file{stderr} should be used
6896 In the following example, the optimization info is output to
6905 gcc -O3 -fopt-info-missed=missed.all
6909 outputs missed optimization report from all the passes into
6910 @file{missed.all}, and this one:
6913 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
6917 prints information about missed optimization opportunities from
6918 vectorization passes on @file{stderr}.
6919 Note that @option{-fopt-info-vec-missed} is equivalent to
6920 @option{-fopt-info-missed-vec}.
6924 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
6928 outputs information about missed optimizations as well as
6929 optimized locations from all the inlining passes into
6935 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
6939 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
6940 in conflict since only one output file is allowed. In this case, only
6941 the first option takes effect and the subsequent options are
6942 ignored. Thus only @file{vec.miss} is produced which contains
6943 dumps from the vectorizer about missed opportunities.
6945 @item -frandom-seed=@var{number}
6946 @opindex frandom-seed
6947 This option provides a seed that GCC uses in place of
6948 random numbers in generating certain symbol names
6949 that have to be different in every compiled file. It is also used to
6950 place unique stamps in coverage data files and the object files that
6951 produce them. You can use the @option{-frandom-seed} option to produce
6952 reproducibly identical object files.
6954 The @var{number} should be different for every file you compile.
6956 @item -fsched-verbose=@var{n}
6957 @opindex fsched-verbose
6958 On targets that use instruction scheduling, this option controls the
6959 amount of debugging output the scheduler prints. This information is
6960 written to standard error, unless @option{-fdump-rtl-sched1} or
6961 @option{-fdump-rtl-sched2} is specified, in which case it is output
6962 to the usual dump listing file, @file{.sched1} or @file{.sched2}
6963 respectively. However for @var{n} greater than nine, the output is
6964 always printed to standard error.
6966 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
6967 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
6968 For @var{n} greater than one, it also output basic block probabilities,
6969 detailed ready list information and unit/insn info. For @var{n} greater
6970 than two, it includes RTL at abort point, control-flow and regions info.
6971 And for @var{n} over four, @option{-fsched-verbose} also includes
6975 @itemx -save-temps=cwd
6977 Store the usual ``temporary'' intermediate files permanently; place them
6978 in the current directory and name them based on the source file. Thus,
6979 compiling @file{foo.c} with @option{-c -save-temps} produces files
6980 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
6981 preprocessed @file{foo.i} output file even though the compiler now
6982 normally uses an integrated preprocessor.
6984 When used in combination with the @option{-x} command-line option,
6985 @option{-save-temps} is sensible enough to avoid over writing an
6986 input source file with the same extension as an intermediate file.
6987 The corresponding intermediate file may be obtained by renaming the
6988 source file before using @option{-save-temps}.
6990 If you invoke GCC in parallel, compiling several different source
6991 files that share a common base name in different subdirectories or the
6992 same source file compiled for multiple output destinations, it is
6993 likely that the different parallel compilers will interfere with each
6994 other, and overwrite the temporary files. For instance:
6997 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
6998 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7001 may result in @file{foo.i} and @file{foo.o} being written to
7002 simultaneously by both compilers.
7004 @item -save-temps=obj
7005 @opindex save-temps=obj
7006 Store the usual ``temporary'' intermediate files permanently. If the
7007 @option{-o} option is used, the temporary files are based on the
7008 object file. If the @option{-o} option is not used, the
7009 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7014 gcc -save-temps=obj -c foo.c
7015 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7016 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7020 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7021 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7022 @file{dir2/yfoobar.o}.
7024 @item -time@r{[}=@var{file}@r{]}
7026 Report the CPU time taken by each subprocess in the compilation
7027 sequence. For C source files, this is the compiler proper and assembler
7028 (plus the linker if linking is done).
7030 Without the specification of an output file, the output looks like this:
7037 The first number on each line is the ``user time'', that is time spent
7038 executing the program itself. The second number is ``system time'',
7039 time spent executing operating system routines on behalf of the program.
7040 Both numbers are in seconds.
7042 With the specification of an output file, the output is appended to the
7043 named file, and it looks like this:
7046 0.12 0.01 cc1 @var{options}
7047 0.00 0.01 as @var{options}
7050 The ``user time'' and the ``system time'' are moved before the program
7051 name, and the options passed to the program are displayed, so that one
7052 can later tell what file was being compiled, and with which options.
7054 @item -fvar-tracking
7055 @opindex fvar-tracking
7056 Run variable tracking pass. It computes where variables are stored at each
7057 position in code. Better debugging information is then generated
7058 (if the debugging information format supports this information).
7060 It is enabled by default when compiling with optimization (@option{-Os},
7061 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7062 the debug info format supports it.
7064 @item -fvar-tracking-assignments
7065 @opindex fvar-tracking-assignments
7066 @opindex fno-var-tracking-assignments
7067 Annotate assignments to user variables early in the compilation and
7068 attempt to carry the annotations over throughout the compilation all the
7069 way to the end, in an attempt to improve debug information while
7070 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7072 It can be enabled even if var-tracking is disabled, in which case
7073 annotations are created and maintained, but discarded at the end.
7075 @item -fvar-tracking-assignments-toggle
7076 @opindex fvar-tracking-assignments-toggle
7077 @opindex fno-var-tracking-assignments-toggle
7078 Toggle @option{-fvar-tracking-assignments}, in the same way that
7079 @option{-gtoggle} toggles @option{-g}.
7081 @item -print-file-name=@var{library}
7082 @opindex print-file-name
7083 Print the full absolute name of the library file @var{library} that
7084 would be used when linking---and don't do anything else. With this
7085 option, GCC does not compile or link anything; it just prints the
7088 @item -print-multi-directory
7089 @opindex print-multi-directory
7090 Print the directory name corresponding to the multilib selected by any
7091 other switches present in the command line. This directory is supposed
7092 to exist in @env{GCC_EXEC_PREFIX}.
7094 @item -print-multi-lib
7095 @opindex print-multi-lib
7096 Print the mapping from multilib directory names to compiler switches
7097 that enable them. The directory name is separated from the switches by
7098 @samp{;}, and each switch starts with an @samp{@@} instead of the
7099 @samp{-}, without spaces between multiple switches. This is supposed to
7100 ease shell processing.
7102 @item -print-multi-os-directory
7103 @opindex print-multi-os-directory
7104 Print the path to OS libraries for the selected
7105 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7106 present in the @file{lib} subdirectory and no multilibs are used, this is
7107 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7108 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7109 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7110 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7112 @item -print-multiarch
7113 @opindex print-multiarch
7114 Print the path to OS libraries for the selected multiarch,
7115 relative to some @file{lib} subdirectory.
7117 @item -print-prog-name=@var{program}
7118 @opindex print-prog-name
7119 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7121 @item -print-libgcc-file-name
7122 @opindex print-libgcc-file-name
7123 Same as @option{-print-file-name=libgcc.a}.
7125 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7126 but you do want to link with @file{libgcc.a}. You can do:
7129 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7132 @item -print-search-dirs
7133 @opindex print-search-dirs
7134 Print the name of the configured installation directory and a list of
7135 program and library directories @command{gcc} searches---and don't do anything else.
7137 This is useful when @command{gcc} prints the error message
7138 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7139 To resolve this you either need to put @file{cpp0} and the other compiler
7140 components where @command{gcc} expects to find them, or you can set the environment
7141 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7142 Don't forget the trailing @samp{/}.
7143 @xref{Environment Variables}.
7145 @item -print-sysroot
7146 @opindex print-sysroot
7147 Print the target sysroot directory that is used during
7148 compilation. This is the target sysroot specified either at configure
7149 time or using the @option{--sysroot} option, possibly with an extra
7150 suffix that depends on compilation options. If no target sysroot is
7151 specified, the option prints nothing.
7153 @item -print-sysroot-headers-suffix
7154 @opindex print-sysroot-headers-suffix
7155 Print the suffix added to the target sysroot when searching for
7156 headers, or give an error if the compiler is not configured with such
7157 a suffix---and don't do anything else.
7160 @opindex dumpmachine
7161 Print the compiler's target machine (for example,
7162 @samp{i686-pc-linux-gnu})---and don't do anything else.
7165 @opindex dumpversion
7166 Print the compiler version (for example, @code{3.0})---and don't do
7171 Print the compiler's built-in specs---and don't do anything else. (This
7172 is used when GCC itself is being built.) @xref{Spec Files}.
7174 @item -fno-eliminate-unused-debug-types
7175 @opindex feliminate-unused-debug-types
7176 @opindex fno-eliminate-unused-debug-types
7177 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7178 output for types that are nowhere used in the source file being compiled.
7179 Sometimes it is useful to have GCC emit debugging
7180 information for all types declared in a compilation
7181 unit, regardless of whether or not they are actually used
7182 in that compilation unit, for example
7183 if, in the debugger, you want to cast a value to a type that is
7184 not actually used in your program (but is declared). More often,
7185 however, this results in a significant amount of wasted space.
7188 @node Optimize Options
7189 @section Options That Control Optimization
7190 @cindex optimize options
7191 @cindex options, optimization
7193 These options control various sorts of optimizations.
7195 Without any optimization option, the compiler's goal is to reduce the
7196 cost of compilation and to make debugging produce the expected
7197 results. Statements are independent: if you stop the program with a
7198 breakpoint between statements, you can then assign a new value to any
7199 variable or change the program counter to any other statement in the
7200 function and get exactly the results you expect from the source
7203 Turning on optimization flags makes the compiler attempt to improve
7204 the performance and/or code size at the expense of compilation time
7205 and possibly the ability to debug the program.
7207 The compiler performs optimization based on the knowledge it has of the
7208 program. Compiling multiple files at once to a single output file mode allows
7209 the compiler to use information gained from all of the files when compiling
7212 Not all optimizations are controlled directly by a flag. Only
7213 optimizations that have a flag are listed in this section.
7215 Most optimizations are only enabled if an @option{-O} level is set on
7216 the command line. Otherwise they are disabled, even if individual
7217 optimization flags are specified.
7219 Depending on the target and how GCC was configured, a slightly different
7220 set of optimizations may be enabled at each @option{-O} level than
7221 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7222 to find out the exact set of optimizations that are enabled at each level.
7223 @xref{Overall Options}, for examples.
7230 Optimize. Optimizing compilation takes somewhat more time, and a lot
7231 more memory for a large function.
7233 With @option{-O}, the compiler tries to reduce code size and execution
7234 time, without performing any optimizations that take a great deal of
7237 @option{-O} turns on the following optimization flags:
7240 -fbranch-count-reg @gol
7241 -fcombine-stack-adjustments @gol
7243 -fcprop-registers @gol
7246 -fdelayed-branch @gol
7248 -fforward-propagate @gol
7249 -fguess-branch-probability @gol
7250 -fif-conversion2 @gol
7251 -fif-conversion @gol
7252 -finline-functions-called-once @gol
7253 -fipa-pure-const @gol
7255 -fipa-reference @gol
7256 -fmerge-constants @gol
7257 -fmove-loop-invariants @gol
7259 -fsplit-wide-types @gol
7264 -ftree-copy-prop @gol
7265 -ftree-copyrename @gol
7267 -ftree-dominator-opts @gol
7269 -ftree-forwprop @gol
7279 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7280 where doing so does not interfere with debugging.
7284 Optimize even more. GCC performs nearly all supported optimizations
7285 that do not involve a space-speed tradeoff.
7286 As compared to @option{-O}, this option increases both compilation time
7287 and the performance of the generated code.
7289 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7290 also turns on the following optimization flags:
7291 @gccoptlist{-fthread-jumps @gol
7292 -falign-functions -falign-jumps @gol
7293 -falign-loops -falign-labels @gol
7296 -fcse-follow-jumps -fcse-skip-blocks @gol
7297 -fdelete-null-pointer-checks @gol
7298 -fdevirtualize -fdevirtualize-speculatively @gol
7299 -fexpensive-optimizations @gol
7300 -fgcse -fgcse-lm @gol
7301 -fhoist-adjacent-loads @gol
7302 -finline-small-functions @gol
7303 -findirect-inlining @gol
7307 -fisolate-erroneous-paths-dereference @gol
7309 -foptimize-sibling-calls @gol
7310 -foptimize-strlen @gol
7311 -fpartial-inlining @gol
7313 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7314 -frerun-cse-after-loop @gol
7315 -fsched-interblock -fsched-spec @gol
7316 -fschedule-insns -fschedule-insns2 @gol
7317 -fstrict-aliasing -fstrict-overflow @gol
7318 -ftree-builtin-call-dce @gol
7319 -ftree-switch-conversion -ftree-tail-merge @gol
7324 Please note the warning under @option{-fgcse} about
7325 invoking @option{-O2} on programs that use computed gotos.
7329 Optimize yet more. @option{-O3} turns on all optimizations specified
7330 by @option{-O2} and also turns on the @option{-finline-functions},
7331 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7332 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7333 @option{-ftree-loop-distribute-patterns},
7334 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7335 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7339 Reduce compilation time and make debugging produce the expected
7340 results. This is the default.
7344 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7345 do not typically increase code size. It also performs further
7346 optimizations designed to reduce code size.
7348 @option{-Os} disables the following optimization flags:
7349 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7350 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7351 -fprefetch-loop-arrays}
7355 Disregard strict standards compliance. @option{-Ofast} enables all
7356 @option{-O3} optimizations. It also enables optimizations that are not
7357 valid for all standard-compliant programs.
7358 It turns on @option{-ffast-math} and the Fortran-specific
7359 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7363 Optimize debugging experience. @option{-Og} enables optimizations
7364 that do not interfere with debugging. It should be the optimization
7365 level of choice for the standard edit-compile-debug cycle, offering
7366 a reasonable level of optimization while maintaining fast compilation
7367 and a good debugging experience.
7369 If you use multiple @option{-O} options, with or without level numbers,
7370 the last such option is the one that is effective.
7373 Options of the form @option{-f@var{flag}} specify machine-independent
7374 flags. Most flags have both positive and negative forms; the negative
7375 form of @option{-ffoo} is @option{-fno-foo}. In the table
7376 below, only one of the forms is listed---the one you typically
7377 use. You can figure out the other form by either removing @samp{no-}
7380 The following options control specific optimizations. They are either
7381 activated by @option{-O} options or are related to ones that are. You
7382 can use the following flags in the rare cases when ``fine-tuning'' of
7383 optimizations to be performed is desired.
7386 @item -fno-defer-pop
7387 @opindex fno-defer-pop
7388 Always pop the arguments to each function call as soon as that function
7389 returns. For machines that must pop arguments after a function call,
7390 the compiler normally lets arguments accumulate on the stack for several
7391 function calls and pops them all at once.
7393 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7395 @item -fforward-propagate
7396 @opindex fforward-propagate
7397 Perform a forward propagation pass on RTL@. The pass tries to combine two
7398 instructions and checks if the result can be simplified. If loop unrolling
7399 is active, two passes are performed and the second is scheduled after
7402 This option is enabled by default at optimization levels @option{-O},
7403 @option{-O2}, @option{-O3}, @option{-Os}.
7405 @item -ffp-contract=@var{style}
7406 @opindex ffp-contract
7407 @option{-ffp-contract=off} disables floating-point expression contraction.
7408 @option{-ffp-contract=fast} enables floating-point expression contraction
7409 such as forming of fused multiply-add operations if the target has
7410 native support for them.
7411 @option{-ffp-contract=on} enables floating-point expression contraction
7412 if allowed by the language standard. This is currently not implemented
7413 and treated equal to @option{-ffp-contract=off}.
7415 The default is @option{-ffp-contract=fast}.
7417 @item -fomit-frame-pointer
7418 @opindex fomit-frame-pointer
7419 Don't keep the frame pointer in a register for functions that
7420 don't need one. This avoids the instructions to save, set up and
7421 restore frame pointers; it also makes an extra register available
7422 in many functions. @strong{It also makes debugging impossible on
7425 On some machines, such as the VAX, this flag has no effect, because
7426 the standard calling sequence automatically handles the frame pointer
7427 and nothing is saved by pretending it doesn't exist. The
7428 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7429 whether a target machine supports this flag. @xref{Registers,,Register
7430 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7432 Starting with GCC version 4.6, the default setting (when not optimizing for
7433 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
7434 @option{-fomit-frame-pointer}. The default can be reverted to
7435 @option{-fno-omit-frame-pointer} by configuring GCC with the
7436 @option{--enable-frame-pointer} configure option.
7438 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7440 @item -foptimize-sibling-calls
7441 @opindex foptimize-sibling-calls
7442 Optimize sibling and tail recursive calls.
7444 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7446 @item -foptimize-strlen
7447 @opindex foptimize-strlen
7448 Optimize various standard C string functions (e.g. @code{strlen},
7449 @code{strchr} or @code{strcpy}) and
7450 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7452 Enabled at levels @option{-O2}, @option{-O3}.
7456 Do not expand any functions inline apart from those marked with
7457 the @code{always_inline} attribute. This is the default when not
7460 Single functions can be exempted from inlining by marking them
7461 with the @code{noinline} attribute.
7463 @item -finline-small-functions
7464 @opindex finline-small-functions
7465 Integrate functions into their callers when their body is smaller than expected
7466 function call code (so overall size of program gets smaller). The compiler
7467 heuristically decides which functions are simple enough to be worth integrating
7468 in this way. This inlining applies to all functions, even those not declared
7471 Enabled at level @option{-O2}.
7473 @item -findirect-inlining
7474 @opindex findirect-inlining
7475 Inline also indirect calls that are discovered to be known at compile
7476 time thanks to previous inlining. This option has any effect only
7477 when inlining itself is turned on by the @option{-finline-functions}
7478 or @option{-finline-small-functions} options.
7480 Enabled at level @option{-O2}.
7482 @item -finline-functions
7483 @opindex finline-functions
7484 Consider all functions for inlining, even if they are not declared inline.
7485 The compiler heuristically decides which functions are worth integrating
7488 If all calls to a given function are integrated, and the function is
7489 declared @code{static}, then the function is normally not output as
7490 assembler code in its own right.
7492 Enabled at level @option{-O3}.
7494 @item -finline-functions-called-once
7495 @opindex finline-functions-called-once
7496 Consider all @code{static} functions called once for inlining into their
7497 caller even if they are not marked @code{inline}. If a call to a given
7498 function is integrated, then the function is not output as assembler code
7501 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7503 @item -fearly-inlining
7504 @opindex fearly-inlining
7505 Inline functions marked by @code{always_inline} and functions whose body seems
7506 smaller than the function call overhead early before doing
7507 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7508 makes profiling significantly cheaper and usually inlining faster on programs
7509 having large chains of nested wrapper functions.
7515 Perform interprocedural scalar replacement of aggregates, removal of
7516 unused parameters and replacement of parameters passed by reference
7517 by parameters passed by value.
7519 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7521 @item -finline-limit=@var{n}
7522 @opindex finline-limit
7523 By default, GCC limits the size of functions that can be inlined. This flag
7524 allows coarse control of this limit. @var{n} is the size of functions that
7525 can be inlined in number of pseudo instructions.
7527 Inlining is actually controlled by a number of parameters, which may be
7528 specified individually by using @option{--param @var{name}=@var{value}}.
7529 The @option{-finline-limit=@var{n}} option sets some of these parameters
7533 @item max-inline-insns-single
7534 is set to @var{n}/2.
7535 @item max-inline-insns-auto
7536 is set to @var{n}/2.
7539 See below for a documentation of the individual
7540 parameters controlling inlining and for the defaults of these parameters.
7542 @emph{Note:} there may be no value to @option{-finline-limit} that results
7543 in default behavior.
7545 @emph{Note:} pseudo instruction represents, in this particular context, an
7546 abstract measurement of function's size. In no way does it represent a count
7547 of assembly instructions and as such its exact meaning might change from one
7548 release to an another.
7550 @item -fno-keep-inline-dllexport
7551 @opindex fno-keep-inline-dllexport
7552 This is a more fine-grained version of @option{-fkeep-inline-functions},
7553 which applies only to functions that are declared using the @code{dllexport}
7554 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7557 @item -fkeep-inline-functions
7558 @opindex fkeep-inline-functions
7559 In C, emit @code{static} functions that are declared @code{inline}
7560 into the object file, even if the function has been inlined into all
7561 of its callers. This switch does not affect functions using the
7562 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7563 inline functions into the object file.
7565 @item -fkeep-static-consts
7566 @opindex fkeep-static-consts
7567 Emit variables declared @code{static const} when optimization isn't turned
7568 on, even if the variables aren't referenced.
7570 GCC enables this option by default. If you want to force the compiler to
7571 check if a variable is referenced, regardless of whether or not
7572 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7574 @item -fmerge-constants
7575 @opindex fmerge-constants
7576 Attempt to merge identical constants (string constants and floating-point
7577 constants) across compilation units.
7579 This option is the default for optimized compilation if the assembler and
7580 linker support it. Use @option{-fno-merge-constants} to inhibit this
7583 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7585 @item -fmerge-all-constants
7586 @opindex fmerge-all-constants
7587 Attempt to merge identical constants and identical variables.
7589 This option implies @option{-fmerge-constants}. In addition to
7590 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7591 arrays or initialized constant variables with integral or floating-point
7592 types. Languages like C or C++ require each variable, including multiple
7593 instances of the same variable in recursive calls, to have distinct locations,
7594 so using this option results in non-conforming
7597 @item -fmodulo-sched
7598 @opindex fmodulo-sched
7599 Perform swing modulo scheduling immediately before the first scheduling
7600 pass. This pass looks at innermost loops and reorders their
7601 instructions by overlapping different iterations.
7603 @item -fmodulo-sched-allow-regmoves
7604 @opindex fmodulo-sched-allow-regmoves
7605 Perform more aggressive SMS-based modulo scheduling with register moves
7606 allowed. By setting this flag certain anti-dependences edges are
7607 deleted, which triggers the generation of reg-moves based on the
7608 life-range analysis. This option is effective only with
7609 @option{-fmodulo-sched} enabled.
7611 @item -fno-branch-count-reg
7612 @opindex fno-branch-count-reg
7613 Do not use ``decrement and branch'' instructions on a count register,
7614 but instead generate a sequence of instructions that decrement a
7615 register, compare it against zero, then branch based upon the result.
7616 This option is only meaningful on architectures that support such
7617 instructions, which include x86, PowerPC, IA-64 and S/390.
7619 Enabled by default at @option{-O1} and higher.
7621 The default is @option{-fbranch-count-reg}.
7623 @item -fno-function-cse
7624 @opindex fno-function-cse
7625 Do not put function addresses in registers; make each instruction that
7626 calls a constant function contain the function's address explicitly.
7628 This option results in less efficient code, but some strange hacks
7629 that alter the assembler output may be confused by the optimizations
7630 performed when this option is not used.
7632 The default is @option{-ffunction-cse}
7634 @item -fno-zero-initialized-in-bss
7635 @opindex fno-zero-initialized-in-bss
7636 If the target supports a BSS section, GCC by default puts variables that
7637 are initialized to zero into BSS@. This can save space in the resulting
7640 This option turns off this behavior because some programs explicitly
7641 rely on variables going to the data section---e.g., so that the
7642 resulting executable can find the beginning of that section and/or make
7643 assumptions based on that.
7645 The default is @option{-fzero-initialized-in-bss}.
7647 @item -fthread-jumps
7648 @opindex fthread-jumps
7649 Perform optimizations that check to see if a jump branches to a
7650 location where another comparison subsumed by the first is found. If
7651 so, the first branch is redirected to either the destination of the
7652 second branch or a point immediately following it, depending on whether
7653 the condition is known to be true or false.
7655 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7657 @item -fsplit-wide-types
7658 @opindex fsplit-wide-types
7659 When using a type that occupies multiple registers, such as @code{long
7660 long} on a 32-bit system, split the registers apart and allocate them
7661 independently. This normally generates better code for those types,
7662 but may make debugging more difficult.
7664 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7667 @item -fcse-follow-jumps
7668 @opindex fcse-follow-jumps
7669 In common subexpression elimination (CSE), scan through jump instructions
7670 when the target of the jump is not reached by any other path. For
7671 example, when CSE encounters an @code{if} statement with an
7672 @code{else} clause, CSE follows the jump when the condition
7675 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7677 @item -fcse-skip-blocks
7678 @opindex fcse-skip-blocks
7679 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7680 follow jumps that conditionally skip over blocks. When CSE
7681 encounters a simple @code{if} statement with no else clause,
7682 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7683 body of the @code{if}.
7685 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7687 @item -frerun-cse-after-loop
7688 @opindex frerun-cse-after-loop
7689 Re-run common subexpression elimination after loop optimizations are
7692 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7696 Perform a global common subexpression elimination pass.
7697 This pass also performs global constant and copy propagation.
7699 @emph{Note:} When compiling a program using computed gotos, a GCC
7700 extension, you may get better run-time performance if you disable
7701 the global common subexpression elimination pass by adding
7702 @option{-fno-gcse} to the command line.
7704 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7708 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7709 attempts to move loads that are only killed by stores into themselves. This
7710 allows a loop containing a load/store sequence to be changed to a load outside
7711 the loop, and a copy/store within the loop.
7713 Enabled by default when @option{-fgcse} is enabled.
7717 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7718 global common subexpression elimination. This pass attempts to move
7719 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7720 loops containing a load/store sequence can be changed to a load before
7721 the loop and a store after the loop.
7723 Not enabled at any optimization level.
7727 When @option{-fgcse-las} is enabled, the global common subexpression
7728 elimination pass eliminates redundant loads that come after stores to the
7729 same memory location (both partial and full redundancies).
7731 Not enabled at any optimization level.
7733 @item -fgcse-after-reload
7734 @opindex fgcse-after-reload
7735 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7736 pass is performed after reload. The purpose of this pass is to clean up
7739 @item -faggressive-loop-optimizations
7740 @opindex faggressive-loop-optimizations
7741 This option tells the loop optimizer to use language constraints to
7742 derive bounds for the number of iterations of a loop. This assumes that
7743 loop code does not invoke undefined behavior by for example causing signed
7744 integer overflows or out-of-bound array accesses. The bounds for the
7745 number of iterations of a loop are used to guide loop unrolling and peeling
7746 and loop exit test optimizations.
7747 This option is enabled by default.
7749 @item -funsafe-loop-optimizations
7750 @opindex funsafe-loop-optimizations
7751 This option tells the loop optimizer to assume that loop indices do not
7752 overflow, and that loops with nontrivial exit condition are not
7753 infinite. This enables a wider range of loop optimizations even if
7754 the loop optimizer itself cannot prove that these assumptions are valid.
7755 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7756 if it finds this kind of loop.
7758 @item -fcrossjumping
7759 @opindex fcrossjumping
7760 Perform cross-jumping transformation.
7761 This transformation unifies equivalent code and saves code size. The
7762 resulting code may or may not perform better than without cross-jumping.
7764 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7766 @item -fauto-inc-dec
7767 @opindex fauto-inc-dec
7768 Combine increments or decrements of addresses with memory accesses.
7769 This pass is always skipped on architectures that do not have
7770 instructions to support this. Enabled by default at @option{-O} and
7771 higher on architectures that support this.
7775 Perform dead code elimination (DCE) on RTL@.
7776 Enabled by default at @option{-O} and higher.
7780 Perform dead store elimination (DSE) on RTL@.
7781 Enabled by default at @option{-O} and higher.
7783 @item -fif-conversion
7784 @opindex fif-conversion
7785 Attempt to transform conditional jumps into branch-less equivalents. This
7786 includes use of conditional moves, min, max, set flags and abs instructions, and
7787 some tricks doable by standard arithmetics. The use of conditional execution
7788 on chips where it is available is controlled by @option{-fif-conversion2}.
7790 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7792 @item -fif-conversion2
7793 @opindex fif-conversion2
7794 Use conditional execution (where available) to transform conditional jumps into
7795 branch-less equivalents.
7797 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7799 @item -fdeclone-ctor-dtor
7800 @opindex fdeclone-ctor-dtor
7801 The C++ ABI requires multiple entry points for constructors and
7802 destructors: one for a base subobject, one for a complete object, and
7803 one for a virtual destructor that calls operator delete afterwards.
7804 For a hierarchy with virtual bases, the base and complete variants are
7805 clones, which means two copies of the function. With this option, the
7806 base and complete variants are changed to be thunks that call a common
7809 Enabled by @option{-Os}.
7811 @item -fdelete-null-pointer-checks
7812 @opindex fdelete-null-pointer-checks
7813 Assume that programs cannot safely dereference null pointers, and that
7814 no code or data element resides there. This enables simple constant
7815 folding optimizations at all optimization levels. In addition, other
7816 optimization passes in GCC use this flag to control global dataflow
7817 analyses that eliminate useless checks for null pointers; these assume
7818 that if a pointer is checked after it has already been dereferenced,
7821 Note however that in some environments this assumption is not true.
7822 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7823 for programs that depend on that behavior.
7825 Some targets, especially embedded ones, disable this option at all levels.
7826 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7827 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7828 are enabled independently at different optimization levels.
7830 @item -fdevirtualize
7831 @opindex fdevirtualize
7832 Attempt to convert calls to virtual functions to direct calls. This
7833 is done both within a procedure and interprocedurally as part of
7834 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7835 propagation (@option{-fipa-cp}).
7836 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7838 @item -fdevirtualize-speculatively
7839 @opindex fdevirtualize-speculatively
7840 Attempt to convert calls to virtual functions to speculative direct calls.
7841 Based on the analysis of the type inheritance graph, determine for a given call
7842 the set of likely targets. If the set is small, preferably of size 1, change
7843 the call into a conditional deciding between direct and indirect calls. The
7844 speculative calls enable more optimizations, such as inlining. When they seem
7845 useless after further optimization, they are converted back into original form.
7847 @item -fdevirtualize-at-ltrans
7848 @opindex fdevirtualize-at-ltrans
7849 Stream extra information needed for aggressive devirtualization when running
7850 the link-time optimizer in local transformation mode.
7851 This option enables more devirtualization but
7852 significantly increases the size of streamed data. For this reason it is
7853 disabled by default.
7855 @item -fexpensive-optimizations
7856 @opindex fexpensive-optimizations
7857 Perform a number of minor optimizations that are relatively expensive.
7859 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7863 Attempt to remove redundant extension instructions. This is especially
7864 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7865 registers after writing to their lower 32-bit half.
7867 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7868 @option{-O3}, @option{-Os}.
7870 @item -flive-range-shrinkage
7871 @opindex flive-range-shrinkage
7872 Attempt to decrease register pressure through register live range
7873 shrinkage. This is helpful for fast processors with small or moderate
7876 @item -fira-algorithm=@var{algorithm}
7877 @opindex fira-algorithm
7878 Use the specified coloring algorithm for the integrated register
7879 allocator. The @var{algorithm} argument can be @samp{priority}, which
7880 specifies Chow's priority coloring, or @samp{CB}, which specifies
7881 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7882 for all architectures, but for those targets that do support it, it is
7883 the default because it generates better code.
7885 @item -fira-region=@var{region}
7886 @opindex fira-region
7887 Use specified regions for the integrated register allocator. The
7888 @var{region} argument should be one of the following:
7893 Use all loops as register allocation regions.
7894 This can give the best results for machines with a small and/or
7895 irregular register set.
7898 Use all loops except for loops with small register pressure
7899 as the regions. This value usually gives
7900 the best results in most cases and for most architectures,
7901 and is enabled by default when compiling with optimization for speed
7902 (@option{-O}, @option{-O2}, @dots{}).
7905 Use all functions as a single region.
7906 This typically results in the smallest code size, and is enabled by default for
7907 @option{-Os} or @option{-O0}.
7911 @item -fira-hoist-pressure
7912 @opindex fira-hoist-pressure
7913 Use IRA to evaluate register pressure in the code hoisting pass for
7914 decisions to hoist expressions. This option usually results in smaller
7915 code, but it can slow the compiler down.
7917 This option is enabled at level @option{-Os} for all targets.
7919 @item -fira-loop-pressure
7920 @opindex fira-loop-pressure
7921 Use IRA to evaluate register pressure in loops for decisions to move
7922 loop invariants. This option usually results in generation
7923 of faster and smaller code on machines with large register files (>= 32
7924 registers), but it can slow the compiler down.
7926 This option is enabled at level @option{-O3} for some targets.
7928 @item -fno-ira-share-save-slots
7929 @opindex fno-ira-share-save-slots
7930 Disable sharing of stack slots used for saving call-used hard
7931 registers living through a call. Each hard register gets a
7932 separate stack slot, and as a result function stack frames are
7935 @item -fno-ira-share-spill-slots
7936 @opindex fno-ira-share-spill-slots
7937 Disable sharing of stack slots allocated for pseudo-registers. Each
7938 pseudo-register that does not get a hard register gets a separate
7939 stack slot, and as a result function stack frames are larger.
7941 @item -fira-verbose=@var{n}
7942 @opindex fira-verbose
7943 Control the verbosity of the dump file for the integrated register allocator.
7944 The default value is 5. If the value @var{n} is greater or equal to 10,
7945 the dump output is sent to stderr using the same format as @var{n} minus 10.
7949 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7950 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7951 values if it is profitable.
7953 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7955 @item -fdelayed-branch
7956 @opindex fdelayed-branch
7957 If supported for the target machine, attempt to reorder instructions
7958 to exploit instruction slots available after delayed branch
7961 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7963 @item -fschedule-insns
7964 @opindex fschedule-insns
7965 If supported for the target machine, attempt to reorder instructions to
7966 eliminate execution stalls due to required data being unavailable. This
7967 helps machines that have slow floating point or memory load instructions
7968 by allowing other instructions to be issued until the result of the load
7969 or floating-point instruction is required.
7971 Enabled at levels @option{-O2}, @option{-O3}.
7973 @item -fschedule-insns2
7974 @opindex fschedule-insns2
7975 Similar to @option{-fschedule-insns}, but requests an additional pass of
7976 instruction scheduling after register allocation has been done. This is
7977 especially useful on machines with a relatively small number of
7978 registers and where memory load instructions take more than one cycle.
7980 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7982 @item -fno-sched-interblock
7983 @opindex fno-sched-interblock
7984 Don't schedule instructions across basic blocks. This is normally
7985 enabled by default when scheduling before register allocation, i.e.@:
7986 with @option{-fschedule-insns} or at @option{-O2} or higher.
7988 @item -fno-sched-spec
7989 @opindex fno-sched-spec
7990 Don't allow speculative motion of non-load instructions. This is normally
7991 enabled by default when scheduling before register allocation, i.e.@:
7992 with @option{-fschedule-insns} or at @option{-O2} or higher.
7994 @item -fsched-pressure
7995 @opindex fsched-pressure
7996 Enable register pressure sensitive insn scheduling before register
7997 allocation. This only makes sense when scheduling before register
7998 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7999 @option{-O2} or higher. Usage of this option can improve the
8000 generated code and decrease its size by preventing register pressure
8001 increase above the number of available hard registers and subsequent
8002 spills in register allocation.
8004 @item -fsched-spec-load
8005 @opindex fsched-spec-load
8006 Allow speculative motion of some load instructions. This only makes
8007 sense when scheduling before register allocation, i.e.@: with
8008 @option{-fschedule-insns} or at @option{-O2} or higher.
8010 @item -fsched-spec-load-dangerous
8011 @opindex fsched-spec-load-dangerous
8012 Allow speculative motion of more load instructions. This only makes
8013 sense when scheduling before register allocation, i.e.@: with
8014 @option{-fschedule-insns} or at @option{-O2} or higher.
8016 @item -fsched-stalled-insns
8017 @itemx -fsched-stalled-insns=@var{n}
8018 @opindex fsched-stalled-insns
8019 Define how many insns (if any) can be moved prematurely from the queue
8020 of stalled insns into the ready list during the second scheduling pass.
8021 @option{-fno-sched-stalled-insns} means that no insns are moved
8022 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8023 on how many queued insns can be moved prematurely.
8024 @option{-fsched-stalled-insns} without a value is equivalent to
8025 @option{-fsched-stalled-insns=1}.
8027 @item -fsched-stalled-insns-dep
8028 @itemx -fsched-stalled-insns-dep=@var{n}
8029 @opindex fsched-stalled-insns-dep
8030 Define how many insn groups (cycles) are examined for a dependency
8031 on a stalled insn that is a candidate for premature removal from the queue
8032 of stalled insns. This has an effect only during the second scheduling pass,
8033 and only if @option{-fsched-stalled-insns} is used.
8034 @option{-fno-sched-stalled-insns-dep} is equivalent to
8035 @option{-fsched-stalled-insns-dep=0}.
8036 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8037 @option{-fsched-stalled-insns-dep=1}.
8039 @item -fsched2-use-superblocks
8040 @opindex fsched2-use-superblocks
8041 When scheduling after register allocation, use superblock scheduling.
8042 This allows motion across basic block boundaries,
8043 resulting in faster schedules. This option is experimental, as not all machine
8044 descriptions used by GCC model the CPU closely enough to avoid unreliable
8045 results from the algorithm.
8047 This only makes sense when scheduling after register allocation, i.e.@: with
8048 @option{-fschedule-insns2} or at @option{-O2} or higher.
8050 @item -fsched-group-heuristic
8051 @opindex fsched-group-heuristic
8052 Enable the group heuristic in the scheduler. This heuristic favors
8053 the instruction that belongs to a schedule group. This is enabled
8054 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8055 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8057 @item -fsched-critical-path-heuristic
8058 @opindex fsched-critical-path-heuristic
8059 Enable the critical-path heuristic in the scheduler. This heuristic favors
8060 instructions on the critical path. This is enabled by default when
8061 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8062 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8064 @item -fsched-spec-insn-heuristic
8065 @opindex fsched-spec-insn-heuristic
8066 Enable the speculative instruction heuristic in the scheduler. This
8067 heuristic favors speculative instructions with greater dependency weakness.
8068 This is enabled by default when scheduling is enabled, i.e.@:
8069 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8070 or at @option{-O2} or higher.
8072 @item -fsched-rank-heuristic
8073 @opindex fsched-rank-heuristic
8074 Enable the rank heuristic in the scheduler. This heuristic favors
8075 the instruction belonging to a basic block with greater size or frequency.
8076 This is enabled by default when scheduling is enabled, i.e.@:
8077 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8078 at @option{-O2} or higher.
8080 @item -fsched-last-insn-heuristic
8081 @opindex fsched-last-insn-heuristic
8082 Enable the last-instruction heuristic in the scheduler. This heuristic
8083 favors the instruction that is less dependent on the last instruction
8084 scheduled. This is enabled by default when scheduling is enabled,
8085 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8086 at @option{-O2} or higher.
8088 @item -fsched-dep-count-heuristic
8089 @opindex fsched-dep-count-heuristic
8090 Enable the dependent-count heuristic in the scheduler. This heuristic
8091 favors the instruction that has more instructions depending on it.
8092 This is enabled by default when scheduling is enabled, i.e.@:
8093 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8094 at @option{-O2} or higher.
8096 @item -freschedule-modulo-scheduled-loops
8097 @opindex freschedule-modulo-scheduled-loops
8098 Modulo scheduling is performed before traditional scheduling. If a loop
8099 is modulo scheduled, later scheduling passes may change its schedule.
8100 Use this option to control that behavior.
8102 @item -fselective-scheduling
8103 @opindex fselective-scheduling
8104 Schedule instructions using selective scheduling algorithm. Selective
8105 scheduling runs instead of the first scheduler pass.
8107 @item -fselective-scheduling2
8108 @opindex fselective-scheduling2
8109 Schedule instructions using selective scheduling algorithm. Selective
8110 scheduling runs instead of the second scheduler pass.
8112 @item -fsel-sched-pipelining
8113 @opindex fsel-sched-pipelining
8114 Enable software pipelining of innermost loops during selective scheduling.
8115 This option has no effect unless one of @option{-fselective-scheduling} or
8116 @option{-fselective-scheduling2} is turned on.
8118 @item -fsel-sched-pipelining-outer-loops
8119 @opindex fsel-sched-pipelining-outer-loops
8120 When pipelining loops during selective scheduling, also pipeline outer loops.
8121 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8123 @item -fsemantic-interposition
8124 @opindex fsemantic-interposition
8125 Some object formats, like ELF, allow interposing of symbols by the
8127 This means that for symbols exported from the DSO, the compiler cannot perform
8128 interprocedural propagation, inlining and other optimizations in anticipation
8129 that the function or variable in question may change. While this feature is
8130 useful, for example, to rewrite memory allocation functions by a debugging
8131 implementation, it is expensive in the terms of code quality.
8132 With @option{-fno-semantic-interposition} the compiler assumes that
8133 if interposition happens for functions the overwriting function will have
8134 precisely the same semantics (and side effects).
8135 Similarly if interposition happens
8136 for variables, the constructor of the variable will be the same. The flag
8137 has no effect for functions explicitly declared inline
8138 (where it is never allowed for interposition to change semantics)
8139 and for symbols explicitly declared weak.
8142 @opindex fshrink-wrap
8143 Emit function prologues only before parts of the function that need it,
8144 rather than at the top of the function. This flag is enabled by default at
8145 @option{-O} and higher.
8147 @item -fcaller-saves
8148 @opindex fcaller-saves
8149 Enable allocation of values to registers that are clobbered by
8150 function calls, by emitting extra instructions to save and restore the
8151 registers around such calls. Such allocation is done only when it
8152 seems to result in better code.
8154 This option is always enabled by default on certain machines, usually
8155 those which have no call-preserved registers to use instead.
8157 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8159 @item -fcombine-stack-adjustments
8160 @opindex fcombine-stack-adjustments
8161 Tracks stack adjustments (pushes and pops) and stack memory references
8162 and then tries to find ways to combine them.
8164 Enabled by default at @option{-O1} and higher.
8168 Use caller save registers for allocation if those registers are not used by
8169 any called function. In that case it is not necessary to save and restore
8170 them around calls. This is only possible if called functions are part of
8171 same compilation unit as current function and they are compiled before it.
8173 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8175 @item -fconserve-stack
8176 @opindex fconserve-stack
8177 Attempt to minimize stack usage. The compiler attempts to use less
8178 stack space, even if that makes the program slower. This option
8179 implies setting the @option{large-stack-frame} parameter to 100
8180 and the @option{large-stack-frame-growth} parameter to 400.
8182 @item -ftree-reassoc
8183 @opindex ftree-reassoc
8184 Perform reassociation on trees. This flag is enabled by default
8185 at @option{-O} and higher.
8189 Perform partial redundancy elimination (PRE) on trees. This flag is
8190 enabled by default at @option{-O2} and @option{-O3}.
8192 @item -ftree-partial-pre
8193 @opindex ftree-partial-pre
8194 Make partial redundancy elimination (PRE) more aggressive. This flag is
8195 enabled by default at @option{-O3}.
8197 @item -ftree-forwprop
8198 @opindex ftree-forwprop
8199 Perform forward propagation on trees. This flag is enabled by default
8200 at @option{-O} and higher.
8204 Perform full redundancy elimination (FRE) on trees. The difference
8205 between FRE and PRE is that FRE only considers expressions
8206 that are computed on all paths leading to the redundant computation.
8207 This analysis is faster than PRE, though it exposes fewer redundancies.
8208 This flag is enabled by default at @option{-O} and higher.
8210 @item -ftree-phiprop
8211 @opindex ftree-phiprop
8212 Perform hoisting of loads from conditional pointers on trees. This
8213 pass is enabled by default at @option{-O} and higher.
8215 @item -fhoist-adjacent-loads
8216 @opindex fhoist-adjacent-loads
8217 Speculatively hoist loads from both branches of an if-then-else if the
8218 loads are from adjacent locations in the same structure and the target
8219 architecture has a conditional move instruction. This flag is enabled
8220 by default at @option{-O2} and higher.
8222 @item -ftree-copy-prop
8223 @opindex ftree-copy-prop
8224 Perform copy propagation on trees. This pass eliminates unnecessary
8225 copy operations. This flag is enabled by default at @option{-O} and
8228 @item -fipa-pure-const
8229 @opindex fipa-pure-const
8230 Discover which functions are pure or constant.
8231 Enabled by default at @option{-O} and higher.
8233 @item -fipa-reference
8234 @opindex fipa-reference
8235 Discover which static variables do not escape the
8237 Enabled by default at @option{-O} and higher.
8241 Perform interprocedural pointer analysis and interprocedural modification
8242 and reference analysis. This option can cause excessive memory and
8243 compile-time usage on large compilation units. It is not enabled by
8244 default at any optimization level.
8247 @opindex fipa-profile
8248 Perform interprocedural profile propagation. The functions called only from
8249 cold functions are marked as cold. Also functions executed once (such as
8250 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8251 functions and loop less parts of functions executed once are then optimized for
8253 Enabled by default at @option{-O} and higher.
8257 Perform interprocedural constant propagation.
8258 This optimization analyzes the program to determine when values passed
8259 to functions are constants and then optimizes accordingly.
8260 This optimization can substantially increase performance
8261 if the application has constants passed to functions.
8262 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8264 @item -fipa-cp-clone
8265 @opindex fipa-cp-clone
8266 Perform function cloning to make interprocedural constant propagation stronger.
8267 When enabled, interprocedural constant propagation performs function cloning
8268 when externally visible function can be called with constant arguments.
8269 Because this optimization can create multiple copies of functions,
8270 it may significantly increase code size
8271 (see @option{--param ipcp-unit-growth=@var{value}}).
8272 This flag is enabled by default at @option{-O3}.
8276 Perform Identical Code Folding for functions and read-only variables.
8277 The optimization reduces code size and may disturb unwind stacks by replacing
8278 a function by equivalent one with a different name. The optimization works
8279 more effectively with link time optimization enabled.
8281 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8282 works on different levels and thus the optimizations are not same - there are
8283 equivalences that are found only by GCC and equivalences found only by Gold.
8285 This flag is enabled by default at @option{-O2} and @option{-Os}.
8287 @item -fisolate-erroneous-paths-dereference
8288 @opindex fisolate-erroneous-paths-dereference
8289 Detect paths that trigger erroneous or undefined behavior due to
8290 dereferencing a null pointer. Isolate those paths from the main control
8291 flow and turn the statement with erroneous or undefined behavior into a trap.
8292 This flag is enabled by default at @option{-O2} and higher.
8294 @item -fisolate-erroneous-paths-attribute
8295 @opindex fisolate-erroneous-paths-attribute
8296 Detect paths that trigger erroneous or undefined behavior due a null value
8297 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8298 attribute. Isolate those paths from the main control flow and turn the
8299 statement with erroneous or undefined behavior into a trap. This is not
8300 currently enabled, but may be enabled by @option{-O2} in the future.
8304 Perform forward store motion on trees. This flag is
8305 enabled by default at @option{-O} and higher.
8307 @item -ftree-bit-ccp
8308 @opindex ftree-bit-ccp
8309 Perform sparse conditional bit constant propagation on trees and propagate
8310 pointer alignment information.
8311 This pass only operates on local scalar variables and is enabled by default
8312 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8316 Perform sparse conditional constant propagation (CCP) on trees. This
8317 pass only operates on local scalar variables and is enabled by default
8318 at @option{-O} and higher.
8321 @opindex fssa-phiopt
8322 Perform pattern matching on SSA PHI nodes to optimize conditional
8323 code. This pass is enabled by default at @option{-O} and higher.
8325 @item -ftree-switch-conversion
8326 @opindex ftree-switch-conversion
8327 Perform conversion of simple initializations in a switch to
8328 initializations from a scalar array. This flag is enabled by default
8329 at @option{-O2} and higher.
8331 @item -ftree-tail-merge
8332 @opindex ftree-tail-merge
8333 Look for identical code sequences. When found, replace one with a jump to the
8334 other. This optimization is known as tail merging or cross jumping. This flag
8335 is enabled by default at @option{-O2} and higher. The compilation time
8337 be limited using @option{max-tail-merge-comparisons} parameter and
8338 @option{max-tail-merge-iterations} parameter.
8342 Perform dead code elimination (DCE) on trees. This flag is enabled by
8343 default at @option{-O} and higher.
8345 @item -ftree-builtin-call-dce
8346 @opindex ftree-builtin-call-dce
8347 Perform conditional dead code elimination (DCE) for calls to built-in functions
8348 that may set @code{errno} but are otherwise side-effect free. This flag is
8349 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8352 @item -ftree-dominator-opts
8353 @opindex ftree-dominator-opts
8354 Perform a variety of simple scalar cleanups (constant/copy
8355 propagation, redundancy elimination, range propagation and expression
8356 simplification) based on a dominator tree traversal. This also
8357 performs jump threading (to reduce jumps to jumps). This flag is
8358 enabled by default at @option{-O} and higher.
8362 Perform dead store elimination (DSE) on trees. A dead store is a store into
8363 a memory location that is later overwritten by another store without
8364 any intervening loads. In this case the earlier store can be deleted. This
8365 flag is enabled by default at @option{-O} and higher.
8369 Perform loop header copying on trees. This is beneficial since it increases
8370 effectiveness of code motion optimizations. It also saves one jump. This flag
8371 is enabled by default at @option{-O} and higher. It is not enabled
8372 for @option{-Os}, since it usually increases code size.
8374 @item -ftree-loop-optimize
8375 @opindex ftree-loop-optimize
8376 Perform loop optimizations on trees. This flag is enabled by default
8377 at @option{-O} and higher.
8379 @item -ftree-loop-linear
8380 @opindex ftree-loop-linear
8381 Perform loop interchange transformations on tree. Same as
8382 @option{-floop-interchange}. To use this code transformation, GCC has
8383 to be configured with @option{--with-isl} to enable the Graphite loop
8384 transformation infrastructure.
8386 @item -floop-interchange
8387 @opindex floop-interchange
8388 Perform loop interchange transformations on loops. Interchanging two
8389 nested loops switches the inner and outer loops. For example, given a
8394 A(J, I) = A(J, I) * C
8399 loop interchange transforms the loop as if it were written:
8403 A(J, I) = A(J, I) * C
8407 which can be beneficial when @code{N} is larger than the caches,
8408 because in Fortran, the elements of an array are stored in memory
8409 contiguously by column, and the original loop iterates over rows,
8410 potentially creating at each access a cache miss. This optimization
8411 applies to all the languages supported by GCC and is not limited to
8412 Fortran. To use this code transformation, GCC has to be configured
8413 with @option{--with-isl} to enable the Graphite loop transformation
8416 @item -floop-strip-mine
8417 @opindex floop-strip-mine
8418 Perform loop strip mining transformations on loops. Strip mining
8419 splits a loop into two nested loops. The outer loop has strides
8420 equal to the strip size and the inner loop has strides of the
8421 original loop within a strip. The strip length can be changed
8422 using the @option{loop-block-tile-size} parameter. For example,
8430 loop strip mining transforms the loop as if it were written:
8433 DO I = II, min (II + 50, N)
8438 This optimization applies to all the languages supported by GCC and is
8439 not limited to Fortran. To use this code transformation, GCC has to
8440 be configured with @option{--with-isl} to enable the Graphite loop
8441 transformation infrastructure.
8444 @opindex floop-block
8445 Perform loop blocking transformations on loops. Blocking strip mines
8446 each loop in the loop nest such that the memory accesses of the
8447 element loops fit inside caches. The strip length can be changed
8448 using the @option{loop-block-tile-size} parameter. For example, given
8453 A(J, I) = B(I) + C(J)
8458 loop blocking transforms the loop as if it were written:
8462 DO I = II, min (II + 50, N)
8463 DO J = JJ, min (JJ + 50, M)
8464 A(J, I) = B(I) + C(J)
8470 which can be beneficial when @code{M} is larger than the caches,
8471 because the innermost loop iterates over a smaller amount of data
8472 which can be kept in the caches. This optimization applies to all the
8473 languages supported by GCC and is not limited to Fortran. To use this
8474 code transformation, GCC has to be configured with @option{--with-isl}
8475 to enable the Graphite loop transformation infrastructure.
8477 @item -fgraphite-identity
8478 @opindex fgraphite-identity
8479 Enable the identity transformation for graphite. For every SCoP we generate
8480 the polyhedral representation and transform it back to gimple. Using
8481 @option{-fgraphite-identity} we can check the costs or benefits of the
8482 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8483 are also performed by the code generator ISL, like index splitting and
8484 dead code elimination in loops.
8486 @item -floop-nest-optimize
8487 @opindex floop-nest-optimize
8488 Enable the ISL based loop nest optimizer. This is a generic loop nest
8489 optimizer based on the Pluto optimization algorithms. It calculates a loop
8490 structure optimized for data-locality and parallelism. This option
8493 @item -floop-unroll-and-jam
8494 @opindex floop-unroll-and-jam
8495 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8496 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8497 The unrolled dimension (counting from the most inner one) can be changed
8498 using the @option{loop-unroll-jam-depth} parameter. .
8500 @item -floop-parallelize-all
8501 @opindex floop-parallelize-all
8502 Use the Graphite data dependence analysis to identify loops that can
8503 be parallelized. Parallelize all the loops that can be analyzed to
8504 not contain loop carried dependences without checking that it is
8505 profitable to parallelize the loops.
8507 @item -fcheck-data-deps
8508 @opindex fcheck-data-deps
8509 Compare the results of several data dependence analyzers. This option
8510 is used for debugging the data dependence analyzers.
8512 @item -ftree-loop-if-convert
8513 @opindex ftree-loop-if-convert
8514 Attempt to transform conditional jumps in the innermost loops to
8515 branch-less equivalents. The intent is to remove control-flow from
8516 the innermost loops in order to improve the ability of the
8517 vectorization pass to handle these loops. This is enabled by default
8518 if vectorization is enabled.
8520 @item -ftree-loop-if-convert-stores
8521 @opindex ftree-loop-if-convert-stores
8522 Attempt to also if-convert conditional jumps containing memory writes.
8523 This transformation can be unsafe for multi-threaded programs as it
8524 transforms conditional memory writes into unconditional memory writes.
8527 for (i = 0; i < N; i++)
8533 for (i = 0; i < N; i++)
8534 A[i] = cond ? expr : A[i];
8536 potentially producing data races.
8538 @item -ftree-loop-distribution
8539 @opindex ftree-loop-distribution
8540 Perform loop distribution. This flag can improve cache performance on
8541 big loop bodies and allow further loop optimizations, like
8542 parallelization or vectorization, to take place. For example, the loop
8559 @item -ftree-loop-distribute-patterns
8560 @opindex ftree-loop-distribute-patterns
8561 Perform loop distribution of patterns that can be code generated with
8562 calls to a library. This flag is enabled by default at @option{-O3}.
8564 This pass distributes the initialization loops and generates a call to
8565 memset zero. For example, the loop
8581 and the initialization loop is transformed into a call to memset zero.
8583 @item -ftree-loop-im
8584 @opindex ftree-loop-im
8585 Perform loop invariant motion on trees. This pass moves only invariants that
8586 are hard to handle at RTL level (function calls, operations that expand to
8587 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8588 operands of conditions that are invariant out of the loop, so that we can use
8589 just trivial invariantness analysis in loop unswitching. The pass also includes
8592 @item -ftree-loop-ivcanon
8593 @opindex ftree-loop-ivcanon
8594 Create a canonical counter for number of iterations in loops for which
8595 determining number of iterations requires complicated analysis. Later
8596 optimizations then may determine the number easily. Useful especially
8597 in connection with unrolling.
8601 Perform induction variable optimizations (strength reduction, induction
8602 variable merging and induction variable elimination) on trees.
8604 @item -ftree-parallelize-loops=n
8605 @opindex ftree-parallelize-loops
8606 Parallelize loops, i.e., split their iteration space to run in n threads.
8607 This is only possible for loops whose iterations are independent
8608 and can be arbitrarily reordered. The optimization is only
8609 profitable on multiprocessor machines, for loops that are CPU-intensive,
8610 rather than constrained e.g.@: by memory bandwidth. This option
8611 implies @option{-pthread}, and thus is only supported on targets
8612 that have support for @option{-pthread}.
8616 Perform function-local points-to analysis on trees. This flag is
8617 enabled by default at @option{-O} and higher.
8621 Perform scalar replacement of aggregates. This pass replaces structure
8622 references with scalars to prevent committing structures to memory too
8623 early. This flag is enabled by default at @option{-O} and higher.
8625 @item -ftree-copyrename
8626 @opindex ftree-copyrename
8627 Perform copy renaming on trees. This pass attempts to rename compiler
8628 temporaries to other variables at copy locations, usually resulting in
8629 variable names which more closely resemble the original variables. This flag
8630 is enabled by default at @option{-O} and higher.
8632 @item -ftree-coalesce-inlined-vars
8633 @opindex ftree-coalesce-inlined-vars
8634 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8635 combine small user-defined variables too, but only if they are inlined
8636 from other functions. It is a more limited form of
8637 @option{-ftree-coalesce-vars}. This may harm debug information of such
8638 inlined variables, but it keeps variables of the inlined-into
8639 function apart from each other, such that they are more likely to
8640 contain the expected values in a debugging session. This was the
8641 default in GCC versions older than 4.7.
8643 @item -ftree-coalesce-vars
8644 @opindex ftree-coalesce-vars
8645 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8646 combine small user-defined variables too, instead of just compiler
8647 temporaries. This may severely limit the ability to debug an optimized
8648 program compiled with @option{-fno-var-tracking-assignments}. In the
8649 negated form, this flag prevents SSA coalescing of user variables,
8650 including inlined ones. This option is enabled by default.
8654 Perform temporary expression replacement during the SSA->normal phase. Single
8655 use/single def temporaries are replaced at their use location with their
8656 defining expression. This results in non-GIMPLE code, but gives the expanders
8657 much more complex trees to work on resulting in better RTL generation. This is
8658 enabled by default at @option{-O} and higher.
8662 Perform straight-line strength reduction on trees. This recognizes related
8663 expressions involving multiplications and replaces them by less expensive
8664 calculations when possible. This is enabled by default at @option{-O} and
8667 @item -ftree-vectorize
8668 @opindex ftree-vectorize
8669 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8670 and @option{-ftree-slp-vectorize} if not explicitly specified.
8672 @item -ftree-loop-vectorize
8673 @opindex ftree-loop-vectorize
8674 Perform loop vectorization on trees. This flag is enabled by default at
8675 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8677 @item -ftree-slp-vectorize
8678 @opindex ftree-slp-vectorize
8679 Perform basic block vectorization on trees. This flag is enabled by default at
8680 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8682 @item -fvect-cost-model=@var{model}
8683 @opindex fvect-cost-model
8684 Alter the cost model used for vectorization. The @var{model} argument
8685 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8686 With the @samp{unlimited} model the vectorized code-path is assumed
8687 to be profitable while with the @samp{dynamic} model a runtime check
8688 guards the vectorized code-path to enable it only for iteration
8689 counts that will likely execute faster than when executing the original
8690 scalar loop. The @samp{cheap} model disables vectorization of
8691 loops where doing so would be cost prohibitive for example due to
8692 required runtime checks for data dependence or alignment but otherwise
8693 is equal to the @samp{dynamic} model.
8694 The default cost model depends on other optimization flags and is
8695 either @samp{dynamic} or @samp{cheap}.
8697 @item -fsimd-cost-model=@var{model}
8698 @opindex fsimd-cost-model
8699 Alter the cost model used for vectorization of loops marked with the OpenMP
8700 or Cilk Plus simd directive. The @var{model} argument should be one of
8701 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8702 have the same meaning as described in @option{-fvect-cost-model} and by
8703 default a cost model defined with @option{-fvect-cost-model} is used.
8707 Perform Value Range Propagation on trees. This is similar to the
8708 constant propagation pass, but instead of values, ranges of values are
8709 propagated. This allows the optimizers to remove unnecessary range
8710 checks like array bound checks and null pointer checks. This is
8711 enabled by default at @option{-O2} and higher. Null pointer check
8712 elimination is only done if @option{-fdelete-null-pointer-checks} is
8717 Perform tail duplication to enlarge superblock size. This transformation
8718 simplifies the control flow of the function allowing other optimizations to do
8721 @item -funroll-loops
8722 @opindex funroll-loops
8723 Unroll loops whose number of iterations can be determined at compile
8724 time or upon entry to the loop. @option{-funroll-loops} implies
8725 @option{-frerun-cse-after-loop}. This option makes code larger,
8726 and may or may not make it run faster.
8728 @item -funroll-all-loops
8729 @opindex funroll-all-loops
8730 Unroll all loops, even if their number of iterations is uncertain when
8731 the loop is entered. This usually makes programs run more slowly.
8732 @option{-funroll-all-loops} implies the same options as
8733 @option{-funroll-loops},
8735 @item -fsplit-ivs-in-unroller
8736 @opindex fsplit-ivs-in-unroller
8737 Enables expression of values of induction variables in later iterations
8738 of the unrolled loop using the value in the first iteration. This breaks
8739 long dependency chains, thus improving efficiency of the scheduling passes.
8741 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8742 same effect. However, that is not reliable in cases where the loop body
8743 is more complicated than a single basic block. It also does not work at all
8744 on some architectures due to restrictions in the CSE pass.
8746 This optimization is enabled by default.
8748 @item -fvariable-expansion-in-unroller
8749 @opindex fvariable-expansion-in-unroller
8750 With this option, the compiler creates multiple copies of some
8751 local variables when unrolling a loop, which can result in superior code.
8753 @item -fpartial-inlining
8754 @opindex fpartial-inlining
8755 Inline parts of functions. This option has any effect only
8756 when inlining itself is turned on by the @option{-finline-functions}
8757 or @option{-finline-small-functions} options.
8759 Enabled at level @option{-O2}.
8761 @item -fpredictive-commoning
8762 @opindex fpredictive-commoning
8763 Perform predictive commoning optimization, i.e., reusing computations
8764 (especially memory loads and stores) performed in previous
8765 iterations of loops.
8767 This option is enabled at level @option{-O3}.
8769 @item -fprefetch-loop-arrays
8770 @opindex fprefetch-loop-arrays
8771 If supported by the target machine, generate instructions to prefetch
8772 memory to improve the performance of loops that access large arrays.
8774 This option may generate better or worse code; results are highly
8775 dependent on the structure of loops within the source code.
8777 Disabled at level @option{-Os}.
8780 @itemx -fno-peephole2
8781 @opindex fno-peephole
8782 @opindex fno-peephole2
8783 Disable any machine-specific peephole optimizations. The difference
8784 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8785 are implemented in the compiler; some targets use one, some use the
8786 other, a few use both.
8788 @option{-fpeephole} is enabled by default.
8789 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8791 @item -fno-guess-branch-probability
8792 @opindex fno-guess-branch-probability
8793 Do not guess branch probabilities using heuristics.
8795 GCC uses heuristics to guess branch probabilities if they are
8796 not provided by profiling feedback (@option{-fprofile-arcs}). These
8797 heuristics are based on the control flow graph. If some branch probabilities
8798 are specified by @code{__builtin_expect}, then the heuristics are
8799 used to guess branch probabilities for the rest of the control flow graph,
8800 taking the @code{__builtin_expect} info into account. The interactions
8801 between the heuristics and @code{__builtin_expect} can be complex, and in
8802 some cases, it may be useful to disable the heuristics so that the effects
8803 of @code{__builtin_expect} are easier to understand.
8805 The default is @option{-fguess-branch-probability} at levels
8806 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8808 @item -freorder-blocks
8809 @opindex freorder-blocks
8810 Reorder basic blocks in the compiled function in order to reduce number of
8811 taken branches and improve code locality.
8813 Enabled at levels @option{-O2}, @option{-O3}.
8815 @item -freorder-blocks-and-partition
8816 @opindex freorder-blocks-and-partition
8817 In addition to reordering basic blocks in the compiled function, in order
8818 to reduce number of taken branches, partitions hot and cold basic blocks
8819 into separate sections of the assembly and .o files, to improve
8820 paging and cache locality performance.
8822 This optimization is automatically turned off in the presence of
8823 exception handling, for linkonce sections, for functions with a user-defined
8824 section attribute and on any architecture that does not support named
8827 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8829 @item -freorder-functions
8830 @opindex freorder-functions
8831 Reorder functions in the object file in order to
8832 improve code locality. This is implemented by using special
8833 subsections @code{.text.hot} for most frequently executed functions and
8834 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8835 the linker so object file format must support named sections and linker must
8836 place them in a reasonable way.
8838 Also profile feedback must be available to make this option effective. See
8839 @option{-fprofile-arcs} for details.
8841 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8843 @item -fstrict-aliasing
8844 @opindex fstrict-aliasing
8845 Allow the compiler to assume the strictest aliasing rules applicable to
8846 the language being compiled. For C (and C++), this activates
8847 optimizations based on the type of expressions. In particular, an
8848 object of one type is assumed never to reside at the same address as an
8849 object of a different type, unless the types are almost the same. For
8850 example, an @code{unsigned int} can alias an @code{int}, but not a
8851 @code{void*} or a @code{double}. A character type may alias any other
8854 @anchor{Type-punning}Pay special attention to code like this:
8867 The practice of reading from a different union member than the one most
8868 recently written to (called ``type-punning'') is common. Even with
8869 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8870 is accessed through the union type. So, the code above works as
8871 expected. @xref{Structures unions enumerations and bit-fields
8872 implementation}. However, this code might not:
8883 Similarly, access by taking the address, casting the resulting pointer
8884 and dereferencing the result has undefined behavior, even if the cast
8885 uses a union type, e.g.:
8889 return ((union a_union *) &d)->i;
8893 The @option{-fstrict-aliasing} option is enabled at levels
8894 @option{-O2}, @option{-O3}, @option{-Os}.
8896 @item -fstrict-overflow
8897 @opindex fstrict-overflow
8898 Allow the compiler to assume strict signed overflow rules, depending
8899 on the language being compiled. For C (and C++) this means that
8900 overflow when doing arithmetic with signed numbers is undefined, which
8901 means that the compiler may assume that it does not happen. This
8902 permits various optimizations. For example, the compiler assumes
8903 that an expression like @code{i + 10 > i} is always true for
8904 signed @code{i}. This assumption is only valid if signed overflow is
8905 undefined, as the expression is false if @code{i + 10} overflows when
8906 using twos complement arithmetic. When this option is in effect any
8907 attempt to determine whether an operation on signed numbers
8908 overflows must be written carefully to not actually involve overflow.
8910 This option also allows the compiler to assume strict pointer
8911 semantics: given a pointer to an object, if adding an offset to that
8912 pointer does not produce a pointer to the same object, the addition is
8913 undefined. This permits the compiler to conclude that @code{p + u >
8914 p} is always true for a pointer @code{p} and unsigned integer
8915 @code{u}. This assumption is only valid because pointer wraparound is
8916 undefined, as the expression is false if @code{p + u} overflows using
8917 twos complement arithmetic.
8919 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8920 that integer signed overflow is fully defined: it wraps. When
8921 @option{-fwrapv} is used, there is no difference between
8922 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8923 integers. With @option{-fwrapv} certain types of overflow are
8924 permitted. For example, if the compiler gets an overflow when doing
8925 arithmetic on constants, the overflowed value can still be used with
8926 @option{-fwrapv}, but not otherwise.
8928 The @option{-fstrict-overflow} option is enabled at levels
8929 @option{-O2}, @option{-O3}, @option{-Os}.
8931 @item -falign-functions
8932 @itemx -falign-functions=@var{n}
8933 @opindex falign-functions
8934 Align the start of functions to the next power-of-two greater than
8935 @var{n}, skipping up to @var{n} bytes. For instance,
8936 @option{-falign-functions=32} aligns functions to the next 32-byte
8937 boundary, but @option{-falign-functions=24} aligns to the next
8938 32-byte boundary only if this can be done by skipping 23 bytes or less.
8940 @option{-fno-align-functions} and @option{-falign-functions=1} are
8941 equivalent and mean that functions are not aligned.
8943 Some assemblers only support this flag when @var{n} is a power of two;
8944 in that case, it is rounded up.
8946 If @var{n} is not specified or is zero, use a machine-dependent default.
8948 Enabled at levels @option{-O2}, @option{-O3}.
8950 @item -falign-labels
8951 @itemx -falign-labels=@var{n}
8952 @opindex falign-labels
8953 Align all branch targets to a power-of-two boundary, skipping up to
8954 @var{n} bytes like @option{-falign-functions}. This option can easily
8955 make code slower, because it must insert dummy operations for when the
8956 branch target is reached in the usual flow of the code.
8958 @option{-fno-align-labels} and @option{-falign-labels=1} are
8959 equivalent and mean that labels are not aligned.
8961 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8962 are greater than this value, then their values are used instead.
8964 If @var{n} is not specified or is zero, use a machine-dependent default
8965 which is very likely to be @samp{1}, meaning no alignment.
8967 Enabled at levels @option{-O2}, @option{-O3}.
8970 @itemx -falign-loops=@var{n}
8971 @opindex falign-loops
8972 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8973 like @option{-falign-functions}. If the loops are
8974 executed many times, this makes up for any execution of the dummy
8977 @option{-fno-align-loops} and @option{-falign-loops=1} are
8978 equivalent and mean that loops are not aligned.
8980 If @var{n} is not specified or is zero, use a machine-dependent default.
8982 Enabled at levels @option{-O2}, @option{-O3}.
8985 @itemx -falign-jumps=@var{n}
8986 @opindex falign-jumps
8987 Align branch targets to a power-of-two boundary, for branch targets
8988 where the targets can only be reached by jumping, skipping up to @var{n}
8989 bytes like @option{-falign-functions}. In this case, no dummy operations
8992 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
8993 equivalent and mean that loops are not aligned.
8995 If @var{n} is not specified or is zero, use a machine-dependent default.
8997 Enabled at levels @option{-O2}, @option{-O3}.
8999 @item -funit-at-a-time
9000 @opindex funit-at-a-time
9001 This option is left for compatibility reasons. @option{-funit-at-a-time}
9002 has no effect, while @option{-fno-unit-at-a-time} implies
9003 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9007 @item -fno-toplevel-reorder
9008 @opindex fno-toplevel-reorder
9009 Do not reorder top-level functions, variables, and @code{asm}
9010 statements. Output them in the same order that they appear in the
9011 input file. When this option is used, unreferenced static variables
9012 are not removed. This option is intended to support existing code
9013 that relies on a particular ordering. For new code, it is better to
9014 use attributes when possible.
9016 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9017 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9022 Constructs webs as commonly used for register allocation purposes and assign
9023 each web individual pseudo register. This allows the register allocation pass
9024 to operate on pseudos directly, but also strengthens several other optimization
9025 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9026 however, make debugging impossible, since variables no longer stay in a
9029 Enabled by default with @option{-funroll-loops}.
9031 @item -fwhole-program
9032 @opindex fwhole-program
9033 Assume that the current compilation unit represents the whole program being
9034 compiled. All public functions and variables with the exception of @code{main}
9035 and those merged by attribute @code{externally_visible} become static functions
9036 and in effect are optimized more aggressively by interprocedural optimizers.
9038 This option should not be used in combination with @option{-flto}.
9039 Instead relying on a linker plugin should provide safer and more precise
9042 @item -flto[=@var{n}]
9044 This option runs the standard link-time optimizer. When invoked
9045 with source code, it generates GIMPLE (one of GCC's internal
9046 representations) and writes it to special ELF sections in the object
9047 file. When the object files are linked together, all the function
9048 bodies are read from these ELF sections and instantiated as if they
9049 had been part of the same translation unit.
9051 To use the link-time optimizer, @option{-flto} and optimization
9052 options should be specified at compile time and during the final link.
9056 gcc -c -O2 -flto foo.c
9057 gcc -c -O2 -flto bar.c
9058 gcc -o myprog -flto -O2 foo.o bar.o
9061 The first two invocations to GCC save a bytecode representation
9062 of GIMPLE into special ELF sections inside @file{foo.o} and
9063 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9064 @file{foo.o} and @file{bar.o}, merges the two files into a single
9065 internal image, and compiles the result as usual. Since both
9066 @file{foo.o} and @file{bar.o} are merged into a single image, this
9067 causes all the interprocedural analyses and optimizations in GCC to
9068 work across the two files as if they were a single one. This means,
9069 for example, that the inliner is able to inline functions in
9070 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9072 Another (simpler) way to enable link-time optimization is:
9075 gcc -o myprog -flto -O2 foo.c bar.c
9078 The above generates bytecode for @file{foo.c} and @file{bar.c},
9079 merges them together into a single GIMPLE representation and optimizes
9080 them as usual to produce @file{myprog}.
9082 The only important thing to keep in mind is that to enable link-time
9083 optimizations you need to use the GCC driver to perform the link-step.
9084 GCC then automatically performs link-time optimization if any of the
9085 objects involved were compiled with the @option{-flto}. You generally
9086 should specify the optimization options to be used for link-time
9087 optimization though GCC tries to be clever at guessing an
9088 optimization level to use from the options used at compile-time
9089 if you fail to specify one at link-time. You can always override
9090 the automatic decision to do link-time optimization at link-time
9091 by passing @option{-fno-lto} to the link command.
9093 To make whole program optimization effective, it is necessary to make
9094 certain whole program assumptions. The compiler needs to know
9095 what functions and variables can be accessed by libraries and runtime
9096 outside of the link-time optimized unit. When supported by the linker,
9097 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9098 to the compiler about used and externally visible symbols. When
9099 the linker plugin is not available, @option{-fwhole-program} should be
9100 used to allow the compiler to make these assumptions, which leads
9101 to more aggressive optimization decisions.
9103 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9104 compiled with @option{-flto}, the generated object file is larger than
9105 a regular object file because it contains GIMPLE bytecodes and the usual
9106 final code (see @option{-ffat-lto-objects}. This means that
9107 object files with LTO information can be linked as normal object
9108 files; if @option{-fno-lto} is passed to the linker, no
9109 interprocedural optimizations are applied. Note that when
9110 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9111 but you cannot perform a regular, non-LTO link on them.
9113 Additionally, the optimization flags used to compile individual files
9114 are not necessarily related to those used at link time. For instance,
9117 gcc -c -O0 -ffat-lto-objects -flto foo.c
9118 gcc -c -O0 -ffat-lto-objects -flto bar.c
9119 gcc -o myprog -O3 foo.o bar.o
9122 This produces individual object files with unoptimized assembler
9123 code, but the resulting binary @file{myprog} is optimized at
9124 @option{-O3}. If, instead, the final binary is generated with
9125 @option{-fno-lto}, then @file{myprog} is not optimized.
9127 When producing the final binary, GCC only
9128 applies link-time optimizations to those files that contain bytecode.
9129 Therefore, you can mix and match object files and libraries with
9130 GIMPLE bytecodes and final object code. GCC automatically selects
9131 which files to optimize in LTO mode and which files to link without
9134 There are some code generation flags preserved by GCC when
9135 generating bytecodes, as they need to be used during the final link
9136 stage. Generally options specified at link-time override those
9137 specified at compile-time.
9139 If you do not specify an optimization level option @option{-O} at
9140 link-time then GCC computes one based on the optimization levels
9141 used when compiling the object files. The highest optimization
9144 Currently, the following options and their setting are take from
9145 the first object file that explicitely specified it:
9146 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9147 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9148 and all the @option{-m} target flags.
9150 Certain ABI changing flags are required to match in all compilation-units
9151 and trying to override this at link-time with a conflicting value
9152 is ignored. This includes options such as @option{-freg-struct-return}
9153 and @option{-fpcc-struct-return}.
9155 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9156 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9157 are passed through to the link stage and merged conservatively for
9158 conflicting translation units. Specifically
9159 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9160 precedence and for example @option{-ffp-contract=off} takes precedence
9161 over @option{-ffp-contract=fast}. You can override them at linke-time.
9163 It is recommended that you compile all the files participating in the
9164 same link with the same options and also specify those options at
9167 If LTO encounters objects with C linkage declared with incompatible
9168 types in separate translation units to be linked together (undefined
9169 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9170 issued. The behavior is still undefined at run time. Similar
9171 diagnostics may be raised for other languages.
9173 Another feature of LTO is that it is possible to apply interprocedural
9174 optimizations on files written in different languages:
9179 gfortran -c -flto baz.f90
9180 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9183 Notice that the final link is done with @command{g++} to get the C++
9184 runtime libraries and @option{-lgfortran} is added to get the Fortran
9185 runtime libraries. In general, when mixing languages in LTO mode, you
9186 should use the same link command options as when mixing languages in a
9187 regular (non-LTO) compilation.
9189 If object files containing GIMPLE bytecode are stored in a library archive, say
9190 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9191 are using a linker with plugin support. To create static libraries suitable
9192 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9193 and @command{ranlib};
9194 to show the symbols of object files with GIMPLE bytecode, use
9195 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9196 and @command{nm} have been compiled with plugin support. At link time, use the the
9197 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9198 the LTO optimization process:
9201 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9204 With the linker plugin enabled, the linker extracts the needed
9205 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9206 to make them part of the aggregated GIMPLE image to be optimized.
9208 If you are not using a linker with plugin support and/or do not
9209 enable the linker plugin, then the objects inside @file{libfoo.a}
9210 are extracted and linked as usual, but they do not participate
9211 in the LTO optimization process. In order to make a static library suitable
9212 for both LTO optimization and usual linkage, compile its object files with
9213 @option{-flto} @option{-ffat-lto-objects}.
9215 Link-time optimizations do not require the presence of the whole program to
9216 operate. If the program does not require any symbols to be exported, it is
9217 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9218 the interprocedural optimizers to use more aggressive assumptions which may
9219 lead to improved optimization opportunities.
9220 Use of @option{-fwhole-program} is not needed when linker plugin is
9221 active (see @option{-fuse-linker-plugin}).
9223 The current implementation of LTO makes no
9224 attempt to generate bytecode that is portable between different
9225 types of hosts. The bytecode files are versioned and there is a
9226 strict version check, so bytecode files generated in one version of
9227 GCC do not work with an older or newer version of GCC.
9229 Link-time optimization does not work well with generation of debugging
9230 information. Combining @option{-flto} with
9231 @option{-g} is currently experimental and expected to produce unexpected
9234 If you specify the optional @var{n}, the optimization and code
9235 generation done at link time is executed in parallel using @var{n}
9236 parallel jobs by utilizing an installed @command{make} program. The
9237 environment variable @env{MAKE} may be used to override the program
9238 used. The default value for @var{n} is 1.
9240 You can also specify @option{-flto=jobserver} to use GNU make's
9241 job server mode to determine the number of parallel jobs. This
9242 is useful when the Makefile calling GCC is already executing in parallel.
9243 You must prepend a @samp{+} to the command recipe in the parent Makefile
9244 for this to work. This option likely only works if @env{MAKE} is
9247 @item -flto-partition=@var{alg}
9248 @opindex flto-partition
9249 Specify the partitioning algorithm used by the link-time optimizer.
9250 The value is either @samp{1to1} to specify a partitioning mirroring
9251 the original source files or @samp{balanced} to specify partitioning
9252 into equally sized chunks (whenever possible) or @samp{max} to create
9253 new partition for every symbol where possible. Specifying @samp{none}
9254 as an algorithm disables partitioning and streaming completely.
9255 The default value is @samp{balanced}. While @samp{1to1} can be used
9256 as an workaround for various code ordering issues, the @samp{max}
9257 partitioning is intended for internal testing only.
9258 The value @samp{one} specifies that exactly one partition should be
9259 used while the value @samp{none} bypasses partitioning and executes
9260 the link-time optimization step directly from the WPA phase.
9262 @item -flto-odr-type-merging
9263 @opindex flto-odr-type-merging
9264 Enable streaming of mangled types names of C++ types and their unification
9265 at linktime. This increases size of LTO object files, but enable
9266 diagnostics about One Definition Rule violations.
9268 @item -flto-compression-level=@var{n}
9269 @opindex flto-compression-level
9270 This option specifies the level of compression used for intermediate
9271 language written to LTO object files, and is only meaningful in
9272 conjunction with LTO mode (@option{-flto}). Valid
9273 values are 0 (no compression) to 9 (maximum compression). Values
9274 outside this range are clamped to either 0 or 9. If the option is not
9275 given, a default balanced compression setting is used.
9278 @opindex flto-report
9279 Prints a report with internal details on the workings of the link-time
9280 optimizer. The contents of this report vary from version to version.
9281 It is meant to be useful to GCC developers when processing object
9282 files in LTO mode (via @option{-flto}).
9284 Disabled by default.
9286 @item -flto-report-wpa
9287 @opindex flto-report-wpa
9288 Like @option{-flto-report}, but only print for the WPA phase of Link
9291 @item -fuse-linker-plugin
9292 @opindex fuse-linker-plugin
9293 Enables the use of a linker plugin during link-time optimization. This
9294 option relies on plugin support in the linker, which is available in gold
9295 or in GNU ld 2.21 or newer.
9297 This option enables the extraction of object files with GIMPLE bytecode out
9298 of library archives. This improves the quality of optimization by exposing
9299 more code to the link-time optimizer. This information specifies what
9300 symbols can be accessed externally (by non-LTO object or during dynamic
9301 linking). Resulting code quality improvements on binaries (and shared
9302 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9303 See @option{-flto} for a description of the effect of this flag and how to
9306 This option is enabled by default when LTO support in GCC is enabled
9307 and GCC was configured for use with
9308 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9310 @item -ffat-lto-objects
9311 @opindex ffat-lto-objects
9312 Fat LTO objects are object files that contain both the intermediate language
9313 and the object code. This makes them usable for both LTO linking and normal
9314 linking. This option is effective only when compiling with @option{-flto}
9315 and is ignored at link time.
9317 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9318 requires the complete toolchain to be aware of LTO. It requires a linker with
9319 linker plugin support for basic functionality. Additionally,
9320 @command{nm}, @command{ar} and @command{ranlib}
9321 need to support linker plugins to allow a full-featured build environment
9322 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9323 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9324 to these tools. With non fat LTO makefiles need to be modified to use them.
9326 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9329 @item -fcompare-elim
9330 @opindex fcompare-elim
9331 After register allocation and post-register allocation instruction splitting,
9332 identify arithmetic instructions that compute processor flags similar to a
9333 comparison operation based on that arithmetic. If possible, eliminate the
9334 explicit comparison operation.
9336 This pass only applies to certain targets that cannot explicitly represent
9337 the comparison operation before register allocation is complete.
9339 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9342 @opindex fuse-ld=bfd
9343 Use the @command{bfd} linker instead of the default linker.
9346 @opindex fuse-ld=gold
9347 Use the @command{gold} linker instead of the default linker.
9349 @item -fcprop-registers
9350 @opindex fcprop-registers
9351 After register allocation and post-register allocation instruction splitting,
9352 perform a copy-propagation pass to try to reduce scheduling dependencies
9353 and occasionally eliminate the copy.
9355 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9357 @item -fprofile-correction
9358 @opindex fprofile-correction
9359 Profiles collected using an instrumented binary for multi-threaded programs may
9360 be inconsistent due to missed counter updates. When this option is specified,
9361 GCC uses heuristics to correct or smooth out such inconsistencies. By
9362 default, GCC emits an error message when an inconsistent profile is detected.
9364 @item -fprofile-dir=@var{path}
9365 @opindex fprofile-dir
9367 Set the directory to search for the profile data files in to @var{path}.
9368 This option affects only the profile data generated by
9369 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9370 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9371 and its related options. Both absolute and relative paths can be used.
9372 By default, GCC uses the current directory as @var{path}, thus the
9373 profile data file appears in the same directory as the object file.
9375 @item -fprofile-generate
9376 @itemx -fprofile-generate=@var{path}
9377 @opindex fprofile-generate
9379 Enable options usually used for instrumenting application to produce
9380 profile useful for later recompilation with profile feedback based
9381 optimization. You must use @option{-fprofile-generate} both when
9382 compiling and when linking your program.
9384 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9386 If @var{path} is specified, GCC looks at the @var{path} to find
9387 the profile feedback data files. See @option{-fprofile-dir}.
9390 @itemx -fprofile-use=@var{path}
9391 @opindex fprofile-use
9392 Enable profile feedback-directed optimizations,
9393 and the following optimizations
9394 which are generally profitable only with profile feedback available:
9395 @option{-fbranch-probabilities}, @option{-fvpt},
9396 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9397 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9399 By default, GCC emits an error message if the feedback profiles do not
9400 match the source code. This error can be turned into a warning by using
9401 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9404 If @var{path} is specified, GCC looks at the @var{path} to find
9405 the profile feedback data files. See @option{-fprofile-dir}.
9407 @item -fauto-profile
9408 @itemx -fauto-profile=@var{path}
9409 @opindex fauto-profile
9410 Enable sampling-based feedback-directed optimizations,
9411 and the following optimizations
9412 which are generally profitable only with profile feedback available:
9413 @option{-fbranch-probabilities}, @option{-fvpt},
9414 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9415 @option{-ftree-vectorize},
9416 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9417 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9418 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9420 @var{path} is the name of a file containing AutoFDO profile information.
9421 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9423 Producing an AutoFDO profile data file requires running your program
9424 with the @command{perf} utility on a supported GNU/Linux target system.
9425 For more information, see @uref{https://perf.wiki.kernel.org/}.
9429 perf record -e br_inst_retired:near_taken -b -o perf.data \
9433 Then use the @command{create_gcov} tool to convert the raw profile data
9434 to a format that can be used by GCC.@ You must also supply the
9435 unstripped binary for your program to this tool.
9436 See @uref{https://github.com/google/autofdo}.
9440 create_gcov --binary=your_program.unstripped --profile=perf.data \
9445 The following options control compiler behavior regarding floating-point
9446 arithmetic. These options trade off between speed and
9447 correctness. All must be specifically enabled.
9451 @opindex ffloat-store
9452 Do not store floating-point variables in registers, and inhibit other
9453 options that might change whether a floating-point value is taken from a
9456 @cindex floating-point precision
9457 This option prevents undesirable excess precision on machines such as
9458 the 68000 where the floating registers (of the 68881) keep more
9459 precision than a @code{double} is supposed to have. Similarly for the
9460 x86 architecture. For most programs, the excess precision does only
9461 good, but a few programs rely on the precise definition of IEEE floating
9462 point. Use @option{-ffloat-store} for such programs, after modifying
9463 them to store all pertinent intermediate computations into variables.
9465 @item -fexcess-precision=@var{style}
9466 @opindex fexcess-precision
9467 This option allows further control over excess precision on machines
9468 where floating-point registers have more precision than the IEEE
9469 @code{float} and @code{double} types and the processor does not
9470 support operations rounding to those types. By default,
9471 @option{-fexcess-precision=fast} is in effect; this means that
9472 operations are carried out in the precision of the registers and that
9473 it is unpredictable when rounding to the types specified in the source
9474 code takes place. When compiling C, if
9475 @option{-fexcess-precision=standard} is specified then excess
9476 precision follows the rules specified in ISO C99; in particular,
9477 both casts and assignments cause values to be rounded to their
9478 semantic types (whereas @option{-ffloat-store} only affects
9479 assignments). This option is enabled by default for C if a strict
9480 conformance option such as @option{-std=c99} is used.
9483 @option{-fexcess-precision=standard} is not implemented for languages
9484 other than C, and has no effect if
9485 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9486 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9487 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9488 semantics apply without excess precision, and in the latter, rounding
9493 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9494 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9495 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9497 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9499 This option is not turned on by any @option{-O} option besides
9500 @option{-Ofast} since it can result in incorrect output for programs
9501 that depend on an exact implementation of IEEE or ISO rules/specifications
9502 for math functions. It may, however, yield faster code for programs
9503 that do not require the guarantees of these specifications.
9505 @item -fno-math-errno
9506 @opindex fno-math-errno
9507 Do not set @code{errno} after calling math functions that are executed
9508 with a single instruction, e.g., @code{sqrt}. A program that relies on
9509 IEEE exceptions for math error handling may want to use this flag
9510 for speed while maintaining IEEE arithmetic compatibility.
9512 This option is not turned on by any @option{-O} option since
9513 it can result in incorrect output for programs that depend on
9514 an exact implementation of IEEE or ISO rules/specifications for
9515 math functions. It may, however, yield faster code for programs
9516 that do not require the guarantees of these specifications.
9518 The default is @option{-fmath-errno}.
9520 On Darwin systems, the math library never sets @code{errno}. There is
9521 therefore no reason for the compiler to consider the possibility that
9522 it might, and @option{-fno-math-errno} is the default.
9524 @item -funsafe-math-optimizations
9525 @opindex funsafe-math-optimizations
9527 Allow optimizations for floating-point arithmetic that (a) assume
9528 that arguments and results are valid and (b) may violate IEEE or
9529 ANSI standards. When used at link-time, it may include libraries
9530 or startup files that change the default FPU control word or other
9531 similar optimizations.
9533 This option is not turned on by any @option{-O} option since
9534 it can result in incorrect output for programs that depend on
9535 an exact implementation of IEEE or ISO rules/specifications for
9536 math functions. It may, however, yield faster code for programs
9537 that do not require the guarantees of these specifications.
9538 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9539 @option{-fassociative-math} and @option{-freciprocal-math}.
9541 The default is @option{-fno-unsafe-math-optimizations}.
9543 @item -fassociative-math
9544 @opindex fassociative-math
9546 Allow re-association of operands in series of floating-point operations.
9547 This violates the ISO C and C++ language standard by possibly changing
9548 computation result. NOTE: re-ordering may change the sign of zero as
9549 well as ignore NaNs and inhibit or create underflow or overflow (and
9550 thus cannot be used on code that relies on rounding behavior like
9551 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9552 and thus may not be used when ordered comparisons are required.
9553 This option requires that both @option{-fno-signed-zeros} and
9554 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9555 much sense with @option{-frounding-math}. For Fortran the option
9556 is automatically enabled when both @option{-fno-signed-zeros} and
9557 @option{-fno-trapping-math} are in effect.
9559 The default is @option{-fno-associative-math}.
9561 @item -freciprocal-math
9562 @opindex freciprocal-math
9564 Allow the reciprocal of a value to be used instead of dividing by
9565 the value if this enables optimizations. For example @code{x / y}
9566 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9567 is subject to common subexpression elimination. Note that this loses
9568 precision and increases the number of flops operating on the value.
9570 The default is @option{-fno-reciprocal-math}.
9572 @item -ffinite-math-only
9573 @opindex ffinite-math-only
9574 Allow optimizations for floating-point arithmetic that assume
9575 that arguments and results are not NaNs or +-Infs.
9577 This option is not turned on by any @option{-O} option since
9578 it can result in incorrect output for programs that depend on
9579 an exact implementation of IEEE or ISO rules/specifications for
9580 math functions. It may, however, yield faster code for programs
9581 that do not require the guarantees of these specifications.
9583 The default is @option{-fno-finite-math-only}.
9585 @item -fno-signed-zeros
9586 @opindex fno-signed-zeros
9587 Allow optimizations for floating-point arithmetic that ignore the
9588 signedness of zero. IEEE arithmetic specifies the behavior of
9589 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9590 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9591 This option implies that the sign of a zero result isn't significant.
9593 The default is @option{-fsigned-zeros}.
9595 @item -fno-trapping-math
9596 @opindex fno-trapping-math
9597 Compile code assuming that floating-point operations cannot generate
9598 user-visible traps. These traps include division by zero, overflow,
9599 underflow, inexact result and invalid operation. This option requires
9600 that @option{-fno-signaling-nans} be in effect. Setting this option may
9601 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9603 This option should never be turned on by any @option{-O} option since
9604 it can result in incorrect output for programs that depend on
9605 an exact implementation of IEEE or ISO rules/specifications for
9608 The default is @option{-ftrapping-math}.
9610 @item -frounding-math
9611 @opindex frounding-math
9612 Disable transformations and optimizations that assume default floating-point
9613 rounding behavior. This is round-to-zero for all floating point
9614 to integer conversions, and round-to-nearest for all other arithmetic
9615 truncations. This option should be specified for programs that change
9616 the FP rounding mode dynamically, or that may be executed with a
9617 non-default rounding mode. This option disables constant folding of
9618 floating-point expressions at compile time (which may be affected by
9619 rounding mode) and arithmetic transformations that are unsafe in the
9620 presence of sign-dependent rounding modes.
9622 The default is @option{-fno-rounding-math}.
9624 This option is experimental and does not currently guarantee to
9625 disable all GCC optimizations that are affected by rounding mode.
9626 Future versions of GCC may provide finer control of this setting
9627 using C99's @code{FENV_ACCESS} pragma. This command-line option
9628 will be used to specify the default state for @code{FENV_ACCESS}.
9630 @item -fsignaling-nans
9631 @opindex fsignaling-nans
9632 Compile code assuming that IEEE signaling NaNs may generate user-visible
9633 traps during floating-point operations. Setting this option disables
9634 optimizations that may change the number of exceptions visible with
9635 signaling NaNs. This option implies @option{-ftrapping-math}.
9637 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9640 The default is @option{-fno-signaling-nans}.
9642 This option is experimental and does not currently guarantee to
9643 disable all GCC optimizations that affect signaling NaN behavior.
9645 @item -fsingle-precision-constant
9646 @opindex fsingle-precision-constant
9647 Treat floating-point constants as single precision instead of
9648 implicitly converting them to double-precision constants.
9650 @item -fcx-limited-range
9651 @opindex fcx-limited-range
9652 When enabled, this option states that a range reduction step is not
9653 needed when performing complex division. Also, there is no checking
9654 whether the result of a complex multiplication or division is @code{NaN
9655 + I*NaN}, with an attempt to rescue the situation in that case. The
9656 default is @option{-fno-cx-limited-range}, but is enabled by
9657 @option{-ffast-math}.
9659 This option controls the default setting of the ISO C99
9660 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9663 @item -fcx-fortran-rules
9664 @opindex fcx-fortran-rules
9665 Complex multiplication and division follow Fortran rules. Range
9666 reduction is done as part of complex division, but there is no checking
9667 whether the result of a complex multiplication or division is @code{NaN
9668 + I*NaN}, with an attempt to rescue the situation in that case.
9670 The default is @option{-fno-cx-fortran-rules}.
9674 The following options control optimizations that may improve
9675 performance, but are not enabled by any @option{-O} options. This
9676 section includes experimental options that may produce broken code.
9679 @item -fbranch-probabilities
9680 @opindex fbranch-probabilities
9681 After running a program compiled with @option{-fprofile-arcs}
9682 (@pxref{Debugging Options,, Options for Debugging Your Program or
9683 @command{gcc}}), you can compile it a second time using
9684 @option{-fbranch-probabilities}, to improve optimizations based on
9685 the number of times each branch was taken. When a program
9686 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9687 counts to a file called @file{@var{sourcename}.gcda} for each source
9688 file. The information in this data file is very dependent on the
9689 structure of the generated code, so you must use the same source code
9690 and the same optimization options for both compilations.
9692 With @option{-fbranch-probabilities}, GCC puts a
9693 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9694 These can be used to improve optimization. Currently, they are only
9695 used in one place: in @file{reorg.c}, instead of guessing which path a
9696 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9697 exactly determine which path is taken more often.
9699 @item -fprofile-values
9700 @opindex fprofile-values
9701 If combined with @option{-fprofile-arcs}, it adds code so that some
9702 data about values of expressions in the program is gathered.
9704 With @option{-fbranch-probabilities}, it reads back the data gathered
9705 from profiling values of expressions for usage in optimizations.
9707 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9709 @item -fprofile-reorder-functions
9710 @opindex fprofile-reorder-functions
9711 Function reordering based on profile instrumentation collects
9712 first time of execution of a function and orders these functions
9715 Enabled with @option{-fprofile-use}.
9719 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9720 to add code to gather information about values of expressions.
9722 With @option{-fbranch-probabilities}, it reads back the data gathered
9723 and actually performs the optimizations based on them.
9724 Currently the optimizations include specialization of division operations
9725 using the knowledge about the value of the denominator.
9727 @item -frename-registers
9728 @opindex frename-registers
9729 Attempt to avoid false dependencies in scheduled code by making use
9730 of registers left over after register allocation. This optimization
9731 most benefits processors with lots of registers. Depending on the
9732 debug information format adopted by the target, however, it can
9733 make debugging impossible, since variables no longer stay in
9734 a ``home register''.
9736 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9738 @item -fschedule-fusion
9739 @opindex fschedule-fusion
9740 Performs a target dependent pass over the instruction stream to schedule
9741 instructions of same type together because target machine can execute them
9742 more efficiently if they are adjacent to each other in the instruction flow.
9744 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9748 Perform tail duplication to enlarge superblock size. This transformation
9749 simplifies the control flow of the function allowing other optimizations to do
9752 Enabled with @option{-fprofile-use}.
9754 @item -funroll-loops
9755 @opindex funroll-loops
9756 Unroll loops whose number of iterations can be determined at compile time or
9757 upon entry to the loop. @option{-funroll-loops} implies
9758 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9759 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9760 a small constant number of iterations). This option makes code larger, and may
9761 or may not make it run faster.
9763 Enabled with @option{-fprofile-use}.
9765 @item -funroll-all-loops
9766 @opindex funroll-all-loops
9767 Unroll all loops, even if their number of iterations is uncertain when
9768 the loop is entered. This usually makes programs run more slowly.
9769 @option{-funroll-all-loops} implies the same options as
9770 @option{-funroll-loops}.
9773 @opindex fpeel-loops
9774 Peels loops for which there is enough information that they do not
9775 roll much (from profile feedback). It also turns on complete loop peeling
9776 (i.e.@: complete removal of loops with small constant number of iterations).
9778 Enabled with @option{-fprofile-use}.
9780 @item -fmove-loop-invariants
9781 @opindex fmove-loop-invariants
9782 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9783 at level @option{-O1}
9785 @item -funswitch-loops
9786 @opindex funswitch-loops
9787 Move branches with loop invariant conditions out of the loop, with duplicates
9788 of the loop on both branches (modified according to result of the condition).
9790 @item -ffunction-sections
9791 @itemx -fdata-sections
9792 @opindex ffunction-sections
9793 @opindex fdata-sections
9794 Place each function or data item into its own section in the output
9795 file if the target supports arbitrary sections. The name of the
9796 function or the name of the data item determines the section's name
9799 Use these options on systems where the linker can perform optimizations
9800 to improve locality of reference in the instruction space. Most systems
9801 using the ELF object format and SPARC processors running Solaris 2 have
9802 linkers with such optimizations. AIX may have these optimizations in
9805 Only use these options when there are significant benefits from doing
9806 so. When you specify these options, the assembler and linker
9807 create larger object and executable files and are also slower.
9808 You cannot use @command{gprof} on all systems if you
9809 specify this option, and you may have problems with debugging if
9810 you specify both this option and @option{-g}.
9812 @item -fbranch-target-load-optimize
9813 @opindex fbranch-target-load-optimize
9814 Perform branch target register load optimization before prologue / epilogue
9816 The use of target registers can typically be exposed only during reload,
9817 thus hoisting loads out of loops and doing inter-block scheduling needs
9818 a separate optimization pass.
9820 @item -fbranch-target-load-optimize2
9821 @opindex fbranch-target-load-optimize2
9822 Perform branch target register load optimization after prologue / epilogue
9825 @item -fbtr-bb-exclusive
9826 @opindex fbtr-bb-exclusive
9827 When performing branch target register load optimization, don't reuse
9828 branch target registers within any basic block.
9830 @item -fstack-protector
9831 @opindex fstack-protector
9832 Emit extra code to check for buffer overflows, such as stack smashing
9833 attacks. This is done by adding a guard variable to functions with
9834 vulnerable objects. This includes functions that call @code{alloca}, and
9835 functions with buffers larger than 8 bytes. The guards are initialized
9836 when a function is entered and then checked when the function exits.
9837 If a guard check fails, an error message is printed and the program exits.
9839 @item -fstack-protector-all
9840 @opindex fstack-protector-all
9841 Like @option{-fstack-protector} except that all functions are protected.
9843 @item -fstack-protector-strong
9844 @opindex fstack-protector-strong
9845 Like @option{-fstack-protector} but includes additional functions to
9846 be protected --- those that have local array definitions, or have
9847 references to local frame addresses.
9849 @item -fsection-anchors
9850 @opindex fsection-anchors
9851 Try to reduce the number of symbolic address calculations by using
9852 shared ``anchor'' symbols to address nearby objects. This transformation
9853 can help to reduce the number of GOT entries and GOT accesses on some
9856 For example, the implementation of the following function @code{foo}:
9860 int foo (void) @{ return a + b + c; @}
9864 usually calculates the addresses of all three variables, but if you
9865 compile it with @option{-fsection-anchors}, it accesses the variables
9866 from a common anchor point instead. The effect is similar to the
9867 following pseudocode (which isn't valid C):
9872 register int *xr = &x;
9873 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9877 Not all targets support this option.
9879 @item --param @var{name}=@var{value}
9881 In some places, GCC uses various constants to control the amount of
9882 optimization that is done. For example, GCC does not inline functions
9883 that contain more than a certain number of instructions. You can
9884 control some of these constants on the command line using the
9885 @option{--param} option.
9887 The names of specific parameters, and the meaning of the values, are
9888 tied to the internals of the compiler, and are subject to change
9889 without notice in future releases.
9891 In each case, the @var{value} is an integer. The allowable choices for
9895 @item predictable-branch-outcome
9896 When branch is predicted to be taken with probability lower than this threshold
9897 (in percent), then it is considered well predictable. The default is 10.
9899 @item max-crossjump-edges
9900 The maximum number of incoming edges to consider for cross-jumping.
9901 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9902 the number of edges incoming to each block. Increasing values mean
9903 more aggressive optimization, making the compilation time increase with
9904 probably small improvement in executable size.
9906 @item min-crossjump-insns
9907 The minimum number of instructions that must be matched at the end
9908 of two blocks before cross-jumping is performed on them. This
9909 value is ignored in the case where all instructions in the block being
9910 cross-jumped from are matched. The default value is 5.
9912 @item max-grow-copy-bb-insns
9913 The maximum code size expansion factor when copying basic blocks
9914 instead of jumping. The expansion is relative to a jump instruction.
9915 The default value is 8.
9917 @item max-goto-duplication-insns
9918 The maximum number of instructions to duplicate to a block that jumps
9919 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9920 passes, GCC factors computed gotos early in the compilation process,
9921 and unfactors them as late as possible. Only computed jumps at the
9922 end of a basic blocks with no more than max-goto-duplication-insns are
9923 unfactored. The default value is 8.
9925 @item max-delay-slot-insn-search
9926 The maximum number of instructions to consider when looking for an
9927 instruction to fill a delay slot. If more than this arbitrary number of
9928 instructions are searched, the time savings from filling the delay slot
9929 are minimal, so stop searching. Increasing values mean more
9930 aggressive optimization, making the compilation time increase with probably
9931 small improvement in execution time.
9933 @item max-delay-slot-live-search
9934 When trying to fill delay slots, the maximum number of instructions to
9935 consider when searching for a block with valid live register
9936 information. Increasing this arbitrarily chosen value means more
9937 aggressive optimization, increasing the compilation time. This parameter
9938 should be removed when the delay slot code is rewritten to maintain the
9941 @item max-gcse-memory
9942 The approximate maximum amount of memory that can be allocated in
9943 order to perform the global common subexpression elimination
9944 optimization. If more memory than specified is required, the
9945 optimization is not done.
9947 @item max-gcse-insertion-ratio
9948 If the ratio of expression insertions to deletions is larger than this value
9949 for any expression, then RTL PRE inserts or removes the expression and thus
9950 leaves partially redundant computations in the instruction stream. The default value is 20.
9952 @item max-pending-list-length
9953 The maximum number of pending dependencies scheduling allows
9954 before flushing the current state and starting over. Large functions
9955 with few branches or calls can create excessively large lists which
9956 needlessly consume memory and resources.
9958 @item max-modulo-backtrack-attempts
9959 The maximum number of backtrack attempts the scheduler should make
9960 when modulo scheduling a loop. Larger values can exponentially increase
9963 @item max-inline-insns-single
9964 Several parameters control the tree inliner used in GCC@.
9965 This number sets the maximum number of instructions (counted in GCC's
9966 internal representation) in a single function that the tree inliner
9967 considers for inlining. This only affects functions declared
9968 inline and methods implemented in a class declaration (C++).
9969 The default value is 400.
9971 @item max-inline-insns-auto
9972 When you use @option{-finline-functions} (included in @option{-O3}),
9973 a lot of functions that would otherwise not be considered for inlining
9974 by the compiler are investigated. To those functions, a different
9975 (more restrictive) limit compared to functions declared inline can
9977 The default value is 40.
9979 @item inline-min-speedup
9980 When estimated performance improvement of caller + callee runtime exceeds this
9981 threshold (in precent), the function can be inlined regardless the limit on
9982 @option{--param max-inline-insns-single} and @option{--param
9983 max-inline-insns-auto}.
9985 @item large-function-insns
9986 The limit specifying really large functions. For functions larger than this
9987 limit after inlining, inlining is constrained by
9988 @option{--param large-function-growth}. This parameter is useful primarily
9989 to avoid extreme compilation time caused by non-linear algorithms used by the
9991 The default value is 2700.
9993 @item large-function-growth
9994 Specifies maximal growth of large function caused by inlining in percents.
9995 The default value is 100 which limits large function growth to 2.0 times
9998 @item large-unit-insns
9999 The limit specifying large translation unit. Growth caused by inlining of
10000 units larger than this limit is limited by @option{--param inline-unit-growth}.
10001 For small units this might be too tight.
10002 For example, consider a unit consisting of function A
10003 that is inline and B that just calls A three times. If B is small relative to
10004 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10005 large units consisting of small inlineable functions, however, the overall unit
10006 growth limit is needed to avoid exponential explosion of code size. Thus for
10007 smaller units, the size is increased to @option{--param large-unit-insns}
10008 before applying @option{--param inline-unit-growth}. The default is 10000.
10010 @item inline-unit-growth
10011 Specifies maximal overall growth of the compilation unit caused by inlining.
10012 The default value is 30 which limits unit growth to 1.3 times the original
10013 size. Cold functions (either marked cold via an attribute or by profile
10014 feedback) are not accounted into the unit size.
10016 @item ipcp-unit-growth
10017 Specifies maximal overall growth of the compilation unit caused by
10018 interprocedural constant propagation. The default value is 10 which limits
10019 unit growth to 1.1 times the original size.
10021 @item large-stack-frame
10022 The limit specifying large stack frames. While inlining the algorithm is trying
10023 to not grow past this limit too much. The default value is 256 bytes.
10025 @item large-stack-frame-growth
10026 Specifies maximal growth of large stack frames caused by inlining in percents.
10027 The default value is 1000 which limits large stack frame growth to 11 times
10030 @item max-inline-insns-recursive
10031 @itemx max-inline-insns-recursive-auto
10032 Specifies the maximum number of instructions an out-of-line copy of a
10033 self-recursive inline
10034 function can grow into by performing recursive inlining.
10036 @option{--param max-inline-insns-recursive} applies to functions
10038 For functions not declared inline, recursive inlining
10039 happens only when @option{-finline-functions} (included in @option{-O3}) is
10040 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10041 default value is 450.
10043 @item max-inline-recursive-depth
10044 @itemx max-inline-recursive-depth-auto
10045 Specifies the maximum recursion depth used for recursive inlining.
10047 @option{--param max-inline-recursive-depth} applies to functions
10048 declared inline. For functions not declared inline, recursive inlining
10049 happens only when @option{-finline-functions} (included in @option{-O3}) is
10050 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10051 default value is 8.
10053 @item min-inline-recursive-probability
10054 Recursive inlining is profitable only for function having deep recursion
10055 in average and can hurt for function having little recursion depth by
10056 increasing the prologue size or complexity of function body to other
10059 When profile feedback is available (see @option{-fprofile-generate}) the actual
10060 recursion depth can be guessed from probability that function recurses via a
10061 given call expression. This parameter limits inlining only to call expressions
10062 whose probability exceeds the given threshold (in percents).
10063 The default value is 10.
10065 @item early-inlining-insns
10066 Specify growth that the early inliner can make. In effect it increases
10067 the amount of inlining for code having a large abstraction penalty.
10068 The default value is 10.
10070 @item max-early-inliner-iterations
10071 Limit of iterations of the early inliner. This basically bounds
10072 the number of nested indirect calls the early inliner can resolve.
10073 Deeper chains are still handled by late inlining.
10075 @item comdat-sharing-probability
10076 Probability (in percent) that C++ inline function with comdat visibility
10077 are shared across multiple compilation units. The default value is 20.
10079 @item profile-func-internal-id
10080 A parameter to control whether to use function internal id in profile
10081 database lookup. If the value is 0, the compiler uses an id that
10082 is based on function assembler name and filename, which makes old profile
10083 data more tolerant to source changes such as function reordering etc.
10084 The default value is 0.
10086 @item min-vect-loop-bound
10087 The minimum number of iterations under which loops are not vectorized
10088 when @option{-ftree-vectorize} is used. The number of iterations after
10089 vectorization needs to be greater than the value specified by this option
10090 to allow vectorization. The default value is 0.
10092 @item gcse-cost-distance-ratio
10093 Scaling factor in calculation of maximum distance an expression
10094 can be moved by GCSE optimizations. This is currently supported only in the
10095 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10096 is with simple expressions, i.e., the expressions that have cost
10097 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10098 hoisting of simple expressions. The default value is 10.
10100 @item gcse-unrestricted-cost
10101 Cost, roughly measured as the cost of a single typical machine
10102 instruction, at which GCSE optimizations do not constrain
10103 the distance an expression can travel. This is currently
10104 supported only in the code hoisting pass. The lesser the cost,
10105 the more aggressive code hoisting is. Specifying 0
10106 allows all expressions to travel unrestricted distances.
10107 The default value is 3.
10109 @item max-hoist-depth
10110 The depth of search in the dominator tree for expressions to hoist.
10111 This is used to avoid quadratic behavior in hoisting algorithm.
10112 The value of 0 does not limit on the search, but may slow down compilation
10113 of huge functions. The default value is 30.
10115 @item max-tail-merge-comparisons
10116 The maximum amount of similar bbs to compare a bb with. This is used to
10117 avoid quadratic behavior in tree tail merging. The default value is 10.
10119 @item max-tail-merge-iterations
10120 The maximum amount of iterations of the pass over the function. This is used to
10121 limit compilation time in tree tail merging. The default value is 2.
10123 @item max-unrolled-insns
10124 The maximum number of instructions that a loop may have to be unrolled.
10125 If a loop is unrolled, this parameter also determines how many times
10126 the loop code is unrolled.
10128 @item max-average-unrolled-insns
10129 The maximum number of instructions biased by probabilities of their execution
10130 that a loop may have to be unrolled. If a loop is unrolled,
10131 this parameter also determines how many times the loop code is unrolled.
10133 @item max-unroll-times
10134 The maximum number of unrollings of a single loop.
10136 @item max-peeled-insns
10137 The maximum number of instructions that a loop may have to be peeled.
10138 If a loop is peeled, this parameter also determines how many times
10139 the loop code is peeled.
10141 @item max-peel-times
10142 The maximum number of peelings of a single loop.
10144 @item max-peel-branches
10145 The maximum number of branches on the hot path through the peeled sequence.
10147 @item max-completely-peeled-insns
10148 The maximum number of insns of a completely peeled loop.
10150 @item max-completely-peel-times
10151 The maximum number of iterations of a loop to be suitable for complete peeling.
10153 @item max-completely-peel-loop-nest-depth
10154 The maximum depth of a loop nest suitable for complete peeling.
10156 @item max-unswitch-insns
10157 The maximum number of insns of an unswitched loop.
10159 @item max-unswitch-level
10160 The maximum number of branches unswitched in a single loop.
10162 @item lim-expensive
10163 The minimum cost of an expensive expression in the loop invariant motion.
10165 @item iv-consider-all-candidates-bound
10166 Bound on number of candidates for induction variables, below which
10167 all candidates are considered for each use in induction variable
10168 optimizations. If there are more candidates than this,
10169 only the most relevant ones are considered to avoid quadratic time complexity.
10171 @item iv-max-considered-uses
10172 The induction variable optimizations give up on loops that contain more
10173 induction variable uses.
10175 @item iv-always-prune-cand-set-bound
10176 If the number of candidates in the set is smaller than this value,
10177 always try to remove unnecessary ivs from the set
10178 when adding a new one.
10180 @item scev-max-expr-size
10181 Bound on size of expressions used in the scalar evolutions analyzer.
10182 Large expressions slow the analyzer.
10184 @item scev-max-expr-complexity
10185 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10186 Complex expressions slow the analyzer.
10188 @item omega-max-vars
10189 The maximum number of variables in an Omega constraint system.
10190 The default value is 128.
10192 @item omega-max-geqs
10193 The maximum number of inequalities in an Omega constraint system.
10194 The default value is 256.
10196 @item omega-max-eqs
10197 The maximum number of equalities in an Omega constraint system.
10198 The default value is 128.
10200 @item omega-max-wild-cards
10201 The maximum number of wildcard variables that the Omega solver is
10202 able to insert. The default value is 18.
10204 @item omega-hash-table-size
10205 The size of the hash table in the Omega solver. The default value is
10208 @item omega-max-keys
10209 The maximal number of keys used by the Omega solver. The default
10212 @item omega-eliminate-redundant-constraints
10213 When set to 1, use expensive methods to eliminate all redundant
10214 constraints. The default value is 0.
10216 @item vect-max-version-for-alignment-checks
10217 The maximum number of run-time checks that can be performed when
10218 doing loop versioning for alignment in the vectorizer.
10220 @item vect-max-version-for-alias-checks
10221 The maximum number of run-time checks that can be performed when
10222 doing loop versioning for alias in the vectorizer.
10224 @item vect-max-peeling-for-alignment
10225 The maximum number of loop peels to enhance access alignment
10226 for vectorizer. Value -1 means 'no limit'.
10228 @item max-iterations-to-track
10229 The maximum number of iterations of a loop the brute-force algorithm
10230 for analysis of the number of iterations of the loop tries to evaluate.
10232 @item hot-bb-count-ws-permille
10233 A basic block profile count is considered hot if it contributes to
10234 the given permillage (i.e. 0...1000) of the entire profiled execution.
10236 @item hot-bb-frequency-fraction
10237 Select fraction of the entry block frequency of executions of basic block in
10238 function given basic block needs to have to be considered hot.
10240 @item max-predicted-iterations
10241 The maximum number of loop iterations we predict statically. This is useful
10242 in cases where a function contains a single loop with known bound and
10243 another loop with unknown bound.
10244 The known number of iterations is predicted correctly, while
10245 the unknown number of iterations average to roughly 10. This means that the
10246 loop without bounds appears artificially cold relative to the other one.
10248 @item builtin-expect-probability
10249 Control the probability of the expression having the specified value. This
10250 parameter takes a percentage (i.e. 0 ... 100) as input.
10251 The default probability of 90 is obtained empirically.
10253 @item align-threshold
10255 Select fraction of the maximal frequency of executions of a basic block in
10256 a function to align the basic block.
10258 @item align-loop-iterations
10260 A loop expected to iterate at least the selected number of iterations is
10263 @item tracer-dynamic-coverage
10264 @itemx tracer-dynamic-coverage-feedback
10266 This value is used to limit superblock formation once the given percentage of
10267 executed instructions is covered. This limits unnecessary code size
10270 The @option{tracer-dynamic-coverage-feedback} is used only when profile
10271 feedback is available. The real profiles (as opposed to statically estimated
10272 ones) are much less balanced allowing the threshold to be larger value.
10274 @item tracer-max-code-growth
10275 Stop tail duplication once code growth has reached given percentage. This is
10276 a rather artificial limit, as most of the duplicates are eliminated later in
10277 cross jumping, so it may be set to much higher values than is the desired code
10280 @item tracer-min-branch-ratio
10282 Stop reverse growth when the reverse probability of best edge is less than this
10283 threshold (in percent).
10285 @item tracer-min-branch-ratio
10286 @itemx tracer-min-branch-ratio-feedback
10288 Stop forward growth if the best edge has probability lower than this
10291 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10292 compilation for profile feedback and one for compilation without. The value
10293 for compilation with profile feedback needs to be more conservative (higher) in
10294 order to make tracer effective.
10296 @item max-cse-path-length
10298 The maximum number of basic blocks on path that CSE considers.
10301 @item max-cse-insns
10302 The maximum number of instructions CSE processes before flushing.
10303 The default is 1000.
10305 @item ggc-min-expand
10307 GCC uses a garbage collector to manage its own memory allocation. This
10308 parameter specifies the minimum percentage by which the garbage
10309 collector's heap should be allowed to expand between collections.
10310 Tuning this may improve compilation speed; it has no effect on code
10313 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10314 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10315 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10316 GCC is not able to calculate RAM on a particular platform, the lower
10317 bound of 30% is used. Setting this parameter and
10318 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10319 every opportunity. This is extremely slow, but can be useful for
10322 @item ggc-min-heapsize
10324 Minimum size of the garbage collector's heap before it begins bothering
10325 to collect garbage. The first collection occurs after the heap expands
10326 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10327 tuning this may improve compilation speed, and has no effect on code
10330 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10331 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10332 with a lower bound of 4096 (four megabytes) and an upper bound of
10333 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10334 particular platform, the lower bound is used. Setting this parameter
10335 very large effectively disables garbage collection. Setting this
10336 parameter and @option{ggc-min-expand} to zero causes a full collection
10337 to occur at every opportunity.
10339 @item max-reload-search-insns
10340 The maximum number of instruction reload should look backward for equivalent
10341 register. Increasing values mean more aggressive optimization, making the
10342 compilation time increase with probably slightly better performance.
10343 The default value is 100.
10345 @item max-cselib-memory-locations
10346 The maximum number of memory locations cselib should take into account.
10347 Increasing values mean more aggressive optimization, making the compilation time
10348 increase with probably slightly better performance. The default value is 500.
10350 @item reorder-blocks-duplicate
10351 @itemx reorder-blocks-duplicate-feedback
10353 Used by the basic block reordering pass to decide whether to use unconditional
10354 branch or duplicate the code on its destination. Code is duplicated when its
10355 estimated size is smaller than this value multiplied by the estimated size of
10356 unconditional jump in the hot spots of the program.
10358 The @option{reorder-block-duplicate-feedback} is used only when profile
10359 feedback is available. It may be set to higher values than
10360 @option{reorder-block-duplicate} since information about the hot spots is more
10363 @item max-sched-ready-insns
10364 The maximum number of instructions ready to be issued the scheduler should
10365 consider at any given time during the first scheduling pass. Increasing
10366 values mean more thorough searches, making the compilation time increase
10367 with probably little benefit. The default value is 100.
10369 @item max-sched-region-blocks
10370 The maximum number of blocks in a region to be considered for
10371 interblock scheduling. The default value is 10.
10373 @item max-pipeline-region-blocks
10374 The maximum number of blocks in a region to be considered for
10375 pipelining in the selective scheduler. The default value is 15.
10377 @item max-sched-region-insns
10378 The maximum number of insns in a region to be considered for
10379 interblock scheduling. The default value is 100.
10381 @item max-pipeline-region-insns
10382 The maximum number of insns in a region to be considered for
10383 pipelining in the selective scheduler. The default value is 200.
10385 @item min-spec-prob
10386 The minimum probability (in percents) of reaching a source block
10387 for interblock speculative scheduling. The default value is 40.
10389 @item max-sched-extend-regions-iters
10390 The maximum number of iterations through CFG to extend regions.
10391 A value of 0 (the default) disables region extensions.
10393 @item max-sched-insn-conflict-delay
10394 The maximum conflict delay for an insn to be considered for speculative motion.
10395 The default value is 3.
10397 @item sched-spec-prob-cutoff
10398 The minimal probability of speculation success (in percents), so that
10399 speculative insns are scheduled.
10400 The default value is 40.
10402 @item sched-spec-state-edge-prob-cutoff
10403 The minimum probability an edge must have for the scheduler to save its
10405 The default value is 10.
10407 @item sched-mem-true-dep-cost
10408 Minimal distance (in CPU cycles) between store and load targeting same
10409 memory locations. The default value is 1.
10411 @item selsched-max-lookahead
10412 The maximum size of the lookahead window of selective scheduling. It is a
10413 depth of search for available instructions.
10414 The default value is 50.
10416 @item selsched-max-sched-times
10417 The maximum number of times that an instruction is scheduled during
10418 selective scheduling. This is the limit on the number of iterations
10419 through which the instruction may be pipelined. The default value is 2.
10421 @item selsched-max-insns-to-rename
10422 The maximum number of best instructions in the ready list that are considered
10423 for renaming in the selective scheduler. The default value is 2.
10426 The minimum value of stage count that swing modulo scheduler
10427 generates. The default value is 2.
10429 @item max-last-value-rtl
10430 The maximum size measured as number of RTLs that can be recorded in an expression
10431 in combiner for a pseudo register as last known value of that register. The default
10434 @item max-combine-insns
10435 The maximum number of instructions the RTL combiner tries to combine.
10436 The default value is 2 at @option{-Og} and 4 otherwise.
10438 @item integer-share-limit
10439 Small integer constants can use a shared data structure, reducing the
10440 compiler's memory usage and increasing its speed. This sets the maximum
10441 value of a shared integer constant. The default value is 256.
10443 @item ssp-buffer-size
10444 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10445 protection when @option{-fstack-protection} is used.
10447 @item min-size-for-stack-sharing
10448 The minimum size of variables taking part in stack slot sharing when not
10449 optimizing. The default value is 32.
10451 @item max-jump-thread-duplication-stmts
10452 Maximum number of statements allowed in a block that needs to be
10453 duplicated when threading jumps.
10455 @item max-fields-for-field-sensitive
10456 Maximum number of fields in a structure treated in
10457 a field sensitive manner during pointer analysis. The default is zero
10458 for @option{-O0} and @option{-O1},
10459 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10461 @item prefetch-latency
10462 Estimate on average number of instructions that are executed before
10463 prefetch finishes. The distance prefetched ahead is proportional
10464 to this constant. Increasing this number may also lead to less
10465 streams being prefetched (see @option{simultaneous-prefetches}).
10467 @item simultaneous-prefetches
10468 Maximum number of prefetches that can run at the same time.
10470 @item l1-cache-line-size
10471 The size of cache line in L1 cache, in bytes.
10473 @item l1-cache-size
10474 The size of L1 cache, in kilobytes.
10476 @item l2-cache-size
10477 The size of L2 cache, in kilobytes.
10479 @item min-insn-to-prefetch-ratio
10480 The minimum ratio between the number of instructions and the
10481 number of prefetches to enable prefetching in a loop.
10483 @item prefetch-min-insn-to-mem-ratio
10484 The minimum ratio between the number of instructions and the
10485 number of memory references to enable prefetching in a loop.
10487 @item use-canonical-types
10488 Whether the compiler should use the ``canonical'' type system. By
10489 default, this should always be 1, which uses a more efficient internal
10490 mechanism for comparing types in C++ and Objective-C++. However, if
10491 bugs in the canonical type system are causing compilation failures,
10492 set this value to 0 to disable canonical types.
10494 @item switch-conversion-max-branch-ratio
10495 Switch initialization conversion refuses to create arrays that are
10496 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10497 branches in the switch.
10499 @item max-partial-antic-length
10500 Maximum length of the partial antic set computed during the tree
10501 partial redundancy elimination optimization (@option{-ftree-pre}) when
10502 optimizing at @option{-O3} and above. For some sorts of source code
10503 the enhanced partial redundancy elimination optimization can run away,
10504 consuming all of the memory available on the host machine. This
10505 parameter sets a limit on the length of the sets that are computed,
10506 which prevents the runaway behavior. Setting a value of 0 for
10507 this parameter allows an unlimited set length.
10509 @item sccvn-max-scc-size
10510 Maximum size of a strongly connected component (SCC) during SCCVN
10511 processing. If this limit is hit, SCCVN processing for the whole
10512 function is not done and optimizations depending on it are
10513 disabled. The default maximum SCC size is 10000.
10515 @item sccvn-max-alias-queries-per-access
10516 Maximum number of alias-oracle queries we perform when looking for
10517 redundancies for loads and stores. If this limit is hit the search
10518 is aborted and the load or store is not considered redundant. The
10519 number of queries is algorithmically limited to the number of
10520 stores on all paths from the load to the function entry.
10521 The default maxmimum number of queries is 1000.
10523 @item ira-max-loops-num
10524 IRA uses regional register allocation by default. If a function
10525 contains more loops than the number given by this parameter, only at most
10526 the given number of the most frequently-executed loops form regions
10527 for regional register allocation. The default value of the
10530 @item ira-max-conflict-table-size
10531 Although IRA uses a sophisticated algorithm to compress the conflict
10532 table, the table can still require excessive amounts of memory for
10533 huge functions. If the conflict table for a function could be more
10534 than the size in MB given by this parameter, the register allocator
10535 instead uses a faster, simpler, and lower-quality
10536 algorithm that does not require building a pseudo-register conflict table.
10537 The default value of the parameter is 2000.
10539 @item ira-loop-reserved-regs
10540 IRA can be used to evaluate more accurate register pressure in loops
10541 for decisions to move loop invariants (see @option{-O3}). The number
10542 of available registers reserved for some other purposes is given
10543 by this parameter. The default value of the parameter is 2, which is
10544 the minimal number of registers needed by typical instructions.
10545 This value is the best found from numerous experiments.
10547 @item loop-invariant-max-bbs-in-loop
10548 Loop invariant motion can be very expensive, both in compilation time and
10549 in amount of needed compile-time memory, with very large loops. Loops
10550 with more basic blocks than this parameter won't have loop invariant
10551 motion optimization performed on them. The default value of the
10552 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10554 @item loop-max-datarefs-for-datadeps
10555 Building data dapendencies is expensive for very large loops. This
10556 parameter limits the number of data references in loops that are
10557 considered for data dependence analysis. These large loops are no
10558 handled by the optimizations using loop data dependencies.
10559 The default value is 1000.
10561 @item max-vartrack-size
10562 Sets a maximum number of hash table slots to use during variable
10563 tracking dataflow analysis of any function. If this limit is exceeded
10564 with variable tracking at assignments enabled, analysis for that
10565 function is retried without it, after removing all debug insns from
10566 the function. If the limit is exceeded even without debug insns, var
10567 tracking analysis is completely disabled for the function. Setting
10568 the parameter to zero makes it unlimited.
10570 @item max-vartrack-expr-depth
10571 Sets a maximum number of recursion levels when attempting to map
10572 variable names or debug temporaries to value expressions. This trades
10573 compilation time for more complete debug information. If this is set too
10574 low, value expressions that are available and could be represented in
10575 debug information may end up not being used; setting this higher may
10576 enable the compiler to find more complex debug expressions, but compile
10577 time and memory use may grow. The default is 12.
10579 @item min-nondebug-insn-uid
10580 Use uids starting at this parameter for nondebug insns. The range below
10581 the parameter is reserved exclusively for debug insns created by
10582 @option{-fvar-tracking-assignments}, but debug insns may get
10583 (non-overlapping) uids above it if the reserved range is exhausted.
10585 @item ipa-sra-ptr-growth-factor
10586 IPA-SRA replaces a pointer to an aggregate with one or more new
10587 parameters only when their cumulative size is less or equal to
10588 @option{ipa-sra-ptr-growth-factor} times the size of the original
10591 @item sra-max-scalarization-size-Ospeed
10592 @item sra-max-scalarization-size-Osize
10593 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10594 replace scalar parts of aggregates with uses of independent scalar
10595 variables. These parameters control the maximum size, in storage units,
10596 of aggregate which is considered for replacement when compiling for
10598 (@option{sra-max-scalarization-size-Ospeed}) or size
10599 (@option{sra-max-scalarization-size-Osize}) respectively.
10601 @item tm-max-aggregate-size
10602 When making copies of thread-local variables in a transaction, this
10603 parameter specifies the size in bytes after which variables are
10604 saved with the logging functions as opposed to save/restore code
10605 sequence pairs. This option only applies when using
10608 @item graphite-max-nb-scop-params
10609 To avoid exponential effects in the Graphite loop transforms, the
10610 number of parameters in a Static Control Part (SCoP) is bounded. The
10611 default value is 10 parameters. A variable whose value is unknown at
10612 compilation time and defined outside a SCoP is a parameter of the SCoP.
10614 @item graphite-max-bbs-per-function
10615 To avoid exponential effects in the detection of SCoPs, the size of
10616 the functions analyzed by Graphite is bounded. The default value is
10619 @item loop-block-tile-size
10620 Loop blocking or strip mining transforms, enabled with
10621 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10622 loop in the loop nest by a given number of iterations. The strip
10623 length can be changed using the @option{loop-block-tile-size}
10624 parameter. The default value is 51 iterations.
10626 @item loop-unroll-jam-size
10627 Specify the unroll factor for the @option{-floop-unroll-and-jam}. The
10628 default value is 4.
10630 @item loop-unroll-jam-depth
10631 Specify the dimension to be unrolled (counting from the most inner loop)
10632 for the @option{-floop-unroll-and-jam}. The default value is 2.
10634 @item ipa-cp-value-list-size
10635 IPA-CP attempts to track all possible values and types passed to a function's
10636 parameter in order to propagate them and perform devirtualization.
10637 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10638 stores per one formal parameter of a function.
10640 @item ipa-cp-eval-threshold
10641 IPA-CP calculates its own score of cloning profitability heuristics
10642 and performs those cloning opportunities with scores that exceed
10643 @option{ipa-cp-eval-threshold}.
10645 @item ipa-max-agg-items
10646 IPA-CP is also capable to propagate a number of scalar values passed
10647 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10648 number of such values per one parameter.
10650 @item ipa-cp-loop-hint-bonus
10651 When IPA-CP determines that a cloning candidate would make the number
10652 of iterations of a loop known, it adds a bonus of
10653 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10656 @item ipa-cp-array-index-hint-bonus
10657 When IPA-CP determines that a cloning candidate would make the index of
10658 an array access known, it adds a bonus of
10659 @option{ipa-cp-array-index-hint-bonus} to the profitability
10660 score of the candidate.
10662 @item ipa-max-aa-steps
10663 During its analysis of function bodies, IPA-CP employs alias analysis
10664 in order to track values pointed to by function parameters. In order
10665 not spend too much time analyzing huge functions, it gives up and
10666 consider all memory clobbered after examining
10667 @option{ipa-max-aa-steps} statements modifying memory.
10669 @item lto-partitions
10670 Specify desired number of partitions produced during WHOPR compilation.
10671 The number of partitions should exceed the number of CPUs used for compilation.
10672 The default value is 32.
10674 @item lto-minpartition
10675 Size of minimal partition for WHOPR (in estimated instructions).
10676 This prevents expenses of splitting very small programs into too many
10679 @item cxx-max-namespaces-for-diagnostic-help
10680 The maximum number of namespaces to consult for suggestions when C++
10681 name lookup fails for an identifier. The default is 1000.
10683 @item sink-frequency-threshold
10684 The maximum relative execution frequency (in percents) of the target block
10685 relative to a statement's original block to allow statement sinking of a
10686 statement. Larger numbers result in more aggressive statement sinking.
10687 The default value is 75. A small positive adjustment is applied for
10688 statements with memory operands as those are even more profitable so sink.
10690 @item max-stores-to-sink
10691 The maximum number of conditional stores paires that can be sunk. Set to 0
10692 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10693 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10695 @item allow-store-data-races
10696 Allow optimizers to introduce new data races on stores.
10697 Set to 1 to allow, otherwise to 0. This option is enabled by default
10698 at optimization level @option{-Ofast}.
10700 @item case-values-threshold
10701 The smallest number of different values for which it is best to use a
10702 jump-table instead of a tree of conditional branches. If the value is
10703 0, use the default for the machine. The default is 0.
10705 @item tree-reassoc-width
10706 Set the maximum number of instructions executed in parallel in
10707 reassociated tree. This parameter overrides target dependent
10708 heuristics used by default if has non zero value.
10710 @item sched-pressure-algorithm
10711 Choose between the two available implementations of
10712 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10713 and is the more likely to prevent instructions from being reordered.
10714 Algorithm 2 was designed to be a compromise between the relatively
10715 conservative approach taken by algorithm 1 and the rather aggressive
10716 approach taken by the default scheduler. It relies more heavily on
10717 having a regular register file and accurate register pressure classes.
10718 See @file{haifa-sched.c} in the GCC sources for more details.
10720 The default choice depends on the target.
10722 @item max-slsr-cand-scan
10723 Set the maximum number of existing candidates that are considered when
10724 seeking a basis for a new straight-line strength reduction candidate.
10727 Enable buffer overflow detection for global objects. This kind
10728 of protection is enabled by default if you are using
10729 @option{-fsanitize=address} option.
10730 To disable global objects protection use @option{--param asan-globals=0}.
10733 Enable buffer overflow detection for stack objects. This kind of
10734 protection is enabled by default when using@option{-fsanitize=address}.
10735 To disable stack protection use @option{--param asan-stack=0} option.
10737 @item asan-instrument-reads
10738 Enable buffer overflow detection for memory reads. This kind of
10739 protection is enabled by default when using @option{-fsanitize=address}.
10740 To disable memory reads protection use
10741 @option{--param asan-instrument-reads=0}.
10743 @item asan-instrument-writes
10744 Enable buffer overflow detection for memory writes. This kind of
10745 protection is enabled by default when using @option{-fsanitize=address}.
10746 To disable memory writes protection use
10747 @option{--param asan-instrument-writes=0} option.
10749 @item asan-memintrin
10750 Enable detection for built-in functions. This kind of protection
10751 is enabled by default when using @option{-fsanitize=address}.
10752 To disable built-in functions protection use
10753 @option{--param asan-memintrin=0}.
10755 @item asan-use-after-return
10756 Enable detection of use-after-return. This kind of protection
10757 is enabled by default when using @option{-fsanitize=address} option.
10758 To disable use-after-return detection use
10759 @option{--param asan-use-after-return=0}.
10761 @item asan-instrumentation-with-call-threshold
10762 If number of memory accesses in function being instrumented
10763 is greater or equal to this number, use callbacks instead of inline checks.
10764 E.g. to disable inline code use
10765 @option{--param asan-instrumentation-with-call-threshold=0}.
10767 @item chkp-max-ctor-size
10768 Static constructors generated by Pointer Bounds Checker may become very
10769 large and significantly increase compile time at optimization level
10770 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10771 in a single generated constructor. Default value is 5000.
10773 @item max-fsm-thread-path-insns
10774 Maximum number of instructions to copy when duplicating blocks on a
10775 finite state automaton jump thread path. The default is 100.
10777 @item max-fsm-thread-length
10778 Maximum number of basic blocks on a finite state automaton jump thread
10779 path. The default is 10.
10781 @item max-fsm-thread-paths
10782 Maximum number of new jump thread paths to create for a finite state
10783 automaton. The default is 50.
10788 @node Preprocessor Options
10789 @section Options Controlling the Preprocessor
10790 @cindex preprocessor options
10791 @cindex options, preprocessor
10793 These options control the C preprocessor, which is run on each C source
10794 file before actual compilation.
10796 If you use the @option{-E} option, nothing is done except preprocessing.
10797 Some of these options make sense only together with @option{-E} because
10798 they cause the preprocessor output to be unsuitable for actual
10802 @item -Wp,@var{option}
10804 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10805 and pass @var{option} directly through to the preprocessor. If
10806 @var{option} contains commas, it is split into multiple options at the
10807 commas. However, many options are modified, translated or interpreted
10808 by the compiler driver before being passed to the preprocessor, and
10809 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10810 interface is undocumented and subject to change, so whenever possible
10811 you should avoid using @option{-Wp} and let the driver handle the
10814 @item -Xpreprocessor @var{option}
10815 @opindex Xpreprocessor
10816 Pass @var{option} as an option to the preprocessor. You can use this to
10817 supply system-specific preprocessor options that GCC does not
10820 If you want to pass an option that takes an argument, you must use
10821 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10823 @item -no-integrated-cpp
10824 @opindex no-integrated-cpp
10825 Perform preprocessing as a separate pass before compilation.
10826 By default, GCC performs preprocessing as an integrated part of
10827 input tokenization and parsing.
10828 If this option is provided, the appropriate language front end
10829 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10830 and Objective-C, respectively) is instead invoked twice,
10831 once for preprocessing only and once for actual compilation
10832 of the preprocessed input.
10833 This option may be useful in conjunction with the @option{-B} or
10834 @option{-wrapper} options to specify an alternate preprocessor or
10835 perform additional processing of the program source between
10836 normal preprocessing and compilation.
10839 @include cppopts.texi
10841 @node Assembler Options
10842 @section Passing Options to the Assembler
10844 @c prevent bad page break with this line
10845 You can pass options to the assembler.
10848 @item -Wa,@var{option}
10850 Pass @var{option} as an option to the assembler. If @var{option}
10851 contains commas, it is split into multiple options at the commas.
10853 @item -Xassembler @var{option}
10854 @opindex Xassembler
10855 Pass @var{option} as an option to the assembler. You can use this to
10856 supply system-specific assembler options that GCC does not
10859 If you want to pass an option that takes an argument, you must use
10860 @option{-Xassembler} twice, once for the option and once for the argument.
10865 @section Options for Linking
10866 @cindex link options
10867 @cindex options, linking
10869 These options come into play when the compiler links object files into
10870 an executable output file. They are meaningless if the compiler is
10871 not doing a link step.
10875 @item @var{object-file-name}
10876 A file name that does not end in a special recognized suffix is
10877 considered to name an object file or library. (Object files are
10878 distinguished from libraries by the linker according to the file
10879 contents.) If linking is done, these object files are used as input
10888 If any of these options is used, then the linker is not run, and
10889 object file names should not be used as arguments. @xref{Overall
10893 @item -l@var{library}
10894 @itemx -l @var{library}
10896 Search the library named @var{library} when linking. (The second
10897 alternative with the library as a separate argument is only for
10898 POSIX compliance and is not recommended.)
10900 It makes a difference where in the command you write this option; the
10901 linker searches and processes libraries and object files in the order they
10902 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10903 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10904 to functions in @samp{z}, those functions may not be loaded.
10906 The linker searches a standard list of directories for the library,
10907 which is actually a file named @file{lib@var{library}.a}. The linker
10908 then uses this file as if it had been specified precisely by name.
10910 The directories searched include several standard system directories
10911 plus any that you specify with @option{-L}.
10913 Normally the files found this way are library files---archive files
10914 whose members are object files. The linker handles an archive file by
10915 scanning through it for members which define symbols that have so far
10916 been referenced but not defined. But if the file that is found is an
10917 ordinary object file, it is linked in the usual fashion. The only
10918 difference between using an @option{-l} option and specifying a file name
10919 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10920 and searches several directories.
10924 You need this special case of the @option{-l} option in order to
10925 link an Objective-C or Objective-C++ program.
10927 @item -nostartfiles
10928 @opindex nostartfiles
10929 Do not use the standard system startup files when linking.
10930 The standard system libraries are used normally, unless @option{-nostdlib}
10931 or @option{-nodefaultlibs} is used.
10933 @item -nodefaultlibs
10934 @opindex nodefaultlibs
10935 Do not use the standard system libraries when linking.
10936 Only the libraries you specify are passed to the linker, and options
10937 specifying linkage of the system libraries, such as @option{-static-libgcc}
10938 or @option{-shared-libgcc}, are ignored.
10939 The standard startup files are used normally, unless @option{-nostartfiles}
10942 The compiler may generate calls to @code{memcmp},
10943 @code{memset}, @code{memcpy} and @code{memmove}.
10944 These entries are usually resolved by entries in
10945 libc. These entry points should be supplied through some other
10946 mechanism when this option is specified.
10950 Do not use the standard system startup files or libraries when linking.
10951 No startup files and only the libraries you specify are passed to
10952 the linker, and options specifying linkage of the system libraries, such as
10953 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
10955 The compiler may generate calls to @code{memcmp}, @code{memset},
10956 @code{memcpy} and @code{memmove}.
10957 These entries are usually resolved by entries in
10958 libc. These entry points should be supplied through some other
10959 mechanism when this option is specified.
10961 @cindex @option{-lgcc}, use with @option{-nostdlib}
10962 @cindex @option{-nostdlib} and unresolved references
10963 @cindex unresolved references and @option{-nostdlib}
10964 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10965 @cindex @option{-nodefaultlibs} and unresolved references
10966 @cindex unresolved references and @option{-nodefaultlibs}
10967 One of the standard libraries bypassed by @option{-nostdlib} and
10968 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10969 which GCC uses to overcome shortcomings of particular machines, or special
10970 needs for some languages.
10971 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10972 Collection (GCC) Internals},
10973 for more discussion of @file{libgcc.a}.)
10974 In most cases, you need @file{libgcc.a} even when you want to avoid
10975 other standard libraries. In other words, when you specify @option{-nostdlib}
10976 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10977 This ensures that you have no unresolved references to internal GCC
10978 library subroutines.
10979 (An example of such an internal subroutine is @code{__main}, used to ensure C++
10980 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10981 GNU Compiler Collection (GCC) Internals}.)
10985 Produce a position independent executable on targets that support it.
10986 For predictable results, you must also specify the same set of options
10987 used for compilation (@option{-fpie}, @option{-fPIE},
10988 or model suboptions) when you specify this linker option.
10992 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10993 that support it. This instructs the linker to add all symbols, not
10994 only used ones, to the dynamic symbol table. This option is needed
10995 for some uses of @code{dlopen} or to allow obtaining backtraces
10996 from within a program.
11000 Remove all symbol table and relocation information from the executable.
11004 On systems that support dynamic linking, this prevents linking with the shared
11005 libraries. On other systems, this option has no effect.
11009 Produce a shared object which can then be linked with other objects to
11010 form an executable. Not all systems support this option. For predictable
11011 results, you must also specify the same set of options used for compilation
11012 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11013 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11014 needs to build supplementary stub code for constructors to work. On
11015 multi-libbed systems, @samp{gcc -shared} must select the correct support
11016 libraries to link against. Failing to supply the correct flags may lead
11017 to subtle defects. Supplying them in cases where they are not necessary
11020 @item -shared-libgcc
11021 @itemx -static-libgcc
11022 @opindex shared-libgcc
11023 @opindex static-libgcc
11024 On systems that provide @file{libgcc} as a shared library, these options
11025 force the use of either the shared or static version, respectively.
11026 If no shared version of @file{libgcc} was built when the compiler was
11027 configured, these options have no effect.
11029 There are several situations in which an application should use the
11030 shared @file{libgcc} instead of the static version. The most common
11031 of these is when the application wishes to throw and catch exceptions
11032 across different shared libraries. In that case, each of the libraries
11033 as well as the application itself should use the shared @file{libgcc}.
11035 Therefore, the G++ and GCJ drivers automatically add
11036 @option{-shared-libgcc} whenever you build a shared library or a main
11037 executable, because C++ and Java programs typically use exceptions, so
11038 this is the right thing to do.
11040 If, instead, you use the GCC driver to create shared libraries, you may
11041 find that they are not always linked with the shared @file{libgcc}.
11042 If GCC finds, at its configuration time, that you have a non-GNU linker
11043 or a GNU linker that does not support option @option{--eh-frame-hdr},
11044 it links the shared version of @file{libgcc} into shared libraries
11045 by default. Otherwise, it takes advantage of the linker and optimizes
11046 away the linking with the shared version of @file{libgcc}, linking with
11047 the static version of libgcc by default. This allows exceptions to
11048 propagate through such shared libraries, without incurring relocation
11049 costs at library load time.
11051 However, if a library or main executable is supposed to throw or catch
11052 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11053 for the languages used in the program, or using the option
11054 @option{-shared-libgcc}, such that it is linked with the shared
11057 @item -static-libasan
11058 @opindex static-libasan
11059 When the @option{-fsanitize=address} option is used to link a program,
11060 the GCC driver automatically links against @option{libasan}. If
11061 @file{libasan} is available as a shared library, and the @option{-static}
11062 option is not used, then this links against the shared version of
11063 @file{libasan}. The @option{-static-libasan} option directs the GCC
11064 driver to link @file{libasan} statically, without necessarily linking
11065 other libraries statically.
11067 @item -static-libtsan
11068 @opindex static-libtsan
11069 When the @option{-fsanitize=thread} option is used to link a program,
11070 the GCC driver automatically links against @option{libtsan}. If
11071 @file{libtsan} is available as a shared library, and the @option{-static}
11072 option is not used, then this links against the shared version of
11073 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11074 driver to link @file{libtsan} statically, without necessarily linking
11075 other libraries statically.
11077 @item -static-liblsan
11078 @opindex static-liblsan
11079 When the @option{-fsanitize=leak} option is used to link a program,
11080 the GCC driver automatically links against @option{liblsan}. If
11081 @file{liblsan} is available as a shared library, and the @option{-static}
11082 option is not used, then this links against the shared version of
11083 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11084 driver to link @file{liblsan} statically, without necessarily linking
11085 other libraries statically.
11087 @item -static-libubsan
11088 @opindex static-libubsan
11089 When the @option{-fsanitize=undefined} option is used to link a program,
11090 the GCC driver automatically links against @option{libubsan}. If
11091 @file{libubsan} is available as a shared library, and the @option{-static}
11092 option is not used, then this links against the shared version of
11093 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11094 driver to link @file{libubsan} statically, without necessarily linking
11095 other libraries statically.
11097 @item -static-libstdc++
11098 @opindex static-libstdc++
11099 When the @command{g++} program is used to link a C++ program, it
11100 normally automatically links against @option{libstdc++}. If
11101 @file{libstdc++} is available as a shared library, and the
11102 @option{-static} option is not used, then this links against the
11103 shared version of @file{libstdc++}. That is normally fine. However, it
11104 is sometimes useful to freeze the version of @file{libstdc++} used by
11105 the program without going all the way to a fully static link. The
11106 @option{-static-libstdc++} option directs the @command{g++} driver to
11107 link @file{libstdc++} statically, without necessarily linking other
11108 libraries statically.
11112 Bind references to global symbols when building a shared object. Warn
11113 about any unresolved references (unless overridden by the link editor
11114 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11117 @item -T @var{script}
11119 @cindex linker script
11120 Use @var{script} as the linker script. This option is supported by most
11121 systems using the GNU linker. On some targets, such as bare-board
11122 targets without an operating system, the @option{-T} option may be required
11123 when linking to avoid references to undefined symbols.
11125 @item -Xlinker @var{option}
11127 Pass @var{option} as an option to the linker. You can use this to
11128 supply system-specific linker options that GCC does not recognize.
11130 If you want to pass an option that takes a separate argument, you must use
11131 @option{-Xlinker} twice, once for the option and once for the argument.
11132 For example, to pass @option{-assert definitions}, you must write
11133 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11134 @option{-Xlinker "-assert definitions"}, because this passes the entire
11135 string as a single argument, which is not what the linker expects.
11137 When using the GNU linker, it is usually more convenient to pass
11138 arguments to linker options using the @option{@var{option}=@var{value}}
11139 syntax than as separate arguments. For example, you can specify
11140 @option{-Xlinker -Map=output.map} rather than
11141 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11142 this syntax for command-line options.
11144 @item -Wl,@var{option}
11146 Pass @var{option} as an option to the linker. If @var{option} contains
11147 commas, it is split into multiple options at the commas. You can use this
11148 syntax to pass an argument to the option.
11149 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11150 linker. When using the GNU linker, you can also get the same effect with
11151 @option{-Wl,-Map=output.map}.
11153 @item -u @var{symbol}
11155 Pretend the symbol @var{symbol} is undefined, to force linking of
11156 library modules to define it. You can use @option{-u} multiple times with
11157 different symbols to force loading of additional library modules.
11159 @item -z @var{keyword}
11161 @option{-z} is passed directly on to the linker along with the keyword
11162 @var{keyword}. See the section in the documentation of your linker for
11163 permitted values and their meanings.
11166 @node Directory Options
11167 @section Options for Directory Search
11168 @cindex directory options
11169 @cindex options, directory search
11170 @cindex search path
11172 These options specify directories to search for header files, for
11173 libraries and for parts of the compiler:
11178 Add the directory @var{dir} to the head of the list of directories to be
11179 searched for header files. This can be used to override a system header
11180 file, substituting your own version, since these directories are
11181 searched before the system header file directories. However, you should
11182 not use this option to add directories that contain vendor-supplied
11183 system header files (use @option{-isystem} for that). If you use more than
11184 one @option{-I} option, the directories are scanned in left-to-right
11185 order; the standard system directories come after.
11187 If a standard system include directory, or a directory specified with
11188 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11189 option is ignored. The directory is still searched but as a
11190 system directory at its normal position in the system include chain.
11191 This is to ensure that GCC's procedure to fix buggy system headers and
11192 the ordering for the @code{include_next} directive are not inadvertently changed.
11193 If you really need to change the search order for system directories,
11194 use the @option{-nostdinc} and/or @option{-isystem} options.
11196 @item -iplugindir=@var{dir}
11197 @opindex iplugindir=
11198 Set the directory to search for plugins that are passed
11199 by @option{-fplugin=@var{name}} instead of
11200 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11201 to be used by the user, but only passed by the driver.
11203 @item -iquote@var{dir}
11205 Add the directory @var{dir} to the head of the list of directories to
11206 be searched for header files only for the case of @code{#include
11207 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11208 otherwise just like @option{-I}.
11212 Add directory @var{dir} to the list of directories to be searched
11215 @item -B@var{prefix}
11217 This option specifies where to find the executables, libraries,
11218 include files, and data files of the compiler itself.
11220 The compiler driver program runs one or more of the subprograms
11221 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11222 @var{prefix} as a prefix for each program it tries to run, both with and
11223 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11225 For each subprogram to be run, the compiler driver first tries the
11226 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11227 is not specified, the driver tries two standard prefixes,
11228 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11229 those results in a file name that is found, the unmodified program
11230 name is searched for using the directories specified in your
11231 @env{PATH} environment variable.
11233 The compiler checks to see if the path provided by the @option{-B}
11234 refers to a directory, and if necessary it adds a directory
11235 separator character at the end of the path.
11237 @option{-B} prefixes that effectively specify directory names also apply
11238 to libraries in the linker, because the compiler translates these
11239 options into @option{-L} options for the linker. They also apply to
11240 include files in the preprocessor, because the compiler translates these
11241 options into @option{-isystem} options for the preprocessor. In this case,
11242 the compiler appends @samp{include} to the prefix.
11244 The runtime support file @file{libgcc.a} can also be searched for using
11245 the @option{-B} prefix, if needed. If it is not found there, the two
11246 standard prefixes above are tried, and that is all. The file is left
11247 out of the link if it is not found by those means.
11249 Another way to specify a prefix much like the @option{-B} prefix is to use
11250 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11253 As a special kludge, if the path provided by @option{-B} is
11254 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11255 9, then it is replaced by @file{[dir/]include}. This is to help
11256 with boot-strapping the compiler.
11258 @item -specs=@var{file}
11260 Process @var{file} after the compiler reads in the standard @file{specs}
11261 file, in order to override the defaults which the @command{gcc} driver
11262 program uses when determining what switches to pass to @command{cc1},
11263 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11264 @option{-specs=@var{file}} can be specified on the command line, and they
11265 are processed in order, from left to right.
11267 @item --sysroot=@var{dir}
11269 Use @var{dir} as the logical root directory for headers and libraries.
11270 For example, if the compiler normally searches for headers in
11271 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11272 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11274 If you use both this option and the @option{-isysroot} option, then
11275 the @option{--sysroot} option applies to libraries, but the
11276 @option{-isysroot} option applies to header files.
11278 The GNU linker (beginning with version 2.16) has the necessary support
11279 for this option. If your linker does not support this option, the
11280 header file aspect of @option{--sysroot} still works, but the
11281 library aspect does not.
11283 @item --no-sysroot-suffix
11284 @opindex no-sysroot-suffix
11285 For some targets, a suffix is added to the root directory specified
11286 with @option{--sysroot}, depending on the other options used, so that
11287 headers may for example be found in
11288 @file{@var{dir}/@var{suffix}/usr/include} instead of
11289 @file{@var{dir}/usr/include}. This option disables the addition of
11294 This option has been deprecated. Please use @option{-iquote} instead for
11295 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
11296 Any directories you specify with @option{-I} options before the @option{-I-}
11297 option are searched only for the case of @code{#include "@var{file}"};
11298 they are not searched for @code{#include <@var{file}>}.
11300 If additional directories are specified with @option{-I} options after
11301 the @option{-I-}, these directories are searched for all @code{#include}
11302 directives. (Ordinarily @emph{all} @option{-I} directories are used
11305 In addition, the @option{-I-} option inhibits the use of the current
11306 directory (where the current input file came from) as the first search
11307 directory for @code{#include "@var{file}"}. There is no way to
11308 override this effect of @option{-I-}. With @option{-I.} you can specify
11309 searching the directory that is current when the compiler is
11310 invoked. That is not exactly the same as what the preprocessor does
11311 by default, but it is often satisfactory.
11313 @option{-I-} does not inhibit the use of the standard system directories
11314 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11321 @section Specifying subprocesses and the switches to pass to them
11324 @command{gcc} is a driver program. It performs its job by invoking a
11325 sequence of other programs to do the work of compiling, assembling and
11326 linking. GCC interprets its command-line parameters and uses these to
11327 deduce which programs it should invoke, and which command-line options
11328 it ought to place on their command lines. This behavior is controlled
11329 by @dfn{spec strings}. In most cases there is one spec string for each
11330 program that GCC can invoke, but a few programs have multiple spec
11331 strings to control their behavior. The spec strings built into GCC can
11332 be overridden by using the @option{-specs=} command-line switch to specify
11335 @dfn{Spec files} are plaintext files that are used to construct spec
11336 strings. They consist of a sequence of directives separated by blank
11337 lines. The type of directive is determined by the first non-whitespace
11338 character on the line, which can be one of the following:
11341 @item %@var{command}
11342 Issues a @var{command} to the spec file processor. The commands that can
11346 @item %include <@var{file}>
11347 @cindex @code{%include}
11348 Search for @var{file} and insert its text at the current point in the
11351 @item %include_noerr <@var{file}>
11352 @cindex @code{%include_noerr}
11353 Just like @samp{%include}, but do not generate an error message if the include
11354 file cannot be found.
11356 @item %rename @var{old_name} @var{new_name}
11357 @cindex @code{%rename}
11358 Rename the spec string @var{old_name} to @var{new_name}.
11362 @item *[@var{spec_name}]:
11363 This tells the compiler to create, override or delete the named spec
11364 string. All lines after this directive up to the next directive or
11365 blank line are considered to be the text for the spec string. If this
11366 results in an empty string then the spec is deleted. (Or, if the
11367 spec did not exist, then nothing happens.) Otherwise, if the spec
11368 does not currently exist a new spec is created. If the spec does
11369 exist then its contents are overridden by the text of this
11370 directive, unless the first character of that text is the @samp{+}
11371 character, in which case the text is appended to the spec.
11373 @item [@var{suffix}]:
11374 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11375 and up to the next directive or blank line are considered to make up the
11376 spec string for the indicated suffix. When the compiler encounters an
11377 input file with the named suffix, it processes the spec string in
11378 order to work out how to compile that file. For example:
11382 z-compile -input %i
11385 This says that any input file whose name ends in @samp{.ZZ} should be
11386 passed to the program @samp{z-compile}, which should be invoked with the
11387 command-line switch @option{-input} and with the result of performing the
11388 @samp{%i} substitution. (See below.)
11390 As an alternative to providing a spec string, the text following a
11391 suffix directive can be one of the following:
11394 @item @@@var{language}
11395 This says that the suffix is an alias for a known @var{language}. This is
11396 similar to using the @option{-x} command-line switch to GCC to specify a
11397 language explicitly. For example:
11404 Says that .ZZ files are, in fact, C++ source files.
11407 This causes an error messages saying:
11410 @var{name} compiler not installed on this system.
11414 GCC already has an extensive list of suffixes built into it.
11415 This directive adds an entry to the end of the list of suffixes, but
11416 since the list is searched from the end backwards, it is effectively
11417 possible to override earlier entries using this technique.
11421 GCC has the following spec strings built into it. Spec files can
11422 override these strings or create their own. Note that individual
11423 targets can also add their own spec strings to this list.
11426 asm Options to pass to the assembler
11427 asm_final Options to pass to the assembler post-processor
11428 cpp Options to pass to the C preprocessor
11429 cc1 Options to pass to the C compiler
11430 cc1plus Options to pass to the C++ compiler
11431 endfile Object files to include at the end of the link
11432 link Options to pass to the linker
11433 lib Libraries to include on the command line to the linker
11434 libgcc Decides which GCC support library to pass to the linker
11435 linker Sets the name of the linker
11436 predefines Defines to be passed to the C preprocessor
11437 signed_char Defines to pass to CPP to say whether @code{char} is signed
11439 startfile Object files to include at the start of the link
11442 Here is a small example of a spec file:
11445 %rename lib old_lib
11448 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11451 This example renames the spec called @samp{lib} to @samp{old_lib} and
11452 then overrides the previous definition of @samp{lib} with a new one.
11453 The new definition adds in some extra command-line options before
11454 including the text of the old definition.
11456 @dfn{Spec strings} are a list of command-line options to be passed to their
11457 corresponding program. In addition, the spec strings can contain
11458 @samp{%}-prefixed sequences to substitute variable text or to
11459 conditionally insert text into the command line. Using these constructs
11460 it is possible to generate quite complex command lines.
11462 Here is a table of all defined @samp{%}-sequences for spec
11463 strings. Note that spaces are not generated automatically around the
11464 results of expanding these sequences. Therefore you can concatenate them
11465 together or combine them with constant text in a single argument.
11469 Substitute one @samp{%} into the program name or argument.
11472 Substitute the name of the input file being processed.
11475 Substitute the basename of the input file being processed.
11476 This is the substring up to (and not including) the last period
11477 and not including the directory.
11480 This is the same as @samp{%b}, but include the file suffix (text after
11484 Marks the argument containing or following the @samp{%d} as a
11485 temporary file name, so that that file is deleted if GCC exits
11486 successfully. Unlike @samp{%g}, this contributes no text to the
11489 @item %g@var{suffix}
11490 Substitute a file name that has suffix @var{suffix} and is chosen
11491 once per compilation, and mark the argument in the same way as
11492 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11493 name is now chosen in a way that is hard to predict even when previously
11494 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11495 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11496 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11497 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11498 was simply substituted with a file name chosen once per compilation,
11499 without regard to any appended suffix (which was therefore treated
11500 just like ordinary text), making such attacks more likely to succeed.
11502 @item %u@var{suffix}
11503 Like @samp{%g}, but generates a new temporary file name
11504 each time it appears instead of once per compilation.
11506 @item %U@var{suffix}
11507 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11508 new one if there is no such last file name. In the absence of any
11509 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11510 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11511 involves the generation of two distinct file names, one
11512 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11513 simply substituted with a file name chosen for the previous @samp{%u},
11514 without regard to any appended suffix.
11516 @item %j@var{suffix}
11517 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11518 writable, and if @option{-save-temps} is not used;
11519 otherwise, substitute the name
11520 of a temporary file, just like @samp{%u}. This temporary file is not
11521 meant for communication between processes, but rather as a junk
11522 disposal mechanism.
11524 @item %|@var{suffix}
11525 @itemx %m@var{suffix}
11526 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11527 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11528 all. These are the two most common ways to instruct a program that it
11529 should read from standard input or write to standard output. If you
11530 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11531 construct: see for example @file{f/lang-specs.h}.
11533 @item %.@var{SUFFIX}
11534 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11535 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11536 terminated by the next space or %.
11539 Marks the argument containing or following the @samp{%w} as the
11540 designated output file of this compilation. This puts the argument
11541 into the sequence of arguments that @samp{%o} substitutes.
11544 Substitutes the names of all the output files, with spaces
11545 automatically placed around them. You should write spaces
11546 around the @samp{%o} as well or the results are undefined.
11547 @samp{%o} is for use in the specs for running the linker.
11548 Input files whose names have no recognized suffix are not compiled
11549 at all, but they are included among the output files, so they are
11553 Substitutes the suffix for object files. Note that this is
11554 handled specially when it immediately follows @samp{%g, %u, or %U},
11555 because of the need for those to form complete file names. The
11556 handling is such that @samp{%O} is treated exactly as if it had already
11557 been substituted, except that @samp{%g, %u, and %U} do not currently
11558 support additional @var{suffix} characters following @samp{%O} as they do
11559 following, for example, @samp{.o}.
11562 Substitutes the standard macro predefinitions for the
11563 current target machine. Use this when running @command{cpp}.
11566 Like @samp{%p}, but puts @samp{__} before and after the name of each
11567 predefined macro, except for macros that start with @samp{__} or with
11568 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11572 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11573 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11574 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11575 and @option{-imultilib} as necessary.
11578 Current argument is the name of a library or startup file of some sort.
11579 Search for that file in a standard list of directories and substitute
11580 the full name found. The current working directory is included in the
11581 list of directories scanned.
11584 Current argument is the name of a linker script. Search for that file
11585 in the current list of directories to scan for libraries. If the file
11586 is located insert a @option{--script} option into the command line
11587 followed by the full path name found. If the file is not found then
11588 generate an error message. Note: the current working directory is not
11592 Print @var{str} as an error message. @var{str} is terminated by a newline.
11593 Use this when inconsistent options are detected.
11595 @item %(@var{name})
11596 Substitute the contents of spec string @var{name} at this point.
11598 @item %x@{@var{option}@}
11599 Accumulate an option for @samp{%X}.
11602 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11606 Output the accumulated assembler options specified by @option{-Wa}.
11609 Output the accumulated preprocessor options specified by @option{-Wp}.
11612 Process the @code{asm} spec. This is used to compute the
11613 switches to be passed to the assembler.
11616 Process the @code{asm_final} spec. This is a spec string for
11617 passing switches to an assembler post-processor, if such a program is
11621 Process the @code{link} spec. This is the spec for computing the
11622 command line passed to the linker. Typically it makes use of the
11623 @samp{%L %G %S %D and %E} sequences.
11626 Dump out a @option{-L} option for each directory that GCC believes might
11627 contain startup files. If the target supports multilibs then the
11628 current multilib directory is prepended to each of these paths.
11631 Process the @code{lib} spec. This is a spec string for deciding which
11632 libraries are included on the command line to the linker.
11635 Process the @code{libgcc} spec. This is a spec string for deciding
11636 which GCC support library is included on the command line to the linker.
11639 Process the @code{startfile} spec. This is a spec for deciding which
11640 object files are the first ones passed to the linker. Typically
11641 this might be a file named @file{crt0.o}.
11644 Process the @code{endfile} spec. This is a spec string that specifies
11645 the last object files that are passed to the linker.
11648 Process the @code{cpp} spec. This is used to construct the arguments
11649 to be passed to the C preprocessor.
11652 Process the @code{cc1} spec. This is used to construct the options to be
11653 passed to the actual C compiler (@command{cc1}).
11656 Process the @code{cc1plus} spec. This is used to construct the options to be
11657 passed to the actual C++ compiler (@command{cc1plus}).
11660 Substitute the variable part of a matched option. See below.
11661 Note that each comma in the substituted string is replaced by
11665 Remove all occurrences of @code{-S} from the command line. Note---this
11666 command is position dependent. @samp{%} commands in the spec string
11667 before this one see @code{-S}, @samp{%} commands in the spec string
11668 after this one do not.
11670 @item %:@var{function}(@var{args})
11671 Call the named function @var{function}, passing it @var{args}.
11672 @var{args} is first processed as a nested spec string, then split
11673 into an argument vector in the usual fashion. The function returns
11674 a string which is processed as if it had appeared literally as part
11675 of the current spec.
11677 The following built-in spec functions are provided:
11680 @item @code{getenv}
11681 The @code{getenv} spec function takes two arguments: an environment
11682 variable name and a string. If the environment variable is not
11683 defined, a fatal error is issued. Otherwise, the return value is the
11684 value of the environment variable concatenated with the string. For
11685 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11688 %:getenv(TOPDIR /include)
11691 expands to @file{/path/to/top/include}.
11693 @item @code{if-exists}
11694 The @code{if-exists} spec function takes one argument, an absolute
11695 pathname to a file. If the file exists, @code{if-exists} returns the
11696 pathname. Here is a small example of its usage:
11700 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11703 @item @code{if-exists-else}
11704 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11705 spec function, except that it takes two arguments. The first argument is
11706 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11707 returns the pathname. If it does not exist, it returns the second argument.
11708 This way, @code{if-exists-else} can be used to select one file or another,
11709 based on the existence of the first. Here is a small example of its usage:
11713 crt0%O%s %:if-exists(crti%O%s) \
11714 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11717 @item @code{replace-outfile}
11718 The @code{replace-outfile} spec function takes two arguments. It looks for the
11719 first argument in the outfiles array and replaces it with the second argument. Here
11720 is a small example of its usage:
11723 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11726 @item @code{remove-outfile}
11727 The @code{remove-outfile} spec function takes one argument. It looks for the
11728 first argument in the outfiles array and removes it. Here is a small example
11732 %:remove-outfile(-lm)
11735 @item @code{pass-through-libs}
11736 The @code{pass-through-libs} spec function takes any number of arguments. It
11737 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11738 assumes are the names of linker input library archive files) and returns a
11739 result containing all the found arguments each prepended by
11740 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11741 intended to be passed to the LTO linker plugin.
11744 %:pass-through-libs(%G %L %G)
11747 @item @code{print-asm-header}
11748 The @code{print-asm-header} function takes no arguments and simply
11749 prints a banner like:
11755 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11758 It is used to separate compiler options from assembler options
11759 in the @option{--target-help} output.
11762 @item %@{@code{S}@}
11763 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11764 If that switch is not specified, this substitutes nothing. Note that
11765 the leading dash is omitted when specifying this option, and it is
11766 automatically inserted if the substitution is performed. Thus the spec
11767 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11768 and outputs the command-line option @option{-foo}.
11770 @item %W@{@code{S}@}
11771 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11772 deleted on failure.
11774 @item %@{@code{S}*@}
11775 Substitutes all the switches specified to GCC whose names start
11776 with @code{-S}, but which also take an argument. This is used for
11777 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11778 GCC considers @option{-o foo} as being
11779 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11780 text, including the space. Thus two arguments are generated.
11782 @item %@{@code{S}*&@code{T}*@}
11783 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11784 (the order of @code{S} and @code{T} in the spec is not significant).
11785 There can be any number of ampersand-separated variables; for each the
11786 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11788 @item %@{@code{S}:@code{X}@}
11789 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11791 @item %@{!@code{S}:@code{X}@}
11792 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11794 @item %@{@code{S}*:@code{X}@}
11795 Substitutes @code{X} if one or more switches whose names start with
11796 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11797 once, no matter how many such switches appeared. However, if @code{%*}
11798 appears somewhere in @code{X}, then @code{X} is substituted once
11799 for each matching switch, with the @code{%*} replaced by the part of
11800 that switch matching the @code{*}.
11802 If @code{%*} appears as the last part of a spec sequence then a space
11803 is added after the end of the last substitution. If there is more
11804 text in the sequence, however, then a space is not generated. This
11805 allows the @code{%*} substitution to be used as part of a larger
11806 string. For example, a spec string like this:
11809 %@{mcu=*:--script=%*/memory.ld@}
11813 when matching an option like @option{-mcu=newchip} produces:
11816 --script=newchip/memory.ld
11819 @item %@{.@code{S}:@code{X}@}
11820 Substitutes @code{X}, if processing a file with suffix @code{S}.
11822 @item %@{!.@code{S}:@code{X}@}
11823 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
11825 @item %@{,@code{S}:@code{X}@}
11826 Substitutes @code{X}, if processing a file for language @code{S}.
11828 @item %@{!,@code{S}:@code{X}@}
11829 Substitutes @code{X}, if not processing a file for language @code{S}.
11831 @item %@{@code{S}|@code{P}:@code{X}@}
11832 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
11833 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
11834 @code{*} sequences as well, although they have a stronger binding than
11835 the @samp{|}. If @code{%*} appears in @code{X}, all of the
11836 alternatives must be starred, and only the first matching alternative
11839 For example, a spec string like this:
11842 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
11846 outputs the following command-line options from the following input
11847 command-line options:
11852 -d fred.c -foo -baz -boggle
11853 -d jim.d -bar -baz -boggle
11856 @item %@{S:X; T:Y; :D@}
11858 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
11859 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
11860 be as many clauses as you need. This may be combined with @code{.},
11861 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
11866 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
11867 construct may contain other nested @samp{%} constructs or spaces, or
11868 even newlines. They are processed as usual, as described above.
11869 Trailing white space in @code{X} is ignored. White space may also
11870 appear anywhere on the left side of the colon in these constructs,
11871 except between @code{.} or @code{*} and the corresponding word.
11873 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
11874 handled specifically in these constructs. If another value of
11875 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
11876 @option{-W} switch is found later in the command line, the earlier
11877 switch value is ignored, except with @{@code{S}*@} where @code{S} is
11878 just one letter, which passes all matching options.
11880 The character @samp{|} at the beginning of the predicate text is used to
11881 indicate that a command should be piped to the following command, but
11882 only if @option{-pipe} is specified.
11884 It is built into GCC which switches take arguments and which do not.
11885 (You might think it would be useful to generalize this to allow each
11886 compiler's spec to say which switches take arguments. But this cannot
11887 be done in a consistent fashion. GCC cannot even decide which input
11888 files have been specified without knowing which switches take arguments,
11889 and it must know which input files to compile in order to tell which
11892 GCC also knows implicitly that arguments starting in @option{-l} are to be
11893 treated as compiler output files, and passed to the linker in their
11894 proper position among the other output files.
11896 @c man begin OPTIONS
11898 @node Target Options
11899 @section Specifying Target Machine and Compiler Version
11900 @cindex target options
11901 @cindex cross compiling
11902 @cindex specifying machine version
11903 @cindex specifying compiler version and target machine
11904 @cindex compiler version, specifying
11905 @cindex target machine, specifying
11907 The usual way to run GCC is to run the executable called @command{gcc}, or
11908 @command{@var{machine}-gcc} when cross-compiling, or
11909 @command{@var{machine}-gcc-@var{version}} to run a version other than the
11910 one that was installed last.
11912 @node Submodel Options
11913 @section Hardware Models and Configurations
11914 @cindex submodel options
11915 @cindex specifying hardware config
11916 @cindex hardware models and configurations, specifying
11917 @cindex machine dependent options
11919 Each target machine types can have its own
11920 special options, starting with @samp{-m}, to choose among various
11921 hardware models or configurations---for example, 68010 vs 68020,
11922 floating coprocessor or none. A single installed version of the
11923 compiler can compile for any model or configuration, according to the
11926 Some configurations of the compiler also support additional special
11927 options, usually for compatibility with other compilers on the same
11930 @c This list is ordered alphanumerically by subsection name.
11931 @c It should be the same order and spelling as these options are listed
11932 @c in Machine Dependent Options
11935 * AArch64 Options::
11936 * Adapteva Epiphany Options::
11940 * Blackfin Options::
11945 * DEC Alpha Options::
11948 * GNU/Linux Options::
11951 * i386 and x86-64 Options::
11952 * i386 and x86-64 Windows Options::
11960 * MicroBlaze Options::
11963 * MN10300 Options::
11967 * Nios II Options::
11969 * picoChip Options::
11970 * PowerPC Options::
11972 * RS/6000 and PowerPC Options::
11974 * S/390 and zSeries Options::
11977 * Solaris 2 Options::
11980 * System V Options::
11981 * TILE-Gx Options::
11982 * TILEPro Options::
11987 * VxWorks Options::
11989 * Xstormy16 Options::
11991 * zSeries Options::
11994 @node AArch64 Options
11995 @subsection AArch64 Options
11996 @cindex AArch64 Options
11998 These options are defined for AArch64 implementations:
12002 @item -mabi=@var{name}
12004 Generate code for the specified data model. Permissible values
12005 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12006 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12007 but long int and pointer are 64-bit.
12009 The default depends on the specific target configuration. Note that
12010 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12011 entire program with the same ABI, and link with a compatible set of libraries.
12014 @opindex mbig-endian
12015 Generate big-endian code. This is the default when GCC is configured for an
12016 @samp{aarch64_be-*-*} target.
12018 @item -mgeneral-regs-only
12019 @opindex mgeneral-regs-only
12020 Generate code which uses only the general registers.
12022 @item -mlittle-endian
12023 @opindex mlittle-endian
12024 Generate little-endian code. This is the default when GCC is configured for an
12025 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12027 @item -mcmodel=tiny
12028 @opindex mcmodel=tiny
12029 Generate code for the tiny code model. The program and its statically defined
12030 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12031 be statically or dynamically linked. This model is not fully implemented and
12032 mostly treated as @samp{small}.
12034 @item -mcmodel=small
12035 @opindex mcmodel=small
12036 Generate code for the small code model. The program and its statically defined
12037 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12038 be statically or dynamically linked. This is the default code model.
12040 @item -mcmodel=large
12041 @opindex mcmodel=large
12042 Generate code for the large code model. This makes no assumptions about
12043 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12044 statically linked only.
12046 @item -mstrict-align
12047 @opindex mstrict-align
12048 Do not assume that unaligned memory references are handled by the system.
12050 @item -momit-leaf-frame-pointer
12051 @itemx -mno-omit-leaf-frame-pointer
12052 @opindex momit-leaf-frame-pointer
12053 @opindex mno-omit-leaf-frame-pointer
12054 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12057 @item -mtls-dialect=desc
12058 @opindex mtls-dialect=desc
12059 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12060 of TLS variables. This is the default.
12062 @item -mtls-dialect=traditional
12063 @opindex mtls-dialect=traditional
12064 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12067 @item -mfix-cortex-a53-835769
12068 @itemx -mno-fix-cortex-a53-835769
12069 @opindex mfix-cortex-a53-835769
12070 @opindex mno-fix-cortex-a53-835769
12071 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12072 This involves inserting a NOP instruction between memory instructions and
12073 64-bit integer multiply-accumulate instructions.
12075 @item -march=@var{name}
12077 Specify the name of the target architecture, optionally suffixed by one or
12078 more feature modifiers. This option has the form
12079 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12080 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
12081 values for @var{feature} are documented in the sub-section below.
12083 Where conflicting feature modifiers are specified, the right-most feature is
12086 GCC uses this name to determine what kind of instructions it can emit when
12087 generating assembly code.
12089 Where @option{-march} is specified without either of @option{-mtune}
12090 or @option{-mcpu} also being specified, the code is tuned to perform
12091 well across a range of target processors implementing the target
12094 @item -mtune=@var{name}
12096 Specify the name of the target processor for which GCC should tune the
12097 performance of the code. Permissible values for this option are:
12098 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{thunderx}.
12100 Additionally, this option can specify that GCC should tune the performance
12101 of the code for a big.LITTLE system. The only permissible value is
12102 @samp{cortex-a57.cortex-a53}.
12104 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12105 are specified, the code is tuned to perform well across a range
12106 of target processors.
12108 This option cannot be suffixed by feature modifiers.
12110 @item -mcpu=@var{name}
12112 Specify the name of the target processor, optionally suffixed by one or more
12113 feature modifiers. This option has the form
12114 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12115 permissible values for @var{cpu} are the same as those available for
12118 The permissible values for @var{feature} are documented in the sub-section
12121 Where conflicting feature modifiers are specified, the right-most feature is
12124 GCC uses this name to determine what kind of instructions it can emit when
12125 generating assembly code (as if by @option{-march}) and to determine
12126 the target processor for which to tune for performance (as if
12127 by @option{-mtune}). Where this option is used in conjunction
12128 with @option{-march} or @option{-mtune}, those options take precedence
12129 over the appropriate part of this option.
12132 @subsubsection @option{-march} and @option{-mcpu} feature modifiers
12133 @cindex @option{-march} feature modifiers
12134 @cindex @option{-mcpu} feature modifiers
12135 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
12140 Enable CRC extension.
12142 Enable Crypto extension. This implies Advanced SIMD is enabled.
12144 Enable floating-point instructions.
12146 Enable Advanced SIMD instructions. This implies floating-point instructions
12147 are enabled. This is the default for all current possible values for options
12148 @option{-march} and @option{-mcpu=}.
12151 @node Adapteva Epiphany Options
12152 @subsection Adapteva Epiphany Options
12154 These @samp{-m} options are defined for Adapteva Epiphany:
12157 @item -mhalf-reg-file
12158 @opindex mhalf-reg-file
12159 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12160 That allows code to run on hardware variants that lack these registers.
12162 @item -mprefer-short-insn-regs
12163 @opindex mprefer-short-insn-regs
12164 Preferrentially allocate registers that allow short instruction generation.
12165 This can result in increased instruction count, so this may either reduce or
12166 increase overall code size.
12168 @item -mbranch-cost=@var{num}
12169 @opindex mbranch-cost
12170 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12171 This cost is only a heuristic and is not guaranteed to produce
12172 consistent results across releases.
12176 Enable the generation of conditional moves.
12178 @item -mnops=@var{num}
12180 Emit @var{num} NOPs before every other generated instruction.
12182 @item -mno-soft-cmpsf
12183 @opindex mno-soft-cmpsf
12184 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12185 and test the flags. This is faster than a software comparison, but can
12186 get incorrect results in the presence of NaNs, or when two different small
12187 numbers are compared such that their difference is calculated as zero.
12188 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12189 software comparisons.
12191 @item -mstack-offset=@var{num}
12192 @opindex mstack-offset
12193 Set the offset between the top of the stack and the stack pointer.
12194 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12195 can be used by leaf functions without stack allocation.
12196 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12197 Note also that this option changes the ABI; compiling a program with a
12198 different stack offset than the libraries have been compiled with
12199 generally does not work.
12200 This option can be useful if you want to evaluate if a different stack
12201 offset would give you better code, but to actually use a different stack
12202 offset to build working programs, it is recommended to configure the
12203 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12205 @item -mno-round-nearest
12206 @opindex mno-round-nearest
12207 Make the scheduler assume that the rounding mode has been set to
12208 truncating. The default is @option{-mround-nearest}.
12211 @opindex mlong-calls
12212 If not otherwise specified by an attribute, assume all calls might be beyond
12213 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12214 function address into a register before performing a (otherwise direct) call.
12215 This is the default.
12217 @item -mshort-calls
12218 @opindex short-calls
12219 If not otherwise specified by an attribute, assume all direct calls are
12220 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12221 for direct calls. The default is @option{-mlong-calls}.
12225 Assume addresses can be loaded as 16-bit unsigned values. This does not
12226 apply to function addresses for which @option{-mlong-calls} semantics
12229 @item -mfp-mode=@var{mode}
12231 Set the prevailing mode of the floating-point unit.
12232 This determines the floating-point mode that is provided and expected
12233 at function call and return time. Making this mode match the mode you
12234 predominantly need at function start can make your programs smaller and
12235 faster by avoiding unnecessary mode switches.
12237 @var{mode} can be set to one the following values:
12241 Any mode at function entry is valid, and retained or restored when
12242 the function returns, and when it calls other functions.
12243 This mode is useful for compiling libraries or other compilation units
12244 you might want to incorporate into different programs with different
12245 prevailing FPU modes, and the convenience of being able to use a single
12246 object file outweighs the size and speed overhead for any extra
12247 mode switching that might be needed, compared with what would be needed
12248 with a more specific choice of prevailing FPU mode.
12251 This is the mode used for floating-point calculations with
12252 truncating (i.e.@: round towards zero) rounding mode. That includes
12253 conversion from floating point to integer.
12255 @item round-nearest
12256 This is the mode used for floating-point calculations with
12257 round-to-nearest-or-even rounding mode.
12260 This is the mode used to perform integer calculations in the FPU, e.g.@:
12261 integer multiply, or integer multiply-and-accumulate.
12264 The default is @option{-mfp-mode=caller}
12266 @item -mnosplit-lohi
12267 @itemx -mno-postinc
12268 @itemx -mno-postmodify
12269 @opindex mnosplit-lohi
12270 @opindex mno-postinc
12271 @opindex mno-postmodify
12272 Code generation tweaks that disable, respectively, splitting of 32-bit
12273 loads, generation of post-increment addresses, and generation of
12274 post-modify addresses. The defaults are @option{msplit-lohi},
12275 @option{-mpost-inc}, and @option{-mpost-modify}.
12277 @item -mnovect-double
12278 @opindex mno-vect-double
12279 Change the preferred SIMD mode to SImode. The default is
12280 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12282 @item -max-vect-align=@var{num}
12283 @opindex max-vect-align
12284 The maximum alignment for SIMD vector mode types.
12285 @var{num} may be 4 or 8. The default is 8.
12286 Note that this is an ABI change, even though many library function
12287 interfaces are unaffected if they don't use SIMD vector modes
12288 in places that affect size and/or alignment of relevant types.
12290 @item -msplit-vecmove-early
12291 @opindex msplit-vecmove-early
12292 Split vector moves into single word moves before reload. In theory this
12293 can give better register allocation, but so far the reverse seems to be
12294 generally the case.
12296 @item -m1reg-@var{reg}
12298 Specify a register to hold the constant @minus{}1, which makes loading small negative
12299 constants and certain bitmasks faster.
12300 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12301 which specify use of that register as a fixed register,
12302 and @samp{none}, which means that no register is used for this
12303 purpose. The default is @option{-m1reg-none}.
12308 @subsection ARC Options
12309 @cindex ARC options
12311 The following options control the architecture variant for which code
12314 @c architecture variants
12317 @item -mbarrel-shifter
12318 @opindex mbarrel-shifter
12319 Generate instructions supported by barrel shifter. This is the default
12320 unless @option{-mcpu=ARC601} is in effect.
12322 @item -mcpu=@var{cpu}
12324 Set architecture type, register usage, and instruction scheduling
12325 parameters for @var{cpu}. There are also shortcut alias options
12326 available for backward compatibility and convenience. Supported
12327 values for @var{cpu} are
12333 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12337 Compile for ARC601. Alias: @option{-mARC601}.
12342 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12343 This is the default when configured with @option{--with-cpu=arc700}@.
12348 @itemx -mdpfp-compact
12349 @opindex mdpfp-compact
12350 FPX: Generate Double Precision FPX instructions, tuned for the compact
12354 @opindex mdpfp-fast
12355 FPX: Generate Double Precision FPX instructions, tuned for the fast
12358 @item -mno-dpfp-lrsr
12359 @opindex mno-dpfp-lrsr
12360 Disable LR and SR instructions from using FPX extension aux registers.
12364 Generate Extended arithmetic instructions. Currently only
12365 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12366 supported. This is always enabled for @option{-mcpu=ARC700}.
12370 Do not generate mpy instructions for ARC700.
12374 Generate 32x16 bit multiply and mac instructions.
12378 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12382 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12387 @itemx -mspfp-compact
12388 @opindex mspfp-compact
12389 FPX: Generate Single Precision FPX instructions, tuned for the compact
12393 @opindex mspfp-fast
12394 FPX: Generate Single Precision FPX instructions, tuned for the fast
12399 Enable generation of ARC SIMD instructions via target-specific
12400 builtins. Only valid for @option{-mcpu=ARC700}.
12403 @opindex msoft-float
12404 This option ignored; it is provided for compatibility purposes only.
12405 Software floating point code is emitted by default, and this default
12406 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12407 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12408 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12412 Generate swap instructions.
12416 The following options are passed through to the assembler, and also
12417 define preprocessor macro symbols.
12419 @c Flags used by the assembler, but for which we define preprocessor
12420 @c macro symbols as well.
12423 @opindex mdsp-packa
12424 Passed down to the assembler to enable the DSP Pack A extensions.
12425 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12429 Passed down to the assembler to enable the dual viterbi butterfly
12430 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12432 @c ARC700 4.10 extension instruction
12435 Passed down to the assembler to enable the Locked Load/Store
12436 Conditional extension. Also sets the preprocessor symbol
12441 Passed down to the assembler. Also sets the preprocessor symbol
12442 @code{__Xxmac_d16}.
12446 Passed down to the assembler. Also sets the preprocessor symbol
12449 @c ARC700 4.10 extension instruction
12452 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12453 extension instruction. Also sets the preprocessor symbol
12456 @c ARC700 4.10 extension instruction
12459 Passed down to the assembler to enable the swap byte ordering
12460 extension instruction. Also sets the preprocessor symbol
12464 @opindex mtelephony
12465 Passed down to the assembler to enable dual and single operand
12466 instructions for telephony. Also sets the preprocessor symbol
12467 @code{__Xtelephony}.
12471 Passed down to the assembler to enable the XY Memory extension. Also
12472 sets the preprocessor symbol @code{__Xxy}.
12476 The following options control how the assembly code is annotated:
12478 @c Assembly annotation options
12482 Annotate assembler instructions with estimated addresses.
12484 @item -mannotate-align
12485 @opindex mannotate-align
12486 Explain what alignment considerations lead to the decision to make an
12487 instruction short or long.
12491 The following options are passed through to the linker:
12493 @c options passed through to the linker
12497 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12498 This option is enabled by default in tool chains built for
12499 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12500 when profiling is not requested.
12502 @item -marclinux_prof
12503 @opindex marclinux_prof
12504 Passed through to the linker, to specify use of the
12505 @code{arclinux_prof} emulation. This option is enabled by default in
12506 tool chains built for @w{@code{arc-linux-uclibc}} and
12507 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12511 The following options control the semantics of generated code:
12513 @c semantically relevant code generation options
12515 @item -mepilogue-cfi
12516 @opindex mepilogue-cfi
12517 Enable generation of call frame information for epilogues.
12519 @item -mno-epilogue-cfi
12520 @opindex mno-epilogue-cfi
12521 Disable generation of call frame information for epilogues.
12524 @opindex mlong-calls
12525 Generate call insns as register indirect calls, thus providing access
12526 to the full 32-bit address range.
12528 @item -mmedium-calls
12529 @opindex mmedium-calls
12530 Don't use less than 25 bit addressing range for calls, which is the
12531 offset available for an unconditional branch-and-link
12532 instruction. Conditional execution of function calls is suppressed, to
12533 allow use of the 25-bit range, rather than the 21-bit range with
12534 conditional branch-and-link. This is the default for tool chains built
12535 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12539 Do not generate sdata references. This is the default for tool chains
12540 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12544 @opindex mucb-mcount
12545 Instrument with mcount calls as used in UCB code. I.e. do the
12546 counting in the callee, not the caller. By default ARC instrumentation
12547 counts in the caller.
12549 @item -mvolatile-cache
12550 @opindex mvolatile-cache
12551 Use ordinarily cached memory accesses for volatile references. This is the
12554 @item -mno-volatile-cache
12555 @opindex mno-volatile-cache
12556 Enable cache bypass for volatile references.
12560 The following options fine tune code generation:
12561 @c code generation tuning options
12564 @opindex malign-call
12565 Do alignment optimizations for call instructions.
12567 @item -mauto-modify-reg
12568 @opindex mauto-modify-reg
12569 Enable the use of pre/post modify with register displacement.
12571 @item -mbbit-peephole
12572 @opindex mbbit-peephole
12573 Enable bbit peephole2.
12577 This option disables a target-specific pass in @file{arc_reorg} to
12578 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12579 generation driven by the combiner pass.
12581 @item -mcase-vector-pcrel
12582 @opindex mcase-vector-pcrel
12583 Use pc-relative switch case tables - this enables case table shortening.
12584 This is the default for @option{-Os}.
12586 @item -mcompact-casesi
12587 @opindex mcompact-casesi
12588 Enable compact casesi pattern.
12589 This is the default for @option{-Os}.
12591 @item -mno-cond-exec
12592 @opindex mno-cond-exec
12593 Disable ARCompact specific pass to generate conditional execution instructions.
12594 Due to delay slot scheduling and interactions between operand numbers,
12595 literal sizes, instruction lengths, and the support for conditional execution,
12596 the target-independent pass to generate conditional execution is often lacking,
12597 so the ARC port has kept a special pass around that tries to find more
12598 conditional execution generating opportunities after register allocation,
12599 branch shortening, and delay slot scheduling have been done. This pass
12600 generally, but not always, improves performance and code size, at the cost of
12601 extra compilation time, which is why there is an option to switch it off.
12602 If you have a problem with call instructions exceeding their allowable
12603 offset range because they are conditionalized, you should consider using
12604 @option{-mmedium-calls} instead.
12606 @item -mearly-cbranchsi
12607 @opindex mearly-cbranchsi
12608 Enable pre-reload use of the cbranchsi pattern.
12610 @item -mexpand-adddi
12611 @opindex mexpand-adddi
12612 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12613 @code{add.f}, @code{adc} etc.
12615 @item -mindexed-loads
12616 @opindex mindexed-loads
12617 Enable the use of indexed loads. This can be problematic because some
12618 optimizers then assume that indexed stores exist, which is not
12623 Enable Local Register Allocation. This is still experimental for ARC,
12624 so by default the compiler uses standard reload
12625 (i.e. @option{-mno-lra}).
12627 @item -mlra-priority-none
12628 @opindex mlra-priority-none
12629 Don't indicate any priority for target registers.
12631 @item -mlra-priority-compact
12632 @opindex mlra-priority-compact
12633 Indicate target register priority for r0..r3 / r12..r15.
12635 @item -mlra-priority-noncompact
12636 @opindex mlra-priority-noncompact
12637 Reduce target regsiter priority for r0..r3 / r12..r15.
12639 @item -mno-millicode
12640 @opindex mno-millicode
12641 When optimizing for size (using @option{-Os}), prologues and epilogues
12642 that have to save or restore a large number of registers are often
12643 shortened by using call to a special function in libgcc; this is
12644 referred to as a @emph{millicode} call. As these calls can pose
12645 performance issues, and/or cause linking issues when linking in a
12646 nonstandard way, this option is provided to turn off millicode call
12650 @opindex mmixed-code
12651 Tweak register allocation to help 16-bit instruction generation.
12652 This generally has the effect of decreasing the average instruction size
12653 while increasing the instruction count.
12657 Enable 'q' instruction alternatives.
12658 This is the default for @option{-Os}.
12662 Enable Rcq constraint handling - most short code generation depends on this.
12663 This is the default.
12667 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12668 This is the default.
12670 @item -msize-level=@var{level}
12671 @opindex msize-level
12672 Fine-tune size optimization with regards to instruction lengths and alignment.
12673 The recognized values for @var{level} are:
12676 No size optimization. This level is deprecated and treated like @samp{1}.
12679 Short instructions are used opportunistically.
12682 In addition, alignment of loops and of code after barriers are dropped.
12685 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12689 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12690 the behavior when this is not set is equivalent to level @samp{1}.
12692 @item -mtune=@var{cpu}
12694 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12695 by @option{-mcpu=}.
12697 Supported values for @var{cpu} are
12701 Tune for ARC600 cpu.
12704 Tune for ARC601 cpu.
12707 Tune for ARC700 cpu with standard multiplier block.
12710 Tune for ARC700 cpu with XMAC block.
12713 Tune for ARC725D cpu.
12716 Tune for ARC750D cpu.
12720 @item -mmultcost=@var{num}
12722 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12723 normal instruction.
12725 @item -munalign-prob-threshold=@var{probability}
12726 @opindex munalign-prob-threshold
12727 Set probability threshold for unaligning branches.
12728 When tuning for @samp{ARC700} and optimizing for speed, branches without
12729 filled delay slot are preferably emitted unaligned and long, unless
12730 profiling indicates that the probability for the branch to be taken
12731 is below @var{probability}. @xref{Cross-profiling}.
12732 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12736 The following options are maintained for backward compatibility, but
12737 are now deprecated and will be removed in a future release:
12739 @c Deprecated options
12747 @opindex mbig-endian
12750 Compile code for big endian targets. Use of these options is now
12751 deprecated. Users wanting big-endian code, should use the
12752 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12753 building the tool chain, for which big-endian is the default.
12755 @item -mlittle-endian
12756 @opindex mlittle-endian
12759 Compile code for little endian targets. Use of these options is now
12760 deprecated. Users wanting little-endian code should use the
12761 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12762 building the tool chain, for which little-endian is the default.
12764 @item -mbarrel_shifter
12765 @opindex mbarrel_shifter
12766 Replaced by @option{-mbarrel-shifter}.
12768 @item -mdpfp_compact
12769 @opindex mdpfp_compact
12770 Replaced by @option{-mdpfp-compact}.
12773 @opindex mdpfp_fast
12774 Replaced by @option{-mdpfp-fast}.
12777 @opindex mdsp_packa
12778 Replaced by @option{-mdsp-packa}.
12782 Replaced by @option{-mea}.
12786 Replaced by @option{-mmac-24}.
12790 Replaced by @option{-mmac-d16}.
12792 @item -mspfp_compact
12793 @opindex mspfp_compact
12794 Replaced by @option{-mspfp-compact}.
12797 @opindex mspfp_fast
12798 Replaced by @option{-mspfp-fast}.
12800 @item -mtune=@var{cpu}
12802 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
12803 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
12804 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
12806 @item -multcost=@var{num}
12808 Replaced by @option{-mmultcost}.
12813 @subsection ARM Options
12814 @cindex ARM options
12816 These @samp{-m} options are defined for the ARM port:
12819 @item -mabi=@var{name}
12821 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
12822 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
12825 @opindex mapcs-frame
12826 Generate a stack frame that is compliant with the ARM Procedure Call
12827 Standard for all functions, even if this is not strictly necessary for
12828 correct execution of the code. Specifying @option{-fomit-frame-pointer}
12829 with this option causes the stack frames not to be generated for
12830 leaf functions. The default is @option{-mno-apcs-frame}.
12834 This is a synonym for @option{-mapcs-frame}.
12837 @c not currently implemented
12838 @item -mapcs-stack-check
12839 @opindex mapcs-stack-check
12840 Generate code to check the amount of stack space available upon entry to
12841 every function (that actually uses some stack space). If there is
12842 insufficient space available then either the function
12843 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
12844 called, depending upon the amount of stack space required. The runtime
12845 system is required to provide these functions. The default is
12846 @option{-mno-apcs-stack-check}, since this produces smaller code.
12848 @c not currently implemented
12850 @opindex mapcs-float
12851 Pass floating-point arguments using the floating-point registers. This is
12852 one of the variants of the APCS@. This option is recommended if the
12853 target hardware has a floating-point unit or if a lot of floating-point
12854 arithmetic is going to be performed by the code. The default is
12855 @option{-mno-apcs-float}, since the size of integer-only code is
12856 slightly increased if @option{-mapcs-float} is used.
12858 @c not currently implemented
12859 @item -mapcs-reentrant
12860 @opindex mapcs-reentrant
12861 Generate reentrant, position-independent code. The default is
12862 @option{-mno-apcs-reentrant}.
12865 @item -mthumb-interwork
12866 @opindex mthumb-interwork
12867 Generate code that supports calling between the ARM and Thumb
12868 instruction sets. Without this option, on pre-v5 architectures, the
12869 two instruction sets cannot be reliably used inside one program. The
12870 default is @option{-mno-thumb-interwork}, since slightly larger code
12871 is generated when @option{-mthumb-interwork} is specified. In AAPCS
12872 configurations this option is meaningless.
12874 @item -mno-sched-prolog
12875 @opindex mno-sched-prolog
12876 Prevent the reordering of instructions in the function prologue, or the
12877 merging of those instruction with the instructions in the function's
12878 body. This means that all functions start with a recognizable set
12879 of instructions (or in fact one of a choice from a small set of
12880 different function prologues), and this information can be used to
12881 locate the start of functions inside an executable piece of code. The
12882 default is @option{-msched-prolog}.
12884 @item -mfloat-abi=@var{name}
12885 @opindex mfloat-abi
12886 Specifies which floating-point ABI to use. Permissible values
12887 are: @samp{soft}, @samp{softfp} and @samp{hard}.
12889 Specifying @samp{soft} causes GCC to generate output containing
12890 library calls for floating-point operations.
12891 @samp{softfp} allows the generation of code using hardware floating-point
12892 instructions, but still uses the soft-float calling conventions.
12893 @samp{hard} allows generation of floating-point instructions
12894 and uses FPU-specific calling conventions.
12896 The default depends on the specific target configuration. Note that
12897 the hard-float and soft-float ABIs are not link-compatible; you must
12898 compile your entire program with the same ABI, and link with a
12899 compatible set of libraries.
12901 @item -mlittle-endian
12902 @opindex mlittle-endian
12903 Generate code for a processor running in little-endian mode. This is
12904 the default for all standard configurations.
12907 @opindex mbig-endian
12908 Generate code for a processor running in big-endian mode; the default is
12909 to compile code for a little-endian processor.
12911 @item -march=@var{name}
12913 This specifies the name of the target ARM architecture. GCC uses this
12914 name to determine what kind of instructions it can emit when generating
12915 assembly code. This option can be used in conjunction with or instead
12916 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
12917 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
12918 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
12919 @samp{armv6}, @samp{armv6j},
12920 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
12921 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
12922 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
12923 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
12925 @option{-march=armv7ve} is the armv7-a architecture with virtualization
12928 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
12929 architecture together with the optional CRC32 extensions.
12931 @option{-march=native} causes the compiler to auto-detect the architecture
12932 of the build computer. At present, this feature is only supported on
12933 GNU/Linux, and not all architectures are recognized. If the auto-detect
12934 is unsuccessful the option has no effect.
12936 @item -mtune=@var{name}
12938 This option specifies the name of the target ARM processor for
12939 which GCC should tune the performance of the code.
12940 For some ARM implementations better performance can be obtained by using
12942 Permissible names are: @samp{arm2}, @samp{arm250},
12943 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
12944 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
12945 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
12946 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
12948 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
12949 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
12950 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
12951 @samp{strongarm1110},
12952 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
12953 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
12954 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
12955 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
12956 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
12957 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
12958 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
12959 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
12960 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53}, @samp{cortex-a57},
12962 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
12967 @samp{cortex-m0plus},
12968 @samp{cortex-m1.small-multiply},
12969 @samp{cortex-m0.small-multiply},
12970 @samp{cortex-m0plus.small-multiply},
12971 @samp{marvell-pj4},
12972 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
12973 @samp{fa526}, @samp{fa626},
12974 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
12976 Additionally, this option can specify that GCC should tune the performance
12977 of the code for a big.LITTLE system. Permissible names are:
12978 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53}.
12980 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
12981 performance for a blend of processors within architecture @var{arch}.
12982 The aim is to generate code that run well on the current most popular
12983 processors, balancing between optimizations that benefit some CPUs in the
12984 range, and avoiding performance pitfalls of other CPUs. The effects of
12985 this option may change in future GCC versions as CPU models come and go.
12987 @option{-mtune=native} causes the compiler to auto-detect the CPU
12988 of the build computer. At present, this feature is only supported on
12989 GNU/Linux, and not all architectures are recognized. If the auto-detect is
12990 unsuccessful the option has no effect.
12992 @item -mcpu=@var{name}
12994 This specifies the name of the target ARM processor. GCC uses this name
12995 to derive the name of the target ARM architecture (as if specified
12996 by @option{-march}) and the ARM processor type for which to tune for
12997 performance (as if specified by @option{-mtune}). Where this option
12998 is used in conjunction with @option{-march} or @option{-mtune},
12999 those options take precedence over the appropriate part of this option.
13001 Permissible names for this option are the same as those for
13004 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13005 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13006 See @option{-mtune} for more information.
13008 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13009 of the build computer. At present, this feature is only supported on
13010 GNU/Linux, and not all architectures are recognized. If the auto-detect
13011 is unsuccessful the option has no effect.
13013 @item -mfpu=@var{name}
13015 This specifies what floating-point hardware (or hardware emulation) is
13016 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13017 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13018 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13019 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13020 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13021 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13023 If @option{-msoft-float} is specified this specifies the format of
13024 floating-point values.
13026 If the selected floating-point hardware includes the NEON extension
13027 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13028 operations are not generated by GCC's auto-vectorization pass unless
13029 @option{-funsafe-math-optimizations} is also specified. This is
13030 because NEON hardware does not fully implement the IEEE 754 standard for
13031 floating-point arithmetic (in particular denormal values are treated as
13032 zero), so the use of NEON instructions may lead to a loss of precision.
13034 @item -mfp16-format=@var{name}
13035 @opindex mfp16-format
13036 Specify the format of the @code{__fp16} half-precision floating-point type.
13037 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13038 the default is @samp{none}, in which case the @code{__fp16} type is not
13039 defined. @xref{Half-Precision}, for more information.
13041 @item -mstructure-size-boundary=@var{n}
13042 @opindex mstructure-size-boundary
13043 The sizes of all structures and unions are rounded up to a multiple
13044 of the number of bits set by this option. Permissible values are 8, 32
13045 and 64. The default value varies for different toolchains. For the COFF
13046 targeted toolchain the default value is 8. A value of 64 is only allowed
13047 if the underlying ABI supports it.
13049 Specifying a larger number can produce faster, more efficient code, but
13050 can also increase the size of the program. Different values are potentially
13051 incompatible. Code compiled with one value cannot necessarily expect to
13052 work with code or libraries compiled with another value, if they exchange
13053 information using structures or unions.
13055 @item -mabort-on-noreturn
13056 @opindex mabort-on-noreturn
13057 Generate a call to the function @code{abort} at the end of a
13058 @code{noreturn} function. It is executed if the function tries to
13062 @itemx -mno-long-calls
13063 @opindex mlong-calls
13064 @opindex mno-long-calls
13065 Tells the compiler to perform function calls by first loading the
13066 address of the function into a register and then performing a subroutine
13067 call on this register. This switch is needed if the target function
13068 lies outside of the 64-megabyte addressing range of the offset-based
13069 version of subroutine call instruction.
13071 Even if this switch is enabled, not all function calls are turned
13072 into long calls. The heuristic is that static functions, functions
13073 that have the @code{short_call} attribute, functions that are inside
13074 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13075 definitions have already been compiled within the current compilation
13076 unit are not turned into long calls. The exceptions to this rule are
13077 that weak function definitions, functions with the @code{long_call}
13078 attribute or the @code{section} attribute, and functions that are within
13079 the scope of a @code{#pragma long_calls} directive are always
13080 turned into long calls.
13082 This feature is not enabled by default. Specifying
13083 @option{-mno-long-calls} restores the default behavior, as does
13084 placing the function calls within the scope of a @code{#pragma
13085 long_calls_off} directive. Note these switches have no effect on how
13086 the compiler generates code to handle function calls via function
13089 @item -msingle-pic-base
13090 @opindex msingle-pic-base
13091 Treat the register used for PIC addressing as read-only, rather than
13092 loading it in the prologue for each function. The runtime system is
13093 responsible for initializing this register with an appropriate value
13094 before execution begins.
13096 @item -mpic-register=@var{reg}
13097 @opindex mpic-register
13098 Specify the register to be used for PIC addressing.
13099 For standard PIC base case, the default is any suitable register
13100 determined by compiler. For single PIC base case, the default is
13101 @samp{R9} if target is EABI based or stack-checking is enabled,
13102 otherwise the default is @samp{R10}.
13104 @item -mpic-data-is-text-relative
13105 @opindex mpic-data-is-text-relative
13106 Assume that each data segments are relative to text segment at load time.
13107 Therefore, it permits addressing data using PC-relative operations.
13108 This option is on by default for targets other than VxWorks RTP.
13110 @item -mpoke-function-name
13111 @opindex mpoke-function-name
13112 Write the name of each function into the text section, directly
13113 preceding the function prologue. The generated code is similar to this:
13117 .ascii "arm_poke_function_name", 0
13120 .word 0xff000000 + (t1 - t0)
13121 arm_poke_function_name
13123 stmfd sp!, @{fp, ip, lr, pc@}
13127 When performing a stack backtrace, code can inspect the value of
13128 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13129 location @code{pc - 12} and the top 8 bits are set, then we know that
13130 there is a function name embedded immediately preceding this location
13131 and has length @code{((pc[-3]) & 0xff000000)}.
13138 Select between generating code that executes in ARM and Thumb
13139 states. The default for most configurations is to generate code
13140 that executes in ARM state, but the default can be changed by
13141 configuring GCC with the @option{--with-mode=}@var{state}
13145 @opindex mtpcs-frame
13146 Generate a stack frame that is compliant with the Thumb Procedure Call
13147 Standard for all non-leaf functions. (A leaf function is one that does
13148 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13150 @item -mtpcs-leaf-frame
13151 @opindex mtpcs-leaf-frame
13152 Generate a stack frame that is compliant with the Thumb Procedure Call
13153 Standard for all leaf functions. (A leaf function is one that does
13154 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13156 @item -mcallee-super-interworking
13157 @opindex mcallee-super-interworking
13158 Gives all externally visible functions in the file being compiled an ARM
13159 instruction set header which switches to Thumb mode before executing the
13160 rest of the function. This allows these functions to be called from
13161 non-interworking code. This option is not valid in AAPCS configurations
13162 because interworking is enabled by default.
13164 @item -mcaller-super-interworking
13165 @opindex mcaller-super-interworking
13166 Allows calls via function pointers (including virtual functions) to
13167 execute correctly regardless of whether the target code has been
13168 compiled for interworking or not. There is a small overhead in the cost
13169 of executing a function pointer if this option is enabled. This option
13170 is not valid in AAPCS configurations because interworking is enabled
13173 @item -mtp=@var{name}
13175 Specify the access model for the thread local storage pointer. The valid
13176 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13177 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13178 (supported in the arm6k architecture), and @samp{auto}, which uses the
13179 best available method for the selected processor. The default setting is
13182 @item -mtls-dialect=@var{dialect}
13183 @opindex mtls-dialect
13184 Specify the dialect to use for accessing thread local storage. Two
13185 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13186 @samp{gnu} dialect selects the original GNU scheme for supporting
13187 local and global dynamic TLS models. The @samp{gnu2} dialect
13188 selects the GNU descriptor scheme, which provides better performance
13189 for shared libraries. The GNU descriptor scheme is compatible with
13190 the original scheme, but does require new assembler, linker and
13191 library support. Initial and local exec TLS models are unaffected by
13192 this option and always use the original scheme.
13194 @item -mword-relocations
13195 @opindex mword-relocations
13196 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13197 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13198 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13201 @item -mfix-cortex-m3-ldrd
13202 @opindex mfix-cortex-m3-ldrd
13203 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13204 with overlapping destination and base registers are used. This option avoids
13205 generating these instructions. This option is enabled by default when
13206 @option{-mcpu=cortex-m3} is specified.
13208 @item -munaligned-access
13209 @itemx -mno-unaligned-access
13210 @opindex munaligned-access
13211 @opindex mno-unaligned-access
13212 Enables (or disables) reading and writing of 16- and 32- bit values
13213 from addresses that are not 16- or 32- bit aligned. By default
13214 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13215 architectures, and enabled for all other architectures. If unaligned
13216 access is not enabled then words in packed data structures are
13217 accessed a byte at a time.
13219 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13220 generated object file to either true or false, depending upon the
13221 setting of this option. If unaligned access is enabled then the
13222 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13225 @item -mneon-for-64bits
13226 @opindex mneon-for-64bits
13227 Enables using Neon to handle scalar 64-bits operations. This is
13228 disabled by default since the cost of moving data from core registers
13231 @item -mslow-flash-data
13232 @opindex mslow-flash-data
13233 Assume loading data from flash is slower than fetching instruction.
13234 Therefore literal load is minimized for better performance.
13235 This option is only supported when compiling for ARMv7 M-profile and
13238 @item -masm-syntax-unified
13239 @opindex masm-syntax-unified
13240 Assume inline assembler is using unified asm syntax. The default is
13241 currently off which implies divided syntax. Currently this option is
13242 available only for Thumb1 and has no effect on ARM state and Thumb2.
13243 However, this may change in future releases of GCC. Divided syntax
13244 should be considered deprecated.
13246 @item -mrestrict-it
13247 @opindex mrestrict-it
13248 Restricts generation of IT blocks to conform to the rules of ARMv8.
13249 IT blocks can only contain a single 16-bit instruction from a select
13250 set of instructions. This option is on by default for ARMv8 Thumb mode.
13254 @subsection AVR Options
13255 @cindex AVR Options
13257 These options are defined for AVR implementations:
13260 @item -mmcu=@var{mcu}
13262 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13264 The default for this option is@tie{}@samp{avr2}.
13266 GCC supports the following AVR devices and ISAs:
13268 @include avr-mmcu.texi
13270 @item -maccumulate-args
13271 @opindex maccumulate-args
13272 Accumulate outgoing function arguments and acquire/release the needed
13273 stack space for outgoing function arguments once in function
13274 prologue/epilogue. Without this option, outgoing arguments are pushed
13275 before calling a function and popped afterwards.
13277 Popping the arguments after the function call can be expensive on
13278 AVR so that accumulating the stack space might lead to smaller
13279 executables because arguments need not to be removed from the
13280 stack after such a function call.
13282 This option can lead to reduced code size for functions that perform
13283 several calls to functions that get their arguments on the stack like
13284 calls to printf-like functions.
13286 @item -mbranch-cost=@var{cost}
13287 @opindex mbranch-cost
13288 Set the branch costs for conditional branch instructions to
13289 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13290 integers. The default branch cost is 0.
13292 @item -mcall-prologues
13293 @opindex mcall-prologues
13294 Functions prologues/epilogues are expanded as calls to appropriate
13295 subroutines. Code size is smaller.
13299 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13300 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13301 and @code{long long} is 4 bytes. Please note that this option does not
13302 conform to the C standards, but it results in smaller code
13305 @item -mno-interrupts
13306 @opindex mno-interrupts
13307 Generated code is not compatible with hardware interrupts.
13308 Code size is smaller.
13312 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13313 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13314 Setting @option{-mrelax} just adds the @option{--relax} option to the
13315 linker command line when the linker is called.
13317 Jump relaxing is performed by the linker because jump offsets are not
13318 known before code is located. Therefore, the assembler code generated by the
13319 compiler is the same, but the instructions in the executable may
13320 differ from instructions in the assembler code.
13322 Relaxing must be turned on if linker stubs are needed, see the
13323 section on @code{EIND} and linker stubs below.
13327 Treat the stack pointer register as an 8-bit register,
13328 i.e.@: assume the high byte of the stack pointer is zero.
13329 In general, you don't need to set this option by hand.
13331 This option is used internally by the compiler to select and
13332 build multilibs for architectures @code{avr2} and @code{avr25}.
13333 These architectures mix devices with and without @code{SPH}.
13334 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13335 the compiler driver adds or removes this option from the compiler
13336 proper's command line, because the compiler then knows if the device
13337 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13342 Use address register @code{X} in a way proposed by the hardware. This means
13343 that @code{X} is only used in indirect, post-increment or
13344 pre-decrement addressing.
13346 Without this option, the @code{X} register may be used in the same way
13347 as @code{Y} or @code{Z} which then is emulated by additional
13349 For example, loading a value with @code{X+const} addressing with a
13350 small non-negative @code{const < 64} to a register @var{Rn} is
13354 adiw r26, const ; X += const
13355 ld @var{Rn}, X ; @var{Rn} = *X
13356 sbiw r26, const ; X -= const
13360 @opindex mtiny-stack
13361 Only change the lower 8@tie{}bits of the stack pointer.
13363 @item -Waddr-space-convert
13364 @opindex Waddr-space-convert
13365 Warn about conversions between address spaces in the case where the
13366 resulting address space is not contained in the incoming address space.
13369 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
13370 @cindex @code{EIND}
13371 Pointers in the implementation are 16@tie{}bits wide.
13372 The address of a function or label is represented as word address so
13373 that indirect jumps and calls can target any code address in the
13374 range of 64@tie{}Ki words.
13376 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13377 bytes of program memory space, there is a special function register called
13378 @code{EIND} that serves as most significant part of the target address
13379 when @code{EICALL} or @code{EIJMP} instructions are used.
13381 Indirect jumps and calls on these devices are handled as follows by
13382 the compiler and are subject to some limitations:
13387 The compiler never sets @code{EIND}.
13390 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13391 instructions or might read @code{EIND} directly in order to emulate an
13392 indirect call/jump by means of a @code{RET} instruction.
13395 The compiler assumes that @code{EIND} never changes during the startup
13396 code or during the application. In particular, @code{EIND} is not
13397 saved/restored in function or interrupt service routine
13401 For indirect calls to functions and computed goto, the linker
13402 generates @emph{stubs}. Stubs are jump pads sometimes also called
13403 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13404 The stub contains a direct jump to the desired address.
13407 Linker relaxation must be turned on so that the linker generates
13408 the stubs correctly in all situations. See the compiler option
13409 @option{-mrelax} and the linker option @option{--relax}.
13410 There are corner cases where the linker is supposed to generate stubs
13411 but aborts without relaxation and without a helpful error message.
13414 The default linker script is arranged for code with @code{EIND = 0}.
13415 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13416 linker script has to be used in order to place the sections whose
13417 name start with @code{.trampolines} into the segment where @code{EIND}
13421 The startup code from libgcc never sets @code{EIND}.
13422 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13423 For the impact of AVR-LibC on @code{EIND}, see the
13424 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13427 It is legitimate for user-specific startup code to set up @code{EIND}
13428 early, for example by means of initialization code located in
13429 section @code{.init3}. Such code runs prior to general startup code
13430 that initializes RAM and calls constructors, but after the bit
13431 of startup code from AVR-LibC that sets @code{EIND} to the segment
13432 where the vector table is located.
13434 #include <avr/io.h>
13437 __attribute__((section(".init3"),naked,used,no_instrument_function))
13438 init3_set_eind (void)
13440 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13441 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13446 The @code{__trampolines_start} symbol is defined in the linker script.
13449 Stubs are generated automatically by the linker if
13450 the following two conditions are met:
13453 @item The address of a label is taken by means of the @code{gs} modifier
13454 (short for @emph{generate stubs}) like so:
13456 LDI r24, lo8(gs(@var{func}))
13457 LDI r25, hi8(gs(@var{func}))
13459 @item The final location of that label is in a code segment
13460 @emph{outside} the segment where the stubs are located.
13464 The compiler emits such @code{gs} modifiers for code labels in the
13465 following situations:
13467 @item Taking address of a function or code label.
13468 @item Computed goto.
13469 @item If prologue-save function is used, see @option{-mcall-prologues}
13470 command-line option.
13471 @item Switch/case dispatch tables. If you do not want such dispatch
13472 tables you can specify the @option{-fno-jump-tables} command-line option.
13473 @item C and C++ constructors/destructors called during startup/shutdown.
13474 @item If the tools hit a @code{gs()} modifier explained above.
13478 Jumping to non-symbolic addresses like so is @emph{not} supported:
13483 /* Call function at word address 0x2 */
13484 return ((int(*)(void)) 0x2)();
13488 Instead, a stub has to be set up, i.e.@: the function has to be called
13489 through a symbol (@code{func_4} in the example):
13494 extern int func_4 (void);
13496 /* Call function at byte address 0x4 */
13501 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13502 Alternatively, @code{func_4} can be defined in the linker script.
13505 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13506 @cindex @code{RAMPD}
13507 @cindex @code{RAMPX}
13508 @cindex @code{RAMPY}
13509 @cindex @code{RAMPZ}
13510 Some AVR devices support memories larger than the 64@tie{}KiB range
13511 that can be accessed with 16-bit pointers. To access memory locations
13512 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13513 register is used as high part of the address:
13514 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13515 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13516 register, respectively, to get a wide address. Similarly,
13517 @code{RAMPD} is used together with direct addressing.
13521 The startup code initializes the @code{RAMP} special function
13522 registers with zero.
13525 If a @ref{AVR Named Address Spaces,named address space} other than
13526 generic or @code{__flash} is used, then @code{RAMPZ} is set
13527 as needed before the operation.
13530 If the device supports RAM larger than 64@tie{}KiB and the compiler
13531 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13532 is reset to zero after the operation.
13535 If the device comes with a specific @code{RAMP} register, the ISR
13536 prologue/epilogue saves/restores that SFR and initializes it with
13537 zero in case the ISR code might (implicitly) use it.
13540 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13541 If you use inline assembler to read from locations outside the
13542 16-bit address range and change one of the @code{RAMP} registers,
13543 you must reset it to zero after the access.
13547 @subsubsection AVR Built-in Macros
13549 GCC defines several built-in macros so that the user code can test
13550 for the presence or absence of features. Almost any of the following
13551 built-in macros are deduced from device capabilities and thus
13552 triggered by the @option{-mmcu=} command-line option.
13554 For even more AVR-specific built-in macros see
13555 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13560 Build-in macro that resolves to a decimal number that identifies the
13561 architecture and depends on the @option{-mmcu=@var{mcu}} option.
13562 Possible values are:
13564 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13565 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
13566 @code{105}, @code{106}, @code{107}
13568 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
13569 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
13570 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
13571 @code{avrxmega6}, @code{avrxmega7}, respectively.
13572 If @var{mcu} specifies a device, this built-in macro is set
13573 accordingly. For example, with @option{-mmcu=atmega8} the macro is
13574 defined to @code{4}.
13576 @item __AVR_@var{Device}__
13577 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
13578 the device's name. For example, @option{-mmcu=atmega8} defines the
13579 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
13580 @code{__AVR_ATtiny261A__}, etc.
13582 The built-in macros' names follow
13583 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13584 the device name as from the AVR user manual. The difference between
13585 @var{Device} in the built-in macro and @var{device} in
13586 @option{-mmcu=@var{device}} is that the latter is always lowercase.
13588 If @var{device} is not a device but only a core architecture like
13589 @samp{avr51}, this macro is not defined.
13591 @item __AVR_DEVICE_NAME__
13592 Setting @option{-mmcu=@var{device}} defines this built-in macro to
13593 the device's name. For example, with @option{-mmcu=atmega8} the macro
13594 is defined to @code{atmega8}.
13596 If @var{device} is not a device but only a core architecture like
13597 @samp{avr51}, this macro is not defined.
13599 @item __AVR_XMEGA__
13600 The device / architecture belongs to the XMEGA family of devices.
13602 @item __AVR_HAVE_ELPM__
13603 The device has the the @code{ELPM} instruction.
13605 @item __AVR_HAVE_ELPMX__
13606 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13607 R@var{n},Z+} instructions.
13609 @item __AVR_HAVE_MOVW__
13610 The device has the @code{MOVW} instruction to perform 16-bit
13611 register-register moves.
13613 @item __AVR_HAVE_LPMX__
13614 The device has the @code{LPM R@var{n},Z} and
13615 @code{LPM R@var{n},Z+} instructions.
13617 @item __AVR_HAVE_MUL__
13618 The device has a hardware multiplier.
13620 @item __AVR_HAVE_JMP_CALL__
13621 The device has the @code{JMP} and @code{CALL} instructions.
13622 This is the case for devices with at least 16@tie{}KiB of program
13625 @item __AVR_HAVE_EIJMP_EICALL__
13626 @itemx __AVR_3_BYTE_PC__
13627 The device has the @code{EIJMP} and @code{EICALL} instructions.
13628 This is the case for devices with more than 128@tie{}KiB of program memory.
13629 This also means that the program counter
13630 (PC) is 3@tie{}bytes wide.
13632 @item __AVR_2_BYTE_PC__
13633 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13634 with up to 128@tie{}KiB of program memory.
13636 @item __AVR_HAVE_8BIT_SP__
13637 @itemx __AVR_HAVE_16BIT_SP__
13638 The stack pointer (SP) register is treated as 8-bit respectively
13639 16-bit register by the compiler.
13640 The definition of these macros is affected by @option{-mtiny-stack}.
13642 @item __AVR_HAVE_SPH__
13644 The device has the SPH (high part of stack pointer) special function
13645 register or has an 8-bit stack pointer, respectively.
13646 The definition of these macros is affected by @option{-mmcu=} and
13647 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
13650 @item __AVR_HAVE_RAMPD__
13651 @itemx __AVR_HAVE_RAMPX__
13652 @itemx __AVR_HAVE_RAMPY__
13653 @itemx __AVR_HAVE_RAMPZ__
13654 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13655 @code{RAMPZ} special function register, respectively.
13657 @item __NO_INTERRUPTS__
13658 This macro reflects the @option{-mno-interrupts} command line option.
13660 @item __AVR_ERRATA_SKIP__
13661 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13662 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13663 instructions because of a hardware erratum. Skip instructions are
13664 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13665 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13668 @item __AVR_ISA_RMW__
13669 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13671 @item __AVR_SFR_OFFSET__=@var{offset}
13672 Instructions that can address I/O special function registers directly
13673 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13674 address as if addressed by an instruction to access RAM like @code{LD}
13675 or @code{STS}. This offset depends on the device architecture and has
13676 to be subtracted from the RAM address in order to get the
13677 respective I/O@tie{}address.
13679 @item __WITH_AVRLIBC__
13680 The compiler is configured to be used together with AVR-Libc.
13681 See the @option{--with-avrlibc} configure option.
13685 @node Blackfin Options
13686 @subsection Blackfin Options
13687 @cindex Blackfin Options
13690 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13692 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13693 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13694 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13695 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13696 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13697 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13698 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13699 @samp{bf561}, @samp{bf592}.
13701 The optional @var{sirevision} specifies the silicon revision of the target
13702 Blackfin processor. Any workarounds available for the targeted silicon revision
13703 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13704 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13705 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13706 hexadecimal digits representing the major and minor numbers in the silicon
13707 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13708 is not defined. If @var{sirevision} is @samp{any}, the
13709 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13710 If this optional @var{sirevision} is not used, GCC assumes the latest known
13711 silicon revision of the targeted Blackfin processor.
13713 GCC defines a preprocessor macro for the specified @var{cpu}.
13714 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13715 provided by libgloss to be linked in if @option{-msim} is not given.
13717 Without this option, @samp{bf532} is used as the processor by default.
13719 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13720 only the preprocessor macro is defined.
13724 Specifies that the program will be run on the simulator. This causes
13725 the simulator BSP provided by libgloss to be linked in. This option
13726 has effect only for @samp{bfin-elf} toolchain.
13727 Certain other options, such as @option{-mid-shared-library} and
13728 @option{-mfdpic}, imply @option{-msim}.
13730 @item -momit-leaf-frame-pointer
13731 @opindex momit-leaf-frame-pointer
13732 Don't keep the frame pointer in a register for leaf functions. This
13733 avoids the instructions to save, set up and restore frame pointers and
13734 makes an extra register available in leaf functions. The option
13735 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13736 which might make debugging harder.
13738 @item -mspecld-anomaly
13739 @opindex mspecld-anomaly
13740 When enabled, the compiler ensures that the generated code does not
13741 contain speculative loads after jump instructions. If this option is used,
13742 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13744 @item -mno-specld-anomaly
13745 @opindex mno-specld-anomaly
13746 Don't generate extra code to prevent speculative loads from occurring.
13748 @item -mcsync-anomaly
13749 @opindex mcsync-anomaly
13750 When enabled, the compiler ensures that the generated code does not
13751 contain CSYNC or SSYNC instructions too soon after conditional branches.
13752 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13754 @item -mno-csync-anomaly
13755 @opindex mno-csync-anomaly
13756 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13757 occurring too soon after a conditional branch.
13761 When enabled, the compiler is free to take advantage of the knowledge that
13762 the entire program fits into the low 64k of memory.
13765 @opindex mno-low-64k
13766 Assume that the program is arbitrarily large. This is the default.
13768 @item -mstack-check-l1
13769 @opindex mstack-check-l1
13770 Do stack checking using information placed into L1 scratchpad memory by the
13773 @item -mid-shared-library
13774 @opindex mid-shared-library
13775 Generate code that supports shared libraries via the library ID method.
13776 This allows for execute in place and shared libraries in an environment
13777 without virtual memory management. This option implies @option{-fPIC}.
13778 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13780 @item -mno-id-shared-library
13781 @opindex mno-id-shared-library
13782 Generate code that doesn't assume ID-based shared libraries are being used.
13783 This is the default.
13785 @item -mleaf-id-shared-library
13786 @opindex mleaf-id-shared-library
13787 Generate code that supports shared libraries via the library ID method,
13788 but assumes that this library or executable won't link against any other
13789 ID shared libraries. That allows the compiler to use faster code for jumps
13792 @item -mno-leaf-id-shared-library
13793 @opindex mno-leaf-id-shared-library
13794 Do not assume that the code being compiled won't link against any ID shared
13795 libraries. Slower code is generated for jump and call insns.
13797 @item -mshared-library-id=n
13798 @opindex mshared-library-id
13799 Specifies the identification number of the ID-based shared library being
13800 compiled. Specifying a value of 0 generates more compact code; specifying
13801 other values forces the allocation of that number to the current
13802 library but is no more space- or time-efficient than omitting this option.
13806 Generate code that allows the data segment to be located in a different
13807 area of memory from the text segment. This allows for execute in place in
13808 an environment without virtual memory management by eliminating relocations
13809 against the text section.
13811 @item -mno-sep-data
13812 @opindex mno-sep-data
13813 Generate code that assumes that the data segment follows the text segment.
13814 This is the default.
13817 @itemx -mno-long-calls
13818 @opindex mlong-calls
13819 @opindex mno-long-calls
13820 Tells the compiler to perform function calls by first loading the
13821 address of the function into a register and then performing a subroutine
13822 call on this register. This switch is needed if the target function
13823 lies outside of the 24-bit addressing range of the offset-based
13824 version of subroutine call instruction.
13826 This feature is not enabled by default. Specifying
13827 @option{-mno-long-calls} restores the default behavior. Note these
13828 switches have no effect on how the compiler generates code to handle
13829 function calls via function pointers.
13833 Link with the fast floating-point library. This library relaxes some of
13834 the IEEE floating-point standard's rules for checking inputs against
13835 Not-a-Number (NAN), in the interest of performance.
13838 @opindex minline-plt
13839 Enable inlining of PLT entries in function calls to functions that are
13840 not known to bind locally. It has no effect without @option{-mfdpic}.
13843 @opindex mmulticore
13844 Build a standalone application for multicore Blackfin processors.
13845 This option causes proper start files and link scripts supporting
13846 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
13847 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
13849 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
13850 selects the one-application-per-core programming model. Without
13851 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
13852 programming model is used. In this model, the main function of Core B
13853 should be named as @code{coreb_main}.
13855 If this option is not used, the single-core application programming
13860 Build a standalone application for Core A of BF561 when using
13861 the one-application-per-core programming model. Proper start files
13862 and link scripts are used to support Core A, and the macro
13863 @code{__BFIN_COREA} is defined.
13864 This option can only be used in conjunction with @option{-mmulticore}.
13868 Build a standalone application for Core B of BF561 when using
13869 the one-application-per-core programming model. Proper start files
13870 and link scripts are used to support Core B, and the macro
13871 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
13872 should be used instead of @code{main}.
13873 This option can only be used in conjunction with @option{-mmulticore}.
13877 Build a standalone application for SDRAM. Proper start files and
13878 link scripts are used to put the application into SDRAM, and the macro
13879 @code{__BFIN_SDRAM} is defined.
13880 The loader should initialize SDRAM before loading the application.
13884 Assume that ICPLBs are enabled at run time. This has an effect on certain
13885 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
13886 are enabled; for standalone applications the default is off.
13890 @subsection C6X Options
13891 @cindex C6X Options
13894 @item -march=@var{name}
13896 This specifies the name of the target architecture. GCC uses this
13897 name to determine what kind of instructions it can emit when generating
13898 assembly code. Permissible names are: @samp{c62x},
13899 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
13902 @opindex mbig-endian
13903 Generate code for a big-endian target.
13905 @item -mlittle-endian
13906 @opindex mlittle-endian
13907 Generate code for a little-endian target. This is the default.
13911 Choose startup files and linker script suitable for the simulator.
13913 @item -msdata=default
13914 @opindex msdata=default
13915 Put small global and static data in the @code{.neardata} section,
13916 which is pointed to by register @code{B14}. Put small uninitialized
13917 global and static data in the @code{.bss} section, which is adjacent
13918 to the @code{.neardata} section. Put small read-only data into the
13919 @code{.rodata} section. The corresponding sections used for large
13920 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
13923 @opindex msdata=all
13924 Put all data, not just small objects, into the sections reserved for
13925 small data, and use addressing relative to the @code{B14} register to
13929 @opindex msdata=none
13930 Make no use of the sections reserved for small data, and use absolute
13931 addresses to access all data. Put all initialized global and static
13932 data in the @code{.fardata} section, and all uninitialized data in the
13933 @code{.far} section. Put all constant data into the @code{.const}
13938 @subsection CRIS Options
13939 @cindex CRIS Options
13941 These options are defined specifically for the CRIS ports.
13944 @item -march=@var{architecture-type}
13945 @itemx -mcpu=@var{architecture-type}
13948 Generate code for the specified architecture. The choices for
13949 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
13950 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
13951 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
13954 @item -mtune=@var{architecture-type}
13956 Tune to @var{architecture-type} everything applicable about the generated
13957 code, except for the ABI and the set of available instructions. The
13958 choices for @var{architecture-type} are the same as for
13959 @option{-march=@var{architecture-type}}.
13961 @item -mmax-stack-frame=@var{n}
13962 @opindex mmax-stack-frame
13963 Warn when the stack frame of a function exceeds @var{n} bytes.
13969 The options @option{-metrax4} and @option{-metrax100} are synonyms for
13970 @option{-march=v3} and @option{-march=v8} respectively.
13972 @item -mmul-bug-workaround
13973 @itemx -mno-mul-bug-workaround
13974 @opindex mmul-bug-workaround
13975 @opindex mno-mul-bug-workaround
13976 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
13977 models where it applies. This option is active by default.
13981 Enable CRIS-specific verbose debug-related information in the assembly
13982 code. This option also has the effect of turning off the @samp{#NO_APP}
13983 formatted-code indicator to the assembler at the beginning of the
13988 Do not use condition-code results from previous instruction; always emit
13989 compare and test instructions before use of condition codes.
13991 @item -mno-side-effects
13992 @opindex mno-side-effects
13993 Do not emit instructions with side effects in addressing modes other than
13996 @item -mstack-align
13997 @itemx -mno-stack-align
13998 @itemx -mdata-align
13999 @itemx -mno-data-align
14000 @itemx -mconst-align
14001 @itemx -mno-const-align
14002 @opindex mstack-align
14003 @opindex mno-stack-align
14004 @opindex mdata-align
14005 @opindex mno-data-align
14006 @opindex mconst-align
14007 @opindex mno-const-align
14008 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14009 stack frame, individual data and constants to be aligned for the maximum
14010 single data access size for the chosen CPU model. The default is to
14011 arrange for 32-bit alignment. ABI details such as structure layout are
14012 not affected by these options.
14020 Similar to the stack- data- and const-align options above, these options
14021 arrange for stack frame, writable data and constants to all be 32-bit,
14022 16-bit or 8-bit aligned. The default is 32-bit alignment.
14024 @item -mno-prologue-epilogue
14025 @itemx -mprologue-epilogue
14026 @opindex mno-prologue-epilogue
14027 @opindex mprologue-epilogue
14028 With @option{-mno-prologue-epilogue}, the normal function prologue and
14029 epilogue which set up the stack frame are omitted and no return
14030 instructions or return sequences are generated in the code. Use this
14031 option only together with visual inspection of the compiled code: no
14032 warnings or errors are generated when call-saved registers must be saved,
14033 or storage for local variables needs to be allocated.
14037 @opindex mno-gotplt
14039 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14040 instruction sequences that load addresses for functions from the PLT part
14041 of the GOT rather than (traditional on other architectures) calls to the
14042 PLT@. The default is @option{-mgotplt}.
14046 Legacy no-op option only recognized with the cris-axis-elf and
14047 cris-axis-linux-gnu targets.
14051 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14055 This option, recognized for the cris-axis-elf, arranges
14056 to link with input-output functions from a simulator library. Code,
14057 initialized data and zero-initialized data are allocated consecutively.
14061 Like @option{-sim}, but pass linker options to locate initialized data at
14062 0x40000000 and zero-initialized data at 0x80000000.
14066 @subsection CR16 Options
14067 @cindex CR16 Options
14069 These options are defined specifically for the CR16 ports.
14075 Enable the use of multiply-accumulate instructions. Disabled by default.
14079 @opindex mcr16cplus
14081 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14086 Links the library libsim.a which is in compatible with simulator. Applicable
14087 to ELF compiler only.
14091 Choose integer type as 32-bit wide.
14095 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14097 @item -mdata-model=@var{model}
14098 @opindex mdata-model
14099 Choose a data model. The choices for @var{model} are @samp{near},
14100 @samp{far} or @samp{medium}. @samp{medium} is default.
14101 However, @samp{far} is not valid with @option{-mcr16c}, as the
14102 CR16C architecture does not support the far data model.
14105 @node Darwin Options
14106 @subsection Darwin Options
14107 @cindex Darwin options
14109 These options are defined for all architectures running the Darwin operating
14112 FSF GCC on Darwin does not create ``fat'' object files; it creates
14113 an object file for the single architecture that GCC was built to
14114 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14115 @option{-arch} options are used; it does so by running the compiler or
14116 linker multiple times and joining the results together with
14119 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14120 @samp{i686}) is determined by the flags that specify the ISA
14121 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14122 @option{-force_cpusubtype_ALL} option can be used to override this.
14124 The Darwin tools vary in their behavior when presented with an ISA
14125 mismatch. The assembler, @file{as}, only permits instructions to
14126 be used that are valid for the subtype of the file it is generating,
14127 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14128 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14129 and prints an error if asked to create a shared library with a less
14130 restrictive subtype than its input files (for instance, trying to put
14131 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14132 for executables, @command{ld}, quietly gives the executable the most
14133 restrictive subtype of any of its input files.
14138 Add the framework directory @var{dir} to the head of the list of
14139 directories to be searched for header files. These directories are
14140 interleaved with those specified by @option{-I} options and are
14141 scanned in a left-to-right order.
14143 A framework directory is a directory with frameworks in it. A
14144 framework is a directory with a @file{Headers} and/or
14145 @file{PrivateHeaders} directory contained directly in it that ends
14146 in @file{.framework}. The name of a framework is the name of this
14147 directory excluding the @file{.framework}. Headers associated with
14148 the framework are found in one of those two directories, with
14149 @file{Headers} being searched first. A subframework is a framework
14150 directory that is in a framework's @file{Frameworks} directory.
14151 Includes of subframework headers can only appear in a header of a
14152 framework that contains the subframework, or in a sibling subframework
14153 header. Two subframeworks are siblings if they occur in the same
14154 framework. A subframework should not have the same name as a
14155 framework; a warning is issued if this is violated. Currently a
14156 subframework cannot have subframeworks; in the future, the mechanism
14157 may be extended to support this. The standard frameworks can be found
14158 in @file{/System/Library/Frameworks} and
14159 @file{/Library/Frameworks}. An example include looks like
14160 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14161 the name of the framework and @file{header.h} is found in the
14162 @file{PrivateHeaders} or @file{Headers} directory.
14164 @item -iframework@var{dir}
14165 @opindex iframework
14166 Like @option{-F} except the directory is a treated as a system
14167 directory. The main difference between this @option{-iframework} and
14168 @option{-F} is that with @option{-iframework} the compiler does not
14169 warn about constructs contained within header files found via
14170 @var{dir}. This option is valid only for the C family of languages.
14174 Emit debugging information for symbols that are used. For stabs
14175 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14176 This is by default ON@.
14180 Emit debugging information for all symbols and types.
14182 @item -mmacosx-version-min=@var{version}
14183 The earliest version of MacOS X that this executable will run on
14184 is @var{version}. Typical values of @var{version} include @code{10.1},
14185 @code{10.2}, and @code{10.3.9}.
14187 If the compiler was built to use the system's headers by default,
14188 then the default for this option is the system version on which the
14189 compiler is running, otherwise the default is to make choices that
14190 are compatible with as many systems and code bases as possible.
14194 Enable kernel development mode. The @option{-mkernel} option sets
14195 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14196 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14197 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14198 applicable. This mode also sets @option{-mno-altivec},
14199 @option{-msoft-float}, @option{-fno-builtin} and
14200 @option{-mlong-branch} for PowerPC targets.
14202 @item -mone-byte-bool
14203 @opindex mone-byte-bool
14204 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14205 By default @code{sizeof(bool)} is @code{4} when compiling for
14206 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14207 option has no effect on x86.
14209 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14210 to generate code that is not binary compatible with code generated
14211 without that switch. Using this switch may require recompiling all
14212 other modules in a program, including system libraries. Use this
14213 switch to conform to a non-default data model.
14215 @item -mfix-and-continue
14216 @itemx -ffix-and-continue
14217 @itemx -findirect-data
14218 @opindex mfix-and-continue
14219 @opindex ffix-and-continue
14220 @opindex findirect-data
14221 Generate code suitable for fast turnaround development, such as to
14222 allow GDB to dynamically load @file{.o} files into already-running
14223 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14224 are provided for backwards compatibility.
14228 Loads all members of static archive libraries.
14229 See man ld(1) for more information.
14231 @item -arch_errors_fatal
14232 @opindex arch_errors_fatal
14233 Cause the errors having to do with files that have the wrong architecture
14236 @item -bind_at_load
14237 @opindex bind_at_load
14238 Causes the output file to be marked such that the dynamic linker will
14239 bind all undefined references when the file is loaded or launched.
14243 Produce a Mach-o bundle format file.
14244 See man ld(1) for more information.
14246 @item -bundle_loader @var{executable}
14247 @opindex bundle_loader
14248 This option specifies the @var{executable} that will load the build
14249 output file being linked. See man ld(1) for more information.
14252 @opindex dynamiclib
14253 When passed this option, GCC produces a dynamic library instead of
14254 an executable when linking, using the Darwin @file{libtool} command.
14256 @item -force_cpusubtype_ALL
14257 @opindex force_cpusubtype_ALL
14258 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14259 one controlled by the @option{-mcpu} or @option{-march} option.
14261 @item -allowable_client @var{client_name}
14262 @itemx -client_name
14263 @itemx -compatibility_version
14264 @itemx -current_version
14266 @itemx -dependency-file
14268 @itemx -dylinker_install_name
14270 @itemx -exported_symbols_list
14273 @itemx -flat_namespace
14274 @itemx -force_flat_namespace
14275 @itemx -headerpad_max_install_names
14278 @itemx -install_name
14279 @itemx -keep_private_externs
14280 @itemx -multi_module
14281 @itemx -multiply_defined
14282 @itemx -multiply_defined_unused
14285 @itemx -no_dead_strip_inits_and_terms
14286 @itemx -nofixprebinding
14287 @itemx -nomultidefs
14289 @itemx -noseglinkedit
14290 @itemx -pagezero_size
14292 @itemx -prebind_all_twolevel_modules
14293 @itemx -private_bundle
14295 @itemx -read_only_relocs
14297 @itemx -sectobjectsymbols
14301 @itemx -sectobjectsymbols
14304 @itemx -segs_read_only_addr
14306 @itemx -segs_read_write_addr
14307 @itemx -seg_addr_table
14308 @itemx -seg_addr_table_filename
14309 @itemx -seglinkedit
14311 @itemx -segs_read_only_addr
14312 @itemx -segs_read_write_addr
14313 @itemx -single_module
14315 @itemx -sub_library
14317 @itemx -sub_umbrella
14318 @itemx -twolevel_namespace
14321 @itemx -unexported_symbols_list
14322 @itemx -weak_reference_mismatches
14323 @itemx -whatsloaded
14324 @opindex allowable_client
14325 @opindex client_name
14326 @opindex compatibility_version
14327 @opindex current_version
14328 @opindex dead_strip
14329 @opindex dependency-file
14330 @opindex dylib_file
14331 @opindex dylinker_install_name
14333 @opindex exported_symbols_list
14335 @opindex flat_namespace
14336 @opindex force_flat_namespace
14337 @opindex headerpad_max_install_names
14338 @opindex image_base
14340 @opindex install_name
14341 @opindex keep_private_externs
14342 @opindex multi_module
14343 @opindex multiply_defined
14344 @opindex multiply_defined_unused
14345 @opindex noall_load
14346 @opindex no_dead_strip_inits_and_terms
14347 @opindex nofixprebinding
14348 @opindex nomultidefs
14350 @opindex noseglinkedit
14351 @opindex pagezero_size
14353 @opindex prebind_all_twolevel_modules
14354 @opindex private_bundle
14355 @opindex read_only_relocs
14357 @opindex sectobjectsymbols
14360 @opindex sectcreate
14361 @opindex sectobjectsymbols
14364 @opindex segs_read_only_addr
14365 @opindex segs_read_write_addr
14366 @opindex seg_addr_table
14367 @opindex seg_addr_table_filename
14368 @opindex seglinkedit
14370 @opindex segs_read_only_addr
14371 @opindex segs_read_write_addr
14372 @opindex single_module
14374 @opindex sub_library
14375 @opindex sub_umbrella
14376 @opindex twolevel_namespace
14379 @opindex unexported_symbols_list
14380 @opindex weak_reference_mismatches
14381 @opindex whatsloaded
14382 These options are passed to the Darwin linker. The Darwin linker man page
14383 describes them in detail.
14386 @node DEC Alpha Options
14387 @subsection DEC Alpha Options
14389 These @samp{-m} options are defined for the DEC Alpha implementations:
14392 @item -mno-soft-float
14393 @itemx -msoft-float
14394 @opindex mno-soft-float
14395 @opindex msoft-float
14396 Use (do not use) the hardware floating-point instructions for
14397 floating-point operations. When @option{-msoft-float} is specified,
14398 functions in @file{libgcc.a} are used to perform floating-point
14399 operations. Unless they are replaced by routines that emulate the
14400 floating-point operations, or compiled in such a way as to call such
14401 emulations routines, these routines issue floating-point
14402 operations. If you are compiling for an Alpha without floating-point
14403 operations, you must ensure that the library is built so as not to call
14406 Note that Alpha implementations without floating-point operations are
14407 required to have floating-point registers.
14410 @itemx -mno-fp-regs
14412 @opindex mno-fp-regs
14413 Generate code that uses (does not use) the floating-point register set.
14414 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14415 register set is not used, floating-point operands are passed in integer
14416 registers as if they were integers and floating-point results are passed
14417 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14418 so any function with a floating-point argument or return value called by code
14419 compiled with @option{-mno-fp-regs} must also be compiled with that
14422 A typical use of this option is building a kernel that does not use,
14423 and hence need not save and restore, any floating-point registers.
14427 The Alpha architecture implements floating-point hardware optimized for
14428 maximum performance. It is mostly compliant with the IEEE floating-point
14429 standard. However, for full compliance, software assistance is
14430 required. This option generates code fully IEEE-compliant code
14431 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14432 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14433 defined during compilation. The resulting code is less efficient but is
14434 able to correctly support denormalized numbers and exceptional IEEE
14435 values such as not-a-number and plus/minus infinity. Other Alpha
14436 compilers call this option @option{-ieee_with_no_inexact}.
14438 @item -mieee-with-inexact
14439 @opindex mieee-with-inexact
14440 This is like @option{-mieee} except the generated code also maintains
14441 the IEEE @var{inexact-flag}. Turning on this option causes the
14442 generated code to implement fully-compliant IEEE math. In addition to
14443 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14444 macro. On some Alpha implementations the resulting code may execute
14445 significantly slower than the code generated by default. Since there is
14446 very little code that depends on the @var{inexact-flag}, you should
14447 normally not specify this option. Other Alpha compilers call this
14448 option @option{-ieee_with_inexact}.
14450 @item -mfp-trap-mode=@var{trap-mode}
14451 @opindex mfp-trap-mode
14452 This option controls what floating-point related traps are enabled.
14453 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14454 The trap mode can be set to one of four values:
14458 This is the default (normal) setting. The only traps that are enabled
14459 are the ones that cannot be disabled in software (e.g., division by zero
14463 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14467 Like @samp{u}, but the instructions are marked to be safe for software
14468 completion (see Alpha architecture manual for details).
14471 Like @samp{su}, but inexact traps are enabled as well.
14474 @item -mfp-rounding-mode=@var{rounding-mode}
14475 @opindex mfp-rounding-mode
14476 Selects the IEEE rounding mode. Other Alpha compilers call this option
14477 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14482 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14483 the nearest machine number or towards the even machine number in case
14487 Round towards minus infinity.
14490 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14493 Dynamic rounding mode. A field in the floating-point control register
14494 (@var{fpcr}, see Alpha architecture reference manual) controls the
14495 rounding mode in effect. The C library initializes this register for
14496 rounding towards plus infinity. Thus, unless your program modifies the
14497 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14500 @item -mtrap-precision=@var{trap-precision}
14501 @opindex mtrap-precision
14502 In the Alpha architecture, floating-point traps are imprecise. This
14503 means without software assistance it is impossible to recover from a
14504 floating trap and program execution normally needs to be terminated.
14505 GCC can generate code that can assist operating system trap handlers
14506 in determining the exact location that caused a floating-point trap.
14507 Depending on the requirements of an application, different levels of
14508 precisions can be selected:
14512 Program precision. This option is the default and means a trap handler
14513 can only identify which program caused a floating-point exception.
14516 Function precision. The trap handler can determine the function that
14517 caused a floating-point exception.
14520 Instruction precision. The trap handler can determine the exact
14521 instruction that caused a floating-point exception.
14524 Other Alpha compilers provide the equivalent options called
14525 @option{-scope_safe} and @option{-resumption_safe}.
14527 @item -mieee-conformant
14528 @opindex mieee-conformant
14529 This option marks the generated code as IEEE conformant. You must not
14530 use this option unless you also specify @option{-mtrap-precision=i} and either
14531 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14532 is to emit the line @samp{.eflag 48} in the function prologue of the
14533 generated assembly file.
14535 @item -mbuild-constants
14536 @opindex mbuild-constants
14537 Normally GCC examines a 32- or 64-bit integer constant to
14538 see if it can construct it from smaller constants in two or three
14539 instructions. If it cannot, it outputs the constant as a literal and
14540 generates code to load it from the data segment at run time.
14542 Use this option to require GCC to construct @emph{all} integer constants
14543 using code, even if it takes more instructions (the maximum is six).
14545 You typically use this option to build a shared library dynamic
14546 loader. Itself a shared library, it must relocate itself in memory
14547 before it can find the variables and constants in its own data segment.
14565 Indicate whether GCC should generate code to use the optional BWX,
14566 CIX, FIX and MAX instruction sets. The default is to use the instruction
14567 sets supported by the CPU type specified via @option{-mcpu=} option or that
14568 of the CPU on which GCC was built if none is specified.
14571 @itemx -mfloat-ieee
14572 @opindex mfloat-vax
14573 @opindex mfloat-ieee
14574 Generate code that uses (does not use) VAX F and G floating-point
14575 arithmetic instead of IEEE single and double precision.
14577 @item -mexplicit-relocs
14578 @itemx -mno-explicit-relocs
14579 @opindex mexplicit-relocs
14580 @opindex mno-explicit-relocs
14581 Older Alpha assemblers provided no way to generate symbol relocations
14582 except via assembler macros. Use of these macros does not allow
14583 optimal instruction scheduling. GNU binutils as of version 2.12
14584 supports a new syntax that allows the compiler to explicitly mark
14585 which relocations should apply to which instructions. This option
14586 is mostly useful for debugging, as GCC detects the capabilities of
14587 the assembler when it is built and sets the default accordingly.
14590 @itemx -mlarge-data
14591 @opindex msmall-data
14592 @opindex mlarge-data
14593 When @option{-mexplicit-relocs} is in effect, static data is
14594 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14595 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14596 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14597 16-bit relocations off of the @code{$gp} register. This limits the
14598 size of the small data area to 64KB, but allows the variables to be
14599 directly accessed via a single instruction.
14601 The default is @option{-mlarge-data}. With this option the data area
14602 is limited to just below 2GB@. Programs that require more than 2GB of
14603 data must use @code{malloc} or @code{mmap} to allocate the data in the
14604 heap instead of in the program's data segment.
14606 When generating code for shared libraries, @option{-fpic} implies
14607 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14610 @itemx -mlarge-text
14611 @opindex msmall-text
14612 @opindex mlarge-text
14613 When @option{-msmall-text} is used, the compiler assumes that the
14614 code of the entire program (or shared library) fits in 4MB, and is
14615 thus reachable with a branch instruction. When @option{-msmall-data}
14616 is used, the compiler can assume that all local symbols share the
14617 same @code{$gp} value, and thus reduce the number of instructions
14618 required for a function call from 4 to 1.
14620 The default is @option{-mlarge-text}.
14622 @item -mcpu=@var{cpu_type}
14624 Set the instruction set and instruction scheduling parameters for
14625 machine type @var{cpu_type}. You can specify either the @samp{EV}
14626 style name or the corresponding chip number. GCC supports scheduling
14627 parameters for the EV4, EV5 and EV6 family of processors and
14628 chooses the default values for the instruction set from the processor
14629 you specify. If you do not specify a processor type, GCC defaults
14630 to the processor on which the compiler was built.
14632 Supported values for @var{cpu_type} are
14638 Schedules as an EV4 and has no instruction set extensions.
14642 Schedules as an EV5 and has no instruction set extensions.
14646 Schedules as an EV5 and supports the BWX extension.
14651 Schedules as an EV5 and supports the BWX and MAX extensions.
14655 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14659 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14662 Native toolchains also support the value @samp{native},
14663 which selects the best architecture option for the host processor.
14664 @option{-mcpu=native} has no effect if GCC does not recognize
14667 @item -mtune=@var{cpu_type}
14669 Set only the instruction scheduling parameters for machine type
14670 @var{cpu_type}. The instruction set is not changed.
14672 Native toolchains also support the value @samp{native},
14673 which selects the best architecture option for the host processor.
14674 @option{-mtune=native} has no effect if GCC does not recognize
14677 @item -mmemory-latency=@var{time}
14678 @opindex mmemory-latency
14679 Sets the latency the scheduler should assume for typical memory
14680 references as seen by the application. This number is highly
14681 dependent on the memory access patterns used by the application
14682 and the size of the external cache on the machine.
14684 Valid options for @var{time} are
14688 A decimal number representing clock cycles.
14694 The compiler contains estimates of the number of clock cycles for
14695 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14696 (also called Dcache, Scache, and Bcache), as well as to main memory.
14697 Note that L3 is only valid for EV5.
14703 @subsection FR30 Options
14704 @cindex FR30 Options
14706 These options are defined specifically for the FR30 port.
14710 @item -msmall-model
14711 @opindex msmall-model
14712 Use the small address space model. This can produce smaller code, but
14713 it does assume that all symbolic values and addresses fit into a
14718 Assume that runtime support has been provided and so there is no need
14719 to include the simulator library (@file{libsim.a}) on the linker
14725 @subsection FRV Options
14726 @cindex FRV Options
14732 Only use the first 32 general-purpose registers.
14737 Use all 64 general-purpose registers.
14742 Use only the first 32 floating-point registers.
14747 Use all 64 floating-point registers.
14750 @opindex mhard-float
14752 Use hardware instructions for floating-point operations.
14755 @opindex msoft-float
14757 Use library routines for floating-point operations.
14762 Dynamically allocate condition code registers.
14767 Do not try to dynamically allocate condition code registers, only
14768 use @code{icc0} and @code{fcc0}.
14773 Change ABI to use double word insns.
14778 Do not use double word instructions.
14783 Use floating-point double instructions.
14786 @opindex mno-double
14788 Do not use floating-point double instructions.
14793 Use media instructions.
14798 Do not use media instructions.
14803 Use multiply and add/subtract instructions.
14806 @opindex mno-muladd
14808 Do not use multiply and add/subtract instructions.
14813 Select the FDPIC ABI, which uses function descriptors to represent
14814 pointers to functions. Without any PIC/PIE-related options, it
14815 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
14816 assumes GOT entries and small data are within a 12-bit range from the
14817 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
14818 are computed with 32 bits.
14819 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14822 @opindex minline-plt
14824 Enable inlining of PLT entries in function calls to functions that are
14825 not known to bind locally. It has no effect without @option{-mfdpic}.
14826 It's enabled by default if optimizing for speed and compiling for
14827 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
14828 optimization option such as @option{-O3} or above is present in the
14834 Assume a large TLS segment when generating thread-local code.
14839 Do not assume a large TLS segment when generating thread-local code.
14844 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
14845 that is known to be in read-only sections. It's enabled by default,
14846 except for @option{-fpic} or @option{-fpie}: even though it may help
14847 make the global offset table smaller, it trades 1 instruction for 4.
14848 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
14849 one of which may be shared by multiple symbols, and it avoids the need
14850 for a GOT entry for the referenced symbol, so it's more likely to be a
14851 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
14853 @item -multilib-library-pic
14854 @opindex multilib-library-pic
14856 Link with the (library, not FD) pic libraries. It's implied by
14857 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
14858 @option{-fpic} without @option{-mfdpic}. You should never have to use
14862 @opindex mlinked-fp
14864 Follow the EABI requirement of always creating a frame pointer whenever
14865 a stack frame is allocated. This option is enabled by default and can
14866 be disabled with @option{-mno-linked-fp}.
14869 @opindex mlong-calls
14871 Use indirect addressing to call functions outside the current
14872 compilation unit. This allows the functions to be placed anywhere
14873 within the 32-bit address space.
14875 @item -malign-labels
14876 @opindex malign-labels
14878 Try to align labels to an 8-byte boundary by inserting NOPs into the
14879 previous packet. This option only has an effect when VLIW packing
14880 is enabled. It doesn't create new packets; it merely adds NOPs to
14883 @item -mlibrary-pic
14884 @opindex mlibrary-pic
14886 Generate position-independent EABI code.
14891 Use only the first four media accumulator registers.
14896 Use all eight media accumulator registers.
14901 Pack VLIW instructions.
14906 Do not pack VLIW instructions.
14909 @opindex mno-eflags
14911 Do not mark ABI switches in e_flags.
14914 @opindex mcond-move
14916 Enable the use of conditional-move instructions (default).
14918 This switch is mainly for debugging the compiler and will likely be removed
14919 in a future version.
14921 @item -mno-cond-move
14922 @opindex mno-cond-move
14924 Disable the use of conditional-move instructions.
14926 This switch is mainly for debugging the compiler and will likely be removed
14927 in a future version.
14932 Enable the use of conditional set instructions (default).
14934 This switch is mainly for debugging the compiler and will likely be removed
14935 in a future version.
14940 Disable the use of conditional set instructions.
14942 This switch is mainly for debugging the compiler and will likely be removed
14943 in a future version.
14946 @opindex mcond-exec
14948 Enable the use of conditional execution (default).
14950 This switch is mainly for debugging the compiler and will likely be removed
14951 in a future version.
14953 @item -mno-cond-exec
14954 @opindex mno-cond-exec
14956 Disable the use of conditional execution.
14958 This switch is mainly for debugging the compiler and will likely be removed
14959 in a future version.
14961 @item -mvliw-branch
14962 @opindex mvliw-branch
14964 Run a pass to pack branches into VLIW instructions (default).
14966 This switch is mainly for debugging the compiler and will likely be removed
14967 in a future version.
14969 @item -mno-vliw-branch
14970 @opindex mno-vliw-branch
14972 Do not run a pass to pack branches into VLIW instructions.
14974 This switch is mainly for debugging the compiler and will likely be removed
14975 in a future version.
14977 @item -mmulti-cond-exec
14978 @opindex mmulti-cond-exec
14980 Enable optimization of @code{&&} and @code{||} in conditional execution
14983 This switch is mainly for debugging the compiler and will likely be removed
14984 in a future version.
14986 @item -mno-multi-cond-exec
14987 @opindex mno-multi-cond-exec
14989 Disable optimization of @code{&&} and @code{||} in conditional execution.
14991 This switch is mainly for debugging the compiler and will likely be removed
14992 in a future version.
14994 @item -mnested-cond-exec
14995 @opindex mnested-cond-exec
14997 Enable nested conditional execution optimizations (default).
14999 This switch is mainly for debugging the compiler and will likely be removed
15000 in a future version.
15002 @item -mno-nested-cond-exec
15003 @opindex mno-nested-cond-exec
15005 Disable nested conditional execution optimizations.
15007 This switch is mainly for debugging the compiler and will likely be removed
15008 in a future version.
15010 @item -moptimize-membar
15011 @opindex moptimize-membar
15013 This switch removes redundant @code{membar} instructions from the
15014 compiler-generated code. It is enabled by default.
15016 @item -mno-optimize-membar
15017 @opindex mno-optimize-membar
15019 This switch disables the automatic removal of redundant @code{membar}
15020 instructions from the generated code.
15022 @item -mtomcat-stats
15023 @opindex mtomcat-stats
15025 Cause gas to print out tomcat statistics.
15027 @item -mcpu=@var{cpu}
15030 Select the processor type for which to generate code. Possible values are
15031 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15032 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15036 @node GNU/Linux Options
15037 @subsection GNU/Linux Options
15039 These @samp{-m} options are defined for GNU/Linux targets:
15044 Use the GNU C library. This is the default except
15045 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
15049 Use uClibc C library. This is the default on
15050 @samp{*-*-linux-*uclibc*} targets.
15054 Use Bionic C library. This is the default on
15055 @samp{*-*-linux-*android*} targets.
15059 Compile code compatible with Android platform. This is the default on
15060 @samp{*-*-linux-*android*} targets.
15062 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15063 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15064 this option makes the GCC driver pass Android-specific options to the linker.
15065 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15068 @item -tno-android-cc
15069 @opindex tno-android-cc
15070 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15071 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15072 @option{-fno-rtti} by default.
15074 @item -tno-android-ld
15075 @opindex tno-android-ld
15076 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15077 linking options to the linker.
15081 @node H8/300 Options
15082 @subsection H8/300 Options
15084 These @samp{-m} options are defined for the H8/300 implementations:
15089 Shorten some address references at link time, when possible; uses the
15090 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15091 ld, Using ld}, for a fuller description.
15095 Generate code for the H8/300H@.
15099 Generate code for the H8S@.
15103 Generate code for the H8S and H8/300H in the normal mode. This switch
15104 must be used either with @option{-mh} or @option{-ms}.
15108 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15112 Extended registers are stored on stack before execution of function
15113 with monitor attribute. Default option is @option{-mexr}.
15114 This option is valid only for H8S targets.
15118 Extended registers are not stored on stack before execution of function
15119 with monitor attribute. Default option is @option{-mno-exr}.
15120 This option is valid only for H8S targets.
15124 Make @code{int} data 32 bits by default.
15127 @opindex malign-300
15128 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15129 The default for the H8/300H and H8S is to align longs and floats on
15131 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15132 This option has no effect on the H8/300.
15136 @subsection HPPA Options
15137 @cindex HPPA Options
15139 These @samp{-m} options are defined for the HPPA family of computers:
15142 @item -march=@var{architecture-type}
15144 Generate code for the specified architecture. The choices for
15145 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15146 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15147 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15148 architecture option for your machine. Code compiled for lower numbered
15149 architectures runs on higher numbered architectures, but not the
15152 @item -mpa-risc-1-0
15153 @itemx -mpa-risc-1-1
15154 @itemx -mpa-risc-2-0
15155 @opindex mpa-risc-1-0
15156 @opindex mpa-risc-1-1
15157 @opindex mpa-risc-2-0
15158 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15160 @item -mjump-in-delay
15161 @opindex mjump-in-delay
15162 This option is ignored and provided for compatibility purposes only.
15164 @item -mdisable-fpregs
15165 @opindex mdisable-fpregs
15166 Prevent floating-point registers from being used in any manner. This is
15167 necessary for compiling kernels that perform lazy context switching of
15168 floating-point registers. If you use this option and attempt to perform
15169 floating-point operations, the compiler aborts.
15171 @item -mdisable-indexing
15172 @opindex mdisable-indexing
15173 Prevent the compiler from using indexing address modes. This avoids some
15174 rather obscure problems when compiling MIG generated code under MACH@.
15176 @item -mno-space-regs
15177 @opindex mno-space-regs
15178 Generate code that assumes the target has no space registers. This allows
15179 GCC to generate faster indirect calls and use unscaled index address modes.
15181 Such code is suitable for level 0 PA systems and kernels.
15183 @item -mfast-indirect-calls
15184 @opindex mfast-indirect-calls
15185 Generate code that assumes calls never cross space boundaries. This
15186 allows GCC to emit code that performs faster indirect calls.
15188 This option does not work in the presence of shared libraries or nested
15191 @item -mfixed-range=@var{register-range}
15192 @opindex mfixed-range
15193 Generate code treating the given register range as fixed registers.
15194 A fixed register is one that the register allocator cannot use. This is
15195 useful when compiling kernel code. A register range is specified as
15196 two registers separated by a dash. Multiple register ranges can be
15197 specified separated by a comma.
15199 @item -mlong-load-store
15200 @opindex mlong-load-store
15201 Generate 3-instruction load and store sequences as sometimes required by
15202 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15205 @item -mportable-runtime
15206 @opindex mportable-runtime
15207 Use the portable calling conventions proposed by HP for ELF systems.
15211 Enable the use of assembler directives only GAS understands.
15213 @item -mschedule=@var{cpu-type}
15215 Schedule code according to the constraints for the machine type
15216 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15217 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15218 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15219 proper scheduling option for your machine. The default scheduling is
15223 @opindex mlinker-opt
15224 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15225 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15226 linkers in which they give bogus error messages when linking some programs.
15229 @opindex msoft-float
15230 Generate output containing library calls for floating point.
15231 @strong{Warning:} the requisite libraries are not available for all HPPA
15232 targets. Normally the facilities of the machine's usual C compiler are
15233 used, but this cannot be done directly in cross-compilation. You must make
15234 your own arrangements to provide suitable library functions for
15237 @option{-msoft-float} changes the calling convention in the output file;
15238 therefore, it is only useful if you compile @emph{all} of a program with
15239 this option. In particular, you need to compile @file{libgcc.a}, the
15240 library that comes with GCC, with @option{-msoft-float} in order for
15245 Generate the predefine, @code{_SIO}, for server IO@. The default is
15246 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15247 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15248 options are available under HP-UX and HI-UX@.
15252 Use options specific to GNU @command{ld}.
15253 This passes @option{-shared} to @command{ld} when
15254 building a shared library. It is the default when GCC is configured,
15255 explicitly or implicitly, with the GNU linker. This option does not
15256 affect which @command{ld} is called; it only changes what parameters
15257 are passed to that @command{ld}.
15258 The @command{ld} that is called is determined by the
15259 @option{--with-ld} configure option, GCC's program search path, and
15260 finally by the user's @env{PATH}. The linker used by GCC can be printed
15261 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15262 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15266 Use options specific to HP @command{ld}.
15267 This passes @option{-b} to @command{ld} when building
15268 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15269 links. It is the default when GCC is configured, explicitly or
15270 implicitly, with the HP linker. This option does not affect
15271 which @command{ld} is called; it only changes what parameters are passed to that
15273 The @command{ld} that is called is determined by the @option{--with-ld}
15274 configure option, GCC's program search path, and finally by the user's
15275 @env{PATH}. The linker used by GCC can be printed using @samp{which
15276 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15277 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15280 @opindex mno-long-calls
15281 Generate code that uses long call sequences. This ensures that a call
15282 is always able to reach linker generated stubs. The default is to generate
15283 long calls only when the distance from the call site to the beginning
15284 of the function or translation unit, as the case may be, exceeds a
15285 predefined limit set by the branch type being used. The limits for
15286 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15287 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15290 Distances are measured from the beginning of functions when using the
15291 @option{-ffunction-sections} option, or when using the @option{-mgas}
15292 and @option{-mno-portable-runtime} options together under HP-UX with
15295 It is normally not desirable to use this option as it degrades
15296 performance. However, it may be useful in large applications,
15297 particularly when partial linking is used to build the application.
15299 The types of long calls used depends on the capabilities of the
15300 assembler and linker, and the type of code being generated. The
15301 impact on systems that support long absolute calls, and long pic
15302 symbol-difference or pc-relative calls should be relatively small.
15303 However, an indirect call is used on 32-bit ELF systems in pic code
15304 and it is quite long.
15306 @item -munix=@var{unix-std}
15308 Generate compiler predefines and select a startfile for the specified
15309 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15310 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15311 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15312 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15313 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15316 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15317 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15318 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15319 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15320 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15321 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15323 It is @emph{important} to note that this option changes the interfaces
15324 for various library routines. It also affects the operational behavior
15325 of the C library. Thus, @emph{extreme} care is needed in using this
15328 Library code that is intended to operate with more than one UNIX
15329 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15330 as appropriate. Most GNU software doesn't provide this capability.
15334 Suppress the generation of link options to search libdld.sl when the
15335 @option{-static} option is specified on HP-UX 10 and later.
15339 The HP-UX implementation of setlocale in libc has a dependency on
15340 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15341 when the @option{-static} option is specified, special link options
15342 are needed to resolve this dependency.
15344 On HP-UX 10 and later, the GCC driver adds the necessary options to
15345 link with libdld.sl when the @option{-static} option is specified.
15346 This causes the resulting binary to be dynamic. On the 64-bit port,
15347 the linkers generate dynamic binaries by default in any case. The
15348 @option{-nolibdld} option can be used to prevent the GCC driver from
15349 adding these link options.
15353 Add support for multithreading with the @dfn{dce thread} library
15354 under HP-UX@. This option sets flags for both the preprocessor and
15358 @node i386 and x86-64 Options
15359 @subsection Intel 386 and AMD x86-64 Options
15360 @cindex i386 Options
15361 @cindex x86-64 Options
15362 @cindex Intel 386 Options
15363 @cindex AMD x86-64 Options
15365 These @samp{-m} options are defined for the i386 and x86-64 family of
15370 @item -march=@var{cpu-type}
15372 Generate instructions for the machine type @var{cpu-type}. In contrast to
15373 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
15374 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
15375 to generate code that may not run at all on processors other than the one
15376 indicated. Specifying @option{-march=@var{cpu-type}} implies
15377 @option{-mtune=@var{cpu-type}}.
15379 The choices for @var{cpu-type} are:
15383 This selects the CPU to generate code for at compilation time by determining
15384 the processor type of the compiling machine. Using @option{-march=native}
15385 enables all instruction subsets supported by the local machine (hence
15386 the result might not run on different machines). Using @option{-mtune=native}
15387 produces code optimized for the local machine under the constraints
15388 of the selected instruction set.
15391 Original Intel i386 CPU@.
15394 Intel i486 CPU@. (No scheduling is implemented for this chip.)
15398 Intel Pentium CPU with no MMX support.
15401 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
15404 Intel Pentium Pro CPU@.
15407 When used with @option{-march}, the Pentium Pro
15408 instruction set is used, so the code runs on all i686 family chips.
15409 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
15412 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
15417 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
15421 Intel Pentium M; low-power version of Intel Pentium III CPU
15422 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
15426 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
15429 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
15433 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
15434 SSE2 and SSE3 instruction set support.
15437 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
15438 instruction set support.
15441 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15442 SSE4.1, SSE4.2 and POPCNT instruction set support.
15445 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15446 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
15449 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15450 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
15453 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
15454 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
15455 instruction set support.
15458 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15459 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
15460 BMI, BMI2 and F16C instruction set support.
15463 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15464 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
15465 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
15468 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
15469 instruction set support.
15472 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
15473 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
15476 AMD K6 CPU with MMX instruction set support.
15480 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
15483 @itemx athlon-tbird
15484 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
15490 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
15491 instruction set support.
15497 Processors based on the AMD K8 core with x86-64 instruction set support,
15498 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
15499 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
15500 instruction set extensions.)
15503 @itemx opteron-sse3
15504 @itemx athlon64-sse3
15505 Improved versions of AMD K8 cores with SSE3 instruction set support.
15509 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
15510 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
15511 instruction set extensions.)
15514 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
15515 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
15516 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
15518 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15519 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
15520 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
15523 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15524 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
15525 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
15526 64-bit instruction set extensions.
15528 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
15529 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
15530 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
15531 SSE4.2, ABM and 64-bit instruction set extensions.
15534 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
15535 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
15536 instruction set extensions.)
15539 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
15540 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
15541 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
15544 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
15548 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
15549 instruction set support.
15552 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
15553 implemented for this chip.)
15556 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
15558 implemented for this chip.)
15561 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
15564 @item -mtune=@var{cpu-type}
15566 Tune to @var{cpu-type} everything applicable about the generated code, except
15567 for the ABI and the set of available instructions.
15568 While picking a specific @var{cpu-type} schedules things appropriately
15569 for that particular chip, the compiler does not generate any code that
15570 cannot run on the default machine type unless you use a
15571 @option{-march=@var{cpu-type}} option.
15572 For example, if GCC is configured for i686-pc-linux-gnu
15573 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
15574 but still runs on i686 machines.
15576 The choices for @var{cpu-type} are the same as for @option{-march}.
15577 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
15581 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
15582 If you know the CPU on which your code will run, then you should use
15583 the corresponding @option{-mtune} or @option{-march} option instead of
15584 @option{-mtune=generic}. But, if you do not know exactly what CPU users
15585 of your application will have, then you should use this option.
15587 As new processors are deployed in the marketplace, the behavior of this
15588 option will change. Therefore, if you upgrade to a newer version of
15589 GCC, code generation controlled by this option will change to reflect
15591 that are most common at the time that version of GCC is released.
15593 There is no @option{-march=generic} option because @option{-march}
15594 indicates the instruction set the compiler can use, and there is no
15595 generic instruction set applicable to all processors. In contrast,
15596 @option{-mtune} indicates the processor (or, in this case, collection of
15597 processors) for which the code is optimized.
15600 Produce code optimized for the most current Intel processors, which are
15601 Haswell and Silvermont for this version of GCC. If you know the CPU
15602 on which your code will run, then you should use the corresponding
15603 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
15604 But, if you want your application performs better on both Haswell and
15605 Silvermont, then you should use this option.
15607 As new Intel processors are deployed in the marketplace, the behavior of
15608 this option will change. Therefore, if you upgrade to a newer version of
15609 GCC, code generation controlled by this option will change to reflect
15610 the most current Intel processors at the time that version of GCC is
15613 There is no @option{-march=intel} option because @option{-march} indicates
15614 the instruction set the compiler can use, and there is no common
15615 instruction set applicable to all processors. In contrast,
15616 @option{-mtune} indicates the processor (or, in this case, collection of
15617 processors) for which the code is optimized.
15620 @item -mcpu=@var{cpu-type}
15622 A deprecated synonym for @option{-mtune}.
15624 @item -mfpmath=@var{unit}
15626 Generate floating-point arithmetic for selected unit @var{unit}. The choices
15627 for @var{unit} are:
15631 Use the standard 387 floating-point coprocessor present on the majority of chips and
15632 emulated otherwise. Code compiled with this option runs almost everywhere.
15633 The temporary results are computed in 80-bit precision instead of the precision
15634 specified by the type, resulting in slightly different results compared to most
15635 of other chips. See @option{-ffloat-store} for more detailed description.
15637 This is the default choice for i386 compiler.
15640 Use scalar floating-point instructions present in the SSE instruction set.
15641 This instruction set is supported by Pentium III and newer chips,
15642 and in the AMD line
15643 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
15644 instruction set supports only single-precision arithmetic, thus the double and
15645 extended-precision arithmetic are still done using 387. A later version, present
15646 only in Pentium 4 and AMD x86-64 chips, supports double-precision
15649 For the i386 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
15650 or @option{-msse2} switches to enable SSE extensions and make this option
15651 effective. For the x86-64 compiler, these extensions are enabled by default.
15653 The resulting code should be considerably faster in the majority of cases and avoid
15654 the numerical instability problems of 387 code, but may break some existing
15655 code that expects temporaries to be 80 bits.
15657 This is the default choice for the x86-64 compiler.
15662 Attempt to utilize both instruction sets at once. This effectively doubles the
15663 amount of available registers, and on chips with separate execution units for
15664 387 and SSE the execution resources too. Use this option with care, as it is
15665 still experimental, because the GCC register allocator does not model separate
15666 functional units well, resulting in unstable performance.
15669 @item -masm=@var{dialect}
15670 @opindex masm=@var{dialect}
15671 Output assembly instructions using selected @var{dialect}. Supported
15672 choices are @samp{intel} or @samp{att} (the default). Darwin does
15673 not support @samp{intel}.
15676 @itemx -mno-ieee-fp
15678 @opindex mno-ieee-fp
15679 Control whether or not the compiler uses IEEE floating-point
15680 comparisons. These correctly handle the case where the result of a
15681 comparison is unordered.
15684 @opindex msoft-float
15685 Generate output containing library calls for floating point.
15687 @strong{Warning:} the requisite libraries are not part of GCC@.
15688 Normally the facilities of the machine's usual C compiler are used, but
15689 this can't be done directly in cross-compilation. You must make your
15690 own arrangements to provide suitable library functions for
15693 On machines where a function returns floating-point results in the 80387
15694 register stack, some floating-point opcodes may be emitted even if
15695 @option{-msoft-float} is used.
15697 @item -mno-fp-ret-in-387
15698 @opindex mno-fp-ret-in-387
15699 Do not use the FPU registers for return values of functions.
15701 The usual calling convention has functions return values of types
15702 @code{float} and @code{double} in an FPU register, even if there
15703 is no FPU@. The idea is that the operating system should emulate
15706 The option @option{-mno-fp-ret-in-387} causes such values to be returned
15707 in ordinary CPU registers instead.
15709 @item -mno-fancy-math-387
15710 @opindex mno-fancy-math-387
15711 Some 387 emulators do not support the @code{sin}, @code{cos} and
15712 @code{sqrt} instructions for the 387. Specify this option to avoid
15713 generating those instructions. This option is the default on FreeBSD,
15714 OpenBSD and NetBSD@. This option is overridden when @option{-march}
15715 indicates that the target CPU always has an FPU and so the
15716 instruction does not need emulation. These
15717 instructions are not generated unless you also use the
15718 @option{-funsafe-math-optimizations} switch.
15720 @item -malign-double
15721 @itemx -mno-align-double
15722 @opindex malign-double
15723 @opindex mno-align-double
15724 Control whether GCC aligns @code{double}, @code{long double}, and
15725 @code{long long} variables on a two-word boundary or a one-word
15726 boundary. Aligning @code{double} variables on a two-word boundary
15727 produces code that runs somewhat faster on a Pentium at the
15728 expense of more memory.
15730 On x86-64, @option{-malign-double} is enabled by default.
15732 @strong{Warning:} if you use the @option{-malign-double} switch,
15733 structures containing the above types are aligned differently than
15734 the published application binary interface specifications for the 386
15735 and are not binary compatible with structures in code compiled
15736 without that switch.
15738 @item -m96bit-long-double
15739 @itemx -m128bit-long-double
15740 @opindex m96bit-long-double
15741 @opindex m128bit-long-double
15742 These switches control the size of @code{long double} type. The i386
15743 application binary interface specifies the size to be 96 bits,
15744 so @option{-m96bit-long-double} is the default in 32-bit mode.
15746 Modern architectures (Pentium and newer) prefer @code{long double}
15747 to be aligned to an 8- or 16-byte boundary. In arrays or structures
15748 conforming to the ABI, this is not possible. So specifying
15749 @option{-m128bit-long-double} aligns @code{long double}
15750 to a 16-byte boundary by padding the @code{long double} with an additional
15753 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
15754 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
15756 Notice that neither of these options enable any extra precision over the x87
15757 standard of 80 bits for a @code{long double}.
15759 @strong{Warning:} if you override the default value for your target ABI, this
15760 changes the size of
15761 structures and arrays containing @code{long double} variables,
15762 as well as modifying the function calling convention for functions taking
15763 @code{long double}. Hence they are not binary-compatible
15764 with code compiled without that switch.
15766 @item -mlong-double-64
15767 @itemx -mlong-double-80
15768 @itemx -mlong-double-128
15769 @opindex mlong-double-64
15770 @opindex mlong-double-80
15771 @opindex mlong-double-128
15772 These switches control the size of @code{long double} type. A size
15773 of 64 bits makes the @code{long double} type equivalent to the @code{double}
15774 type. This is the default for 32-bit Bionic C library. A size
15775 of 128 bits makes the @code{long double} type equivalent to the
15776 @code{__float128} type. This is the default for 64-bit Bionic C library.
15778 @strong{Warning:} if you override the default value for your target ABI, this
15779 changes the size of
15780 structures and arrays containing @code{long double} variables,
15781 as well as modifying the function calling convention for functions taking
15782 @code{long double}. Hence they are not binary-compatible
15783 with code compiled without that switch.
15785 @item -malign-data=@var{type}
15786 @opindex malign-data
15787 Control how GCC aligns variables. Supported values for @var{type} are
15788 @samp{compat} uses increased alignment value compatible uses GCC 4.8
15789 and earlier, @samp{abi} uses alignment value as specified by the
15790 psABI, and @samp{cacheline} uses increased alignment value to match
15791 the cache line size. @samp{compat} is the default.
15793 @item -mlarge-data-threshold=@var{threshold}
15794 @opindex mlarge-data-threshold
15795 When @option{-mcmodel=medium} is specified, data objects larger than
15796 @var{threshold} are placed in the large data section. This value must be the
15797 same across all objects linked into the binary, and defaults to 65535.
15801 Use a different function-calling convention, in which functions that
15802 take a fixed number of arguments return with the @code{ret @var{num}}
15803 instruction, which pops their arguments while returning. This saves one
15804 instruction in the caller since there is no need to pop the arguments
15807 You can specify that an individual function is called with this calling
15808 sequence with the function attribute @code{stdcall}. You can also
15809 override the @option{-mrtd} option by using the function attribute
15810 @code{cdecl}. @xref{Function Attributes}.
15812 @strong{Warning:} this calling convention is incompatible with the one
15813 normally used on Unix, so you cannot use it if you need to call
15814 libraries compiled with the Unix compiler.
15816 Also, you must provide function prototypes for all functions that
15817 take variable numbers of arguments (including @code{printf});
15818 otherwise incorrect code is generated for calls to those
15821 In addition, seriously incorrect code results if you call a
15822 function with too many arguments. (Normally, extra arguments are
15823 harmlessly ignored.)
15825 @item -mregparm=@var{num}
15827 Control how many registers are used to pass integer arguments. By
15828 default, no registers are used to pass arguments, and at most 3
15829 registers can be used. You can control this behavior for a specific
15830 function by using the function attribute @code{regparm}.
15831 @xref{Function Attributes}.
15833 @strong{Warning:} if you use this switch, and
15834 @var{num} is nonzero, then you must build all modules with the same
15835 value, including any libraries. This includes the system libraries and
15839 @opindex msseregparm
15840 Use SSE register passing conventions for float and double arguments
15841 and return values. You can control this behavior for a specific
15842 function by using the function attribute @code{sseregparm}.
15843 @xref{Function Attributes}.
15845 @strong{Warning:} if you use this switch then you must build all
15846 modules with the same value, including any libraries. This includes
15847 the system libraries and startup modules.
15849 @item -mvect8-ret-in-mem
15850 @opindex mvect8-ret-in-mem
15851 Return 8-byte vectors in memory instead of MMX registers. This is the
15852 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
15853 Studio compilers until version 12. Later compiler versions (starting
15854 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
15855 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
15856 you need to remain compatible with existing code produced by those
15857 previous compiler versions or older versions of GCC@.
15866 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
15867 is specified, the significands of results of floating-point operations are
15868 rounded to 24 bits (single precision); @option{-mpc64} rounds the
15869 significands of results of floating-point operations to 53 bits (double
15870 precision) and @option{-mpc80} rounds the significands of results of
15871 floating-point operations to 64 bits (extended double precision), which is
15872 the default. When this option is used, floating-point operations in higher
15873 precisions are not available to the programmer without setting the FPU
15874 control word explicitly.
15876 Setting the rounding of floating-point operations to less than the default
15877 80 bits can speed some programs by 2% or more. Note that some mathematical
15878 libraries assume that extended-precision (80-bit) floating-point operations
15879 are enabled by default; routines in such libraries could suffer significant
15880 loss of accuracy, typically through so-called ``catastrophic cancellation'',
15881 when this option is used to set the precision to less than extended precision.
15883 @item -mstackrealign
15884 @opindex mstackrealign
15885 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
15886 option generates an alternate prologue and epilogue that realigns the
15887 run-time stack if necessary. This supports mixing legacy codes that keep
15888 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
15889 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
15890 applicable to individual functions.
15892 @item -mpreferred-stack-boundary=@var{num}
15893 @opindex mpreferred-stack-boundary
15894 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
15895 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
15896 the default is 4 (16 bytes or 128 bits).
15898 @strong{Warning:} When generating code for the x86-64 architecture with
15899 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
15900 used to keep the stack boundary aligned to 8 byte boundary. Since
15901 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
15902 intended to be used in controlled environment where stack space is
15903 important limitation. This option leads to wrong code when functions
15904 compiled with 16 byte stack alignment (such as functions from a standard
15905 library) are called with misaligned stack. In this case, SSE
15906 instructions may lead to misaligned memory access traps. In addition,
15907 variable arguments are handled incorrectly for 16 byte aligned
15908 objects (including x87 long double and __int128), leading to wrong
15909 results. You must build all modules with
15910 @option{-mpreferred-stack-boundary=3}, including any libraries. This
15911 includes the system libraries and startup modules.
15913 @item -mincoming-stack-boundary=@var{num}
15914 @opindex mincoming-stack-boundary
15915 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
15916 boundary. If @option{-mincoming-stack-boundary} is not specified,
15917 the one specified by @option{-mpreferred-stack-boundary} is used.
15919 On Pentium and Pentium Pro, @code{double} and @code{long double} values
15920 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
15921 suffer significant run time performance penalties. On Pentium III, the
15922 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
15923 properly if it is not 16-byte aligned.
15925 To ensure proper alignment of this values on the stack, the stack boundary
15926 must be as aligned as that required by any value stored on the stack.
15927 Further, every function must be generated such that it keeps the stack
15928 aligned. Thus calling a function compiled with a higher preferred
15929 stack boundary from a function compiled with a lower preferred stack
15930 boundary most likely misaligns the stack. It is recommended that
15931 libraries that use callbacks always use the default setting.
15933 This extra alignment does consume extra stack space, and generally
15934 increases code size. Code that is sensitive to stack space usage, such
15935 as embedded systems and operating system kernels, may want to reduce the
15936 preferred alignment to @option{-mpreferred-stack-boundary=2}.
15982 @opindex mclfushopt
16000 @itemx -mprefetchwt1
16001 @opindex mprefetchwt1
16049 These switches enable the use of instructions in the MMX, SSE,
16050 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
16051 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
16052 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX or 3DNow!@:
16053 extended instruction sets. Each has a corresponding @option{-mno-} option
16054 to disable use of these instructions.
16056 These extensions are also available as built-in functions: see
16057 @ref{X86 Built-in Functions}, for details of the functions enabled and
16058 disabled by these switches.
16060 To generate SSE/SSE2 instructions automatically from floating-point
16061 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
16063 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
16064 generates new AVX instructions or AVX equivalence for all SSEx instructions
16067 These options enable GCC to use these extended instructions in
16068 generated code, even without @option{-mfpmath=sse}. Applications that
16069 perform run-time CPU detection must compile separate files for each
16070 supported architecture, using the appropriate flags. In particular,
16071 the file containing the CPU detection code should be compiled without
16074 @item -mdump-tune-features
16075 @opindex mdump-tune-features
16076 This option instructs GCC to dump the names of the x86 performance
16077 tuning features and default settings. The names can be used in
16078 @option{-mtune-ctrl=@var{feature-list}}.
16080 @item -mtune-ctrl=@var{feature-list}
16081 @opindex mtune-ctrl=@var{feature-list}
16082 This option is used to do fine grain control of x86 code generation features.
16083 @var{feature-list} is a comma separated list of @var{feature} names. See also
16084 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
16085 on if it is not preceded with @samp{^}, otherwise, it is turned off.
16086 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
16087 developers. Using it may lead to code paths not covered by testing and can
16088 potentially result in compiler ICEs or runtime errors.
16091 @opindex mno-default
16092 This option instructs GCC to turn off all tunable features. See also
16093 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
16097 This option instructs GCC to emit a @code{cld} instruction in the prologue
16098 of functions that use string instructions. String instructions depend on
16099 the DF flag to select between autoincrement or autodecrement mode. While the
16100 ABI specifies the DF flag to be cleared on function entry, some operating
16101 systems violate this specification by not clearing the DF flag in their
16102 exception dispatchers. The exception handler can be invoked with the DF flag
16103 set, which leads to wrong direction mode when string instructions are used.
16104 This option can be enabled by default on 32-bit x86 targets by configuring
16105 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
16106 instructions can be suppressed with the @option{-mno-cld} compiler option
16110 @opindex mvzeroupper
16111 This option instructs GCC to emit a @code{vzeroupper} instruction
16112 before a transfer of control flow out of the function to minimize
16113 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
16116 @item -mprefer-avx128
16117 @opindex mprefer-avx128
16118 This option instructs GCC to use 128-bit AVX instructions instead of
16119 256-bit AVX instructions in the auto-vectorizer.
16123 This option enables GCC to generate @code{CMPXCHG16B} instructions.
16124 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
16125 (or oword) data types.
16126 This is useful for high-resolution counters that can be updated
16127 by multiple processors (or cores). This instruction is generated as part of
16128 atomic built-in functions: see @ref{__sync Builtins} or
16129 @ref{__atomic Builtins} for details.
16133 This option enables generation of @code{SAHF} instructions in 64-bit code.
16134 Early Intel Pentium 4 CPUs with Intel 64 support,
16135 prior to the introduction of Pentium 4 G1 step in December 2005,
16136 lacked the @code{LAHF} and @code{SAHF} instructions
16137 which are supported by AMD64.
16138 These are load and store instructions, respectively, for certain status flags.
16139 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
16140 @code{drem}, and @code{remainder} built-in functions;
16141 see @ref{Other Builtins} for details.
16145 This option enables use of the @code{movbe} instruction to implement
16146 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
16150 This option enables built-in functions @code{__builtin_ia32_crc32qi},
16151 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
16152 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
16156 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
16157 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
16158 with an additional Newton-Raphson step
16159 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
16160 (and their vectorized
16161 variants) for single-precision floating-point arguments. These instructions
16162 are generated only when @option{-funsafe-math-optimizations} is enabled
16163 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
16164 Note that while the throughput of the sequence is higher than the throughput
16165 of the non-reciprocal instruction, the precision of the sequence can be
16166 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
16168 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
16169 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
16170 combination), and doesn't need @option{-mrecip}.
16172 Also note that GCC emits the above sequence with additional Newton-Raphson step
16173 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
16174 already with @option{-ffast-math} (or the above option combination), and
16175 doesn't need @option{-mrecip}.
16177 @item -mrecip=@var{opt}
16178 @opindex mrecip=opt
16179 This option controls which reciprocal estimate instructions
16180 may be used. @var{opt} is a comma-separated list of options, which may
16181 be preceded by a @samp{!} to invert the option:
16185 Enable all estimate instructions.
16188 Enable the default instructions, equivalent to @option{-mrecip}.
16191 Disable all estimate instructions, equivalent to @option{-mno-recip}.
16194 Enable the approximation for scalar division.
16197 Enable the approximation for vectorized division.
16200 Enable the approximation for scalar square root.
16203 Enable the approximation for vectorized square root.
16206 So, for example, @option{-mrecip=all,!sqrt} enables
16207 all of the reciprocal approximations, except for square root.
16209 @item -mveclibabi=@var{type}
16210 @opindex mveclibabi
16211 Specifies the ABI type to use for vectorizing intrinsics using an
16212 external library. Supported values for @var{type} are @samp{svml}
16213 for the Intel short
16214 vector math library and @samp{acml} for the AMD math core library.
16215 To use this option, both @option{-ftree-vectorize} and
16216 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
16217 ABI-compatible library must be specified at link time.
16219 GCC currently emits calls to @code{vmldExp2},
16220 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
16221 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
16222 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
16223 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
16224 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
16225 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
16226 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
16227 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
16228 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
16229 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
16230 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
16231 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
16232 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
16233 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
16234 when @option{-mveclibabi=acml} is used.
16236 @item -mabi=@var{name}
16238 Generate code for the specified calling convention. Permissible values
16239 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
16240 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
16241 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
16242 You can control this behavior for specific functions by
16243 using the function attributes @code{ms_abi} and @code{sysv_abi}.
16244 @xref{Function Attributes}.
16246 @item -mtls-dialect=@var{type}
16247 @opindex mtls-dialect
16248 Generate code to access thread-local storage using the @samp{gnu} or
16249 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
16250 @samp{gnu2} is more efficient, but it may add compile- and run-time
16251 requirements that cannot be satisfied on all systems.
16254 @itemx -mno-push-args
16255 @opindex mpush-args
16256 @opindex mno-push-args
16257 Use PUSH operations to store outgoing parameters. This method is shorter
16258 and usually equally fast as method using SUB/MOV operations and is enabled
16259 by default. In some cases disabling it may improve performance because of
16260 improved scheduling and reduced dependencies.
16262 @item -maccumulate-outgoing-args
16263 @opindex maccumulate-outgoing-args
16264 If enabled, the maximum amount of space required for outgoing arguments is
16265 computed in the function prologue. This is faster on most modern CPUs
16266 because of reduced dependencies, improved scheduling and reduced stack usage
16267 when the preferred stack boundary is not equal to 2. The drawback is a notable
16268 increase in code size. This switch implies @option{-mno-push-args}.
16272 Support thread-safe exception handling on MinGW. Programs that rely
16273 on thread-safe exception handling must compile and link all code with the
16274 @option{-mthreads} option. When compiling, @option{-mthreads} defines
16275 @option{-D_MT}; when linking, it links in a special thread helper library
16276 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
16278 @item -mno-align-stringops
16279 @opindex mno-align-stringops
16280 Do not align the destination of inlined string operations. This switch reduces
16281 code size and improves performance in case the destination is already aligned,
16282 but GCC doesn't know about it.
16284 @item -minline-all-stringops
16285 @opindex minline-all-stringops
16286 By default GCC inlines string operations only when the destination is
16287 known to be aligned to least a 4-byte boundary.
16288 This enables more inlining and increases code
16289 size, but may improve performance of code that depends on fast
16290 @code{memcpy}, @code{strlen},
16291 and @code{memset} for short lengths.
16293 @item -minline-stringops-dynamically
16294 @opindex minline-stringops-dynamically
16295 For string operations of unknown size, use run-time checks with
16296 inline code for small blocks and a library call for large blocks.
16298 @item -mstringop-strategy=@var{alg}
16299 @opindex mstringop-strategy=@var{alg}
16300 Override the internal decision heuristic for the particular algorithm to use
16301 for inlining string operations. The allowed values for @var{alg} are:
16307 Expand using i386 @code{rep} prefix of the specified size.
16311 @itemx unrolled_loop
16312 Expand into an inline loop.
16315 Always use a library call.
16318 @item -mmemcpy-strategy=@var{strategy}
16319 @opindex mmemcpy-strategy=@var{strategy}
16320 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
16321 should be inlined and what inline algorithm to use when the expected size
16322 of the copy operation is known. @var{strategy}
16323 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
16324 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
16325 the max byte size with which inline algorithm @var{alg} is allowed. For the last
16326 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
16327 in the list must be specified in increasing order. The minimal byte size for
16328 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
16331 @item -mmemset-strategy=@var{strategy}
16332 @opindex mmemset-strategy=@var{strategy}
16333 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
16334 @code{__builtin_memset} expansion.
16336 @item -momit-leaf-frame-pointer
16337 @opindex momit-leaf-frame-pointer
16338 Don't keep the frame pointer in a register for leaf functions. This
16339 avoids the instructions to save, set up, and restore frame pointers and
16340 makes an extra register available in leaf functions. The option
16341 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
16342 which might make debugging harder.
16344 @item -mtls-direct-seg-refs
16345 @itemx -mno-tls-direct-seg-refs
16346 @opindex mtls-direct-seg-refs
16347 Controls whether TLS variables may be accessed with offsets from the
16348 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
16349 or whether the thread base pointer must be added. Whether or not this
16350 is valid depends on the operating system, and whether it maps the
16351 segment to cover the entire TLS area.
16353 For systems that use the GNU C Library, the default is on.
16356 @itemx -mno-sse2avx
16358 Specify that the assembler should encode SSE instructions with VEX
16359 prefix. The option @option{-mavx} turns this on by default.
16364 If profiling is active (@option{-pg}), put the profiling
16365 counter call before the prologue.
16366 Note: On x86 architectures the attribute @code{ms_hook_prologue}
16367 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
16369 @item -mrecord-mcount
16370 @itemx -mno-record-mcount
16371 @opindex mrecord-mcount
16372 If profiling is active (@option{-pg}), generate a __mcount_loc section
16373 that contains pointers to each profiling call. This is useful for
16374 automatically patching and out calls.
16377 @itemx -mno-nop-mcount
16378 @opindex mnop-mcount
16379 If profiling is active (@option{-pg}), generate the calls to
16380 the profiling functions as nops. This is useful when they
16381 should be patched in later dynamically. This is likely only
16382 useful together with @option{-mrecord-mcount}.
16384 @item -mskip-rax-setup
16385 @itemx -mno-skip-rax-setup
16386 @opindex mskip-rax-setup
16387 When generating code for the x86-64 architecture with SSE extensions
16388 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
16389 register when there are no variable arguments passed in vector registers.
16391 @strong{Warning:} Since RAX register is used to avoid unnecessarily
16392 saving vector registers on stack when passing variable arguments, the
16393 impacts of this option are callees may waste some stack space,
16394 misbehave or jump to a random location. GCC 4.4 or newer don't have
16395 those issues, regardless the RAX register value.
16398 @itemx -mno-8bit-idiv
16399 @opindex m8bit-idiv
16400 On some processors, like Intel Atom, 8-bit unsigned integer divide is
16401 much faster than 32-bit/64-bit integer divide. This option generates a
16402 run-time check. If both dividend and divisor are within range of 0
16403 to 255, 8-bit unsigned integer divide is used instead of
16404 32-bit/64-bit integer divide.
16406 @item -mavx256-split-unaligned-load
16407 @itemx -mavx256-split-unaligned-store
16408 @opindex mavx256-split-unaligned-load
16409 @opindex mavx256-split-unaligned-store
16410 Split 32-byte AVX unaligned load and store.
16412 @item -mstack-protector-guard=@var{guard}
16413 @opindex mstack-protector-guard=@var{guard}
16414 Generate stack protection code using canary at @var{guard}. Supported
16415 locations are @samp{global} for global canary or @samp{tls} for per-thread
16416 canary in the TLS block (the default). This option has effect only when
16417 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
16421 These @samp{-m} switches are supported in addition to the above
16422 on x86-64 processors in 64-bit environments.
16433 Generate code for a 16-bit, 32-bit or 64-bit environment.
16434 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
16436 generates code that runs on any i386 system.
16438 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
16439 types to 64 bits, and generates code for the x86-64 architecture.
16440 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
16441 and @option{-mdynamic-no-pic} options.
16443 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
16445 generates code for the x86-64 architecture.
16447 The @option{-m16} option is the same as @option{-m32}, except for that
16448 it outputs the @code{.code16gcc} assembly directive at the beginning of
16449 the assembly output so that the binary can run in 16-bit mode.
16451 @item -mno-red-zone
16452 @opindex mno-red-zone
16453 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
16454 by the x86-64 ABI; it is a 128-byte area beyond the location of the
16455 stack pointer that is not modified by signal or interrupt handlers
16456 and therefore can be used for temporary data without adjusting the stack
16457 pointer. The flag @option{-mno-red-zone} disables this red zone.
16459 @item -mcmodel=small
16460 @opindex mcmodel=small
16461 Generate code for the small code model: the program and its symbols must
16462 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
16463 Programs can be statically or dynamically linked. This is the default
16466 @item -mcmodel=kernel
16467 @opindex mcmodel=kernel
16468 Generate code for the kernel code model. The kernel runs in the
16469 negative 2 GB of the address space.
16470 This model has to be used for Linux kernel code.
16472 @item -mcmodel=medium
16473 @opindex mcmodel=medium
16474 Generate code for the medium model: the program is linked in the lower 2
16475 GB of the address space. Small symbols are also placed there. Symbols
16476 with sizes larger than @option{-mlarge-data-threshold} are put into
16477 large data or BSS sections and can be located above 2GB. Programs can
16478 be statically or dynamically linked.
16480 @item -mcmodel=large
16481 @opindex mcmodel=large
16482 Generate code for the large model. This model makes no assumptions
16483 about addresses and sizes of sections.
16485 @item -maddress-mode=long
16486 @opindex maddress-mode=long
16487 Generate code for long address mode. This is only supported for 64-bit
16488 and x32 environments. It is the default address mode for 64-bit
16491 @item -maddress-mode=short
16492 @opindex maddress-mode=short
16493 Generate code for short address mode. This is only supported for 32-bit
16494 and x32 environments. It is the default address mode for 32-bit and
16498 @node i386 and x86-64 Windows Options
16499 @subsection i386 and x86-64 Windows Options
16500 @cindex i386 and x86-64 Windows Options
16502 These additional options are available for Microsoft Windows targets:
16508 specifies that a console application is to be generated, by
16509 instructing the linker to set the PE header subsystem type
16510 required for console applications.
16511 This option is available for Cygwin and MinGW targets and is
16512 enabled by default on those targets.
16516 This option is available for Cygwin and MinGW targets. It
16517 specifies that a DLL---a dynamic link library---is to be
16518 generated, enabling the selection of the required runtime
16519 startup object and entry point.
16521 @item -mnop-fun-dllimport
16522 @opindex mnop-fun-dllimport
16523 This option is available for Cygwin and MinGW targets. It
16524 specifies that the @code{dllimport} attribute should be ignored.
16528 This option is available for MinGW targets. It specifies
16529 that MinGW-specific thread support is to be used.
16533 This option is available for MinGW-w64 targets. It causes
16534 the @code{UNICODE} preprocessor macro to be predefined, and
16535 chooses Unicode-capable runtime startup code.
16539 This option is available for Cygwin and MinGW targets. It
16540 specifies that the typical Microsoft Windows predefined macros are to
16541 be set in the pre-processor, but does not influence the choice
16542 of runtime library/startup code.
16546 This option is available for Cygwin and MinGW targets. It
16547 specifies that a GUI application is to be generated by
16548 instructing the linker to set the PE header subsystem type
16551 @item -fno-set-stack-executable
16552 @opindex fno-set-stack-executable
16553 This option is available for MinGW targets. It specifies that
16554 the executable flag for the stack used by nested functions isn't
16555 set. This is necessary for binaries running in kernel mode of
16556 Microsoft Windows, as there the User32 API, which is used to set executable
16557 privileges, isn't available.
16559 @item -fwritable-relocated-rdata
16560 @opindex fno-writable-relocated-rdata
16561 This option is available for MinGW and Cygwin targets. It specifies
16562 that relocated-data in read-only section is put into .data
16563 section. This is a necessary for older runtimes not supporting
16564 modification of .rdata sections for pseudo-relocation.
16566 @item -mpe-aligned-commons
16567 @opindex mpe-aligned-commons
16568 This option is available for Cygwin and MinGW targets. It
16569 specifies that the GNU extension to the PE file format that
16570 permits the correct alignment of COMMON variables should be
16571 used when generating code. It is enabled by default if
16572 GCC detects that the target assembler found during configuration
16573 supports the feature.
16576 See also under @ref{i386 and x86-64 Options} for standard options.
16578 @node IA-64 Options
16579 @subsection IA-64 Options
16580 @cindex IA-64 Options
16582 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16586 @opindex mbig-endian
16587 Generate code for a big-endian target. This is the default for HP-UX@.
16589 @item -mlittle-endian
16590 @opindex mlittle-endian
16591 Generate code for a little-endian target. This is the default for AIX5
16597 @opindex mno-gnu-as
16598 Generate (or don't) code for the GNU assembler. This is the default.
16599 @c Also, this is the default if the configure option @option{--with-gnu-as}
16605 @opindex mno-gnu-ld
16606 Generate (or don't) code for the GNU linker. This is the default.
16607 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16612 Generate code that does not use a global pointer register. The result
16613 is not position independent code, and violates the IA-64 ABI@.
16615 @item -mvolatile-asm-stop
16616 @itemx -mno-volatile-asm-stop
16617 @opindex mvolatile-asm-stop
16618 @opindex mno-volatile-asm-stop
16619 Generate (or don't) a stop bit immediately before and after volatile asm
16622 @item -mregister-names
16623 @itemx -mno-register-names
16624 @opindex mregister-names
16625 @opindex mno-register-names
16626 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16627 the stacked registers. This may make assembler output more readable.
16633 Disable (or enable) optimizations that use the small data section. This may
16634 be useful for working around optimizer bugs.
16636 @item -mconstant-gp
16637 @opindex mconstant-gp
16638 Generate code that uses a single constant global pointer value. This is
16639 useful when compiling kernel code.
16643 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16644 This is useful when compiling firmware code.
16646 @item -minline-float-divide-min-latency
16647 @opindex minline-float-divide-min-latency
16648 Generate code for inline divides of floating-point values
16649 using the minimum latency algorithm.
16651 @item -minline-float-divide-max-throughput
16652 @opindex minline-float-divide-max-throughput
16653 Generate code for inline divides of floating-point values
16654 using the maximum throughput algorithm.
16656 @item -mno-inline-float-divide
16657 @opindex mno-inline-float-divide
16658 Do not generate inline code for divides of floating-point values.
16660 @item -minline-int-divide-min-latency
16661 @opindex minline-int-divide-min-latency
16662 Generate code for inline divides of integer values
16663 using the minimum latency algorithm.
16665 @item -minline-int-divide-max-throughput
16666 @opindex minline-int-divide-max-throughput
16667 Generate code for inline divides of integer values
16668 using the maximum throughput algorithm.
16670 @item -mno-inline-int-divide
16671 @opindex mno-inline-int-divide
16672 Do not generate inline code for divides of integer values.
16674 @item -minline-sqrt-min-latency
16675 @opindex minline-sqrt-min-latency
16676 Generate code for inline square roots
16677 using the minimum latency algorithm.
16679 @item -minline-sqrt-max-throughput
16680 @opindex minline-sqrt-max-throughput
16681 Generate code for inline square roots
16682 using the maximum throughput algorithm.
16684 @item -mno-inline-sqrt
16685 @opindex mno-inline-sqrt
16686 Do not generate inline code for @code{sqrt}.
16689 @itemx -mno-fused-madd
16690 @opindex mfused-madd
16691 @opindex mno-fused-madd
16692 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16693 instructions. The default is to use these instructions.
16695 @item -mno-dwarf2-asm
16696 @itemx -mdwarf2-asm
16697 @opindex mno-dwarf2-asm
16698 @opindex mdwarf2-asm
16699 Don't (or do) generate assembler code for the DWARF 2 line number debugging
16700 info. This may be useful when not using the GNU assembler.
16702 @item -mearly-stop-bits
16703 @itemx -mno-early-stop-bits
16704 @opindex mearly-stop-bits
16705 @opindex mno-early-stop-bits
16706 Allow stop bits to be placed earlier than immediately preceding the
16707 instruction that triggered the stop bit. This can improve instruction
16708 scheduling, but does not always do so.
16710 @item -mfixed-range=@var{register-range}
16711 @opindex mfixed-range
16712 Generate code treating the given register range as fixed registers.
16713 A fixed register is one that the register allocator cannot use. This is
16714 useful when compiling kernel code. A register range is specified as
16715 two registers separated by a dash. Multiple register ranges can be
16716 specified separated by a comma.
16718 @item -mtls-size=@var{tls-size}
16720 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16723 @item -mtune=@var{cpu-type}
16725 Tune the instruction scheduling for a particular CPU, Valid values are
16726 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16727 and @samp{mckinley}.
16733 Generate code for a 32-bit or 64-bit environment.
16734 The 32-bit environment sets int, long and pointer to 32 bits.
16735 The 64-bit environment sets int to 32 bits and long and pointer
16736 to 64 bits. These are HP-UX specific flags.
16738 @item -mno-sched-br-data-spec
16739 @itemx -msched-br-data-spec
16740 @opindex mno-sched-br-data-spec
16741 @opindex msched-br-data-spec
16742 (Dis/En)able data speculative scheduling before reload.
16743 This results in generation of @code{ld.a} instructions and
16744 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16745 The default is 'disable'.
16747 @item -msched-ar-data-spec
16748 @itemx -mno-sched-ar-data-spec
16749 @opindex msched-ar-data-spec
16750 @opindex mno-sched-ar-data-spec
16751 (En/Dis)able data speculative scheduling after reload.
16752 This results in generation of @code{ld.a} instructions and
16753 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16754 The default is 'enable'.
16756 @item -mno-sched-control-spec
16757 @itemx -msched-control-spec
16758 @opindex mno-sched-control-spec
16759 @opindex msched-control-spec
16760 (Dis/En)able control speculative scheduling. This feature is
16761 available only during region scheduling (i.e.@: before reload).
16762 This results in generation of the @code{ld.s} instructions and
16763 the corresponding check instructions @code{chk.s}.
16764 The default is 'disable'.
16766 @item -msched-br-in-data-spec
16767 @itemx -mno-sched-br-in-data-spec
16768 @opindex msched-br-in-data-spec
16769 @opindex mno-sched-br-in-data-spec
16770 (En/Dis)able speculative scheduling of the instructions that
16771 are dependent on the data speculative loads before reload.
16772 This is effective only with @option{-msched-br-data-spec} enabled.
16773 The default is 'enable'.
16775 @item -msched-ar-in-data-spec
16776 @itemx -mno-sched-ar-in-data-spec
16777 @opindex msched-ar-in-data-spec
16778 @opindex mno-sched-ar-in-data-spec
16779 (En/Dis)able speculative scheduling of the instructions that
16780 are dependent on the data speculative loads after reload.
16781 This is effective only with @option{-msched-ar-data-spec} enabled.
16782 The default is 'enable'.
16784 @item -msched-in-control-spec
16785 @itemx -mno-sched-in-control-spec
16786 @opindex msched-in-control-spec
16787 @opindex mno-sched-in-control-spec
16788 (En/Dis)able speculative scheduling of the instructions that
16789 are dependent on the control speculative loads.
16790 This is effective only with @option{-msched-control-spec} enabled.
16791 The default is 'enable'.
16793 @item -mno-sched-prefer-non-data-spec-insns
16794 @itemx -msched-prefer-non-data-spec-insns
16795 @opindex mno-sched-prefer-non-data-spec-insns
16796 @opindex msched-prefer-non-data-spec-insns
16797 If enabled, data-speculative instructions are chosen for schedule
16798 only if there are no other choices at the moment. This makes
16799 the use of the data speculation much more conservative.
16800 The default is 'disable'.
16802 @item -mno-sched-prefer-non-control-spec-insns
16803 @itemx -msched-prefer-non-control-spec-insns
16804 @opindex mno-sched-prefer-non-control-spec-insns
16805 @opindex msched-prefer-non-control-spec-insns
16806 If enabled, control-speculative instructions are chosen for schedule
16807 only if there are no other choices at the moment. This makes
16808 the use of the control speculation much more conservative.
16809 The default is 'disable'.
16811 @item -mno-sched-count-spec-in-critical-path
16812 @itemx -msched-count-spec-in-critical-path
16813 @opindex mno-sched-count-spec-in-critical-path
16814 @opindex msched-count-spec-in-critical-path
16815 If enabled, speculative dependencies are considered during
16816 computation of the instructions priorities. This makes the use of the
16817 speculation a bit more conservative.
16818 The default is 'disable'.
16820 @item -msched-spec-ldc
16821 @opindex msched-spec-ldc
16822 Use a simple data speculation check. This option is on by default.
16824 @item -msched-control-spec-ldc
16825 @opindex msched-spec-ldc
16826 Use a simple check for control speculation. This option is on by default.
16828 @item -msched-stop-bits-after-every-cycle
16829 @opindex msched-stop-bits-after-every-cycle
16830 Place a stop bit after every cycle when scheduling. This option is on
16833 @item -msched-fp-mem-deps-zero-cost
16834 @opindex msched-fp-mem-deps-zero-cost
16835 Assume that floating-point stores and loads are not likely to cause a conflict
16836 when placed into the same instruction group. This option is disabled by
16839 @item -msel-sched-dont-check-control-spec
16840 @opindex msel-sched-dont-check-control-spec
16841 Generate checks for control speculation in selective scheduling.
16842 This flag is disabled by default.
16844 @item -msched-max-memory-insns=@var{max-insns}
16845 @opindex msched-max-memory-insns
16846 Limit on the number of memory insns per instruction group, giving lower
16847 priority to subsequent memory insns attempting to schedule in the same
16848 instruction group. Frequently useful to prevent cache bank conflicts.
16849 The default value is 1.
16851 @item -msched-max-memory-insns-hard-limit
16852 @opindex msched-max-memory-insns-hard-limit
16853 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16854 disallowing more than that number in an instruction group.
16855 Otherwise, the limit is ``soft'', meaning that non-memory operations
16856 are preferred when the limit is reached, but memory operations may still
16862 @subsection LM32 Options
16863 @cindex LM32 options
16865 These @option{-m} options are defined for the LatticeMico32 architecture:
16868 @item -mbarrel-shift-enabled
16869 @opindex mbarrel-shift-enabled
16870 Enable barrel-shift instructions.
16872 @item -mdivide-enabled
16873 @opindex mdivide-enabled
16874 Enable divide and modulus instructions.
16876 @item -mmultiply-enabled
16877 @opindex multiply-enabled
16878 Enable multiply instructions.
16880 @item -msign-extend-enabled
16881 @opindex msign-extend-enabled
16882 Enable sign extend instructions.
16884 @item -muser-enabled
16885 @opindex muser-enabled
16886 Enable user-defined instructions.
16891 @subsection M32C Options
16892 @cindex M32C options
16895 @item -mcpu=@var{name}
16897 Select the CPU for which code is generated. @var{name} may be one of
16898 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16899 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16900 the M32C/80 series.
16904 Specifies that the program will be run on the simulator. This causes
16905 an alternate runtime library to be linked in which supports, for
16906 example, file I/O@. You must not use this option when generating
16907 programs that will run on real hardware; you must provide your own
16908 runtime library for whatever I/O functions are needed.
16910 @item -memregs=@var{number}
16912 Specifies the number of memory-based pseudo-registers GCC uses
16913 during code generation. These pseudo-registers are used like real
16914 registers, so there is a tradeoff between GCC's ability to fit the
16915 code into available registers, and the performance penalty of using
16916 memory instead of registers. Note that all modules in a program must
16917 be compiled with the same value for this option. Because of that, you
16918 must not use this option with GCC's default runtime libraries.
16922 @node M32R/D Options
16923 @subsection M32R/D Options
16924 @cindex M32R/D options
16926 These @option{-m} options are defined for Renesas M32R/D architectures:
16931 Generate code for the M32R/2@.
16935 Generate code for the M32R/X@.
16939 Generate code for the M32R@. This is the default.
16941 @item -mmodel=small
16942 @opindex mmodel=small
16943 Assume all objects live in the lower 16MB of memory (so that their addresses
16944 can be loaded with the @code{ld24} instruction), and assume all subroutines
16945 are reachable with the @code{bl} instruction.
16946 This is the default.
16948 The addressability of a particular object can be set with the
16949 @code{model} attribute.
16951 @item -mmodel=medium
16952 @opindex mmodel=medium
16953 Assume objects may be anywhere in the 32-bit address space (the compiler
16954 generates @code{seth/add3} instructions to load their addresses), and
16955 assume all subroutines are reachable with the @code{bl} instruction.
16957 @item -mmodel=large
16958 @opindex mmodel=large
16959 Assume objects may be anywhere in the 32-bit address space (the compiler
16960 generates @code{seth/add3} instructions to load their addresses), and
16961 assume subroutines may not be reachable with the @code{bl} instruction
16962 (the compiler generates the much slower @code{seth/add3/jl}
16963 instruction sequence).
16966 @opindex msdata=none
16967 Disable use of the small data area. Variables are put into
16968 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16969 @code{section} attribute has been specified).
16970 This is the default.
16972 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16973 Objects may be explicitly put in the small data area with the
16974 @code{section} attribute using one of these sections.
16976 @item -msdata=sdata
16977 @opindex msdata=sdata
16978 Put small global and static data in the small data area, but do not
16979 generate special code to reference them.
16982 @opindex msdata=use
16983 Put small global and static data in the small data area, and generate
16984 special instructions to reference them.
16988 @cindex smaller data references
16989 Put global and static objects less than or equal to @var{num} bytes
16990 into the small data or BSS sections instead of the normal data or BSS
16991 sections. The default value of @var{num} is 8.
16992 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16993 for this option to have any effect.
16995 All modules should be compiled with the same @option{-G @var{num}} value.
16996 Compiling with different values of @var{num} may or may not work; if it
16997 doesn't the linker gives an error message---incorrect code is not
17002 Makes the M32R-specific code in the compiler display some statistics
17003 that might help in debugging programs.
17005 @item -malign-loops
17006 @opindex malign-loops
17007 Align all loops to a 32-byte boundary.
17009 @item -mno-align-loops
17010 @opindex mno-align-loops
17011 Do not enforce a 32-byte alignment for loops. This is the default.
17013 @item -missue-rate=@var{number}
17014 @opindex missue-rate=@var{number}
17015 Issue @var{number} instructions per cycle. @var{number} can only be 1
17018 @item -mbranch-cost=@var{number}
17019 @opindex mbranch-cost=@var{number}
17020 @var{number} can only be 1 or 2. If it is 1 then branches are
17021 preferred over conditional code, if it is 2, then the opposite applies.
17023 @item -mflush-trap=@var{number}
17024 @opindex mflush-trap=@var{number}
17025 Specifies the trap number to use to flush the cache. The default is
17026 12. Valid numbers are between 0 and 15 inclusive.
17028 @item -mno-flush-trap
17029 @opindex mno-flush-trap
17030 Specifies that the cache cannot be flushed by using a trap.
17032 @item -mflush-func=@var{name}
17033 @opindex mflush-func=@var{name}
17034 Specifies the name of the operating system function to call to flush
17035 the cache. The default is @samp{_flush_cache}, but a function call
17036 is only used if a trap is not available.
17038 @item -mno-flush-func
17039 @opindex mno-flush-func
17040 Indicates that there is no OS function for flushing the cache.
17044 @node M680x0 Options
17045 @subsection M680x0 Options
17046 @cindex M680x0 options
17048 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
17049 The default settings depend on which architecture was selected when
17050 the compiler was configured; the defaults for the most common choices
17054 @item -march=@var{arch}
17056 Generate code for a specific M680x0 or ColdFire instruction set
17057 architecture. Permissible values of @var{arch} for M680x0
17058 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
17059 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
17060 architectures are selected according to Freescale's ISA classification
17061 and the permissible values are: @samp{isaa}, @samp{isaaplus},
17062 @samp{isab} and @samp{isac}.
17064 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
17065 code for a ColdFire target. The @var{arch} in this macro is one of the
17066 @option{-march} arguments given above.
17068 When used together, @option{-march} and @option{-mtune} select code
17069 that runs on a family of similar processors but that is optimized
17070 for a particular microarchitecture.
17072 @item -mcpu=@var{cpu}
17074 Generate code for a specific M680x0 or ColdFire processor.
17075 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
17076 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
17077 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
17078 below, which also classifies the CPUs into families:
17080 @multitable @columnfractions 0.20 0.80
17081 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
17082 @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}
17083 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
17084 @item @samp{5206e} @tab @samp{5206e}
17085 @item @samp{5208} @tab @samp{5207} @samp{5208}
17086 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
17087 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
17088 @item @samp{5216} @tab @samp{5214} @samp{5216}
17089 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
17090 @item @samp{5225} @tab @samp{5224} @samp{5225}
17091 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
17092 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
17093 @item @samp{5249} @tab @samp{5249}
17094 @item @samp{5250} @tab @samp{5250}
17095 @item @samp{5271} @tab @samp{5270} @samp{5271}
17096 @item @samp{5272} @tab @samp{5272}
17097 @item @samp{5275} @tab @samp{5274} @samp{5275}
17098 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
17099 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
17100 @item @samp{5307} @tab @samp{5307}
17101 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
17102 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
17103 @item @samp{5407} @tab @samp{5407}
17104 @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}
17107 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
17108 @var{arch} is compatible with @var{cpu}. Other combinations of
17109 @option{-mcpu} and @option{-march} are rejected.
17111 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
17112 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
17113 where the value of @var{family} is given by the table above.
17115 @item -mtune=@var{tune}
17117 Tune the code for a particular microarchitecture within the
17118 constraints set by @option{-march} and @option{-mcpu}.
17119 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
17120 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
17121 and @samp{cpu32}. The ColdFire microarchitectures
17122 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
17124 You can also use @option{-mtune=68020-40} for code that needs
17125 to run relatively well on 68020, 68030 and 68040 targets.
17126 @option{-mtune=68020-60} is similar but includes 68060 targets
17127 as well. These two options select the same tuning decisions as
17128 @option{-m68020-40} and @option{-m68020-60} respectively.
17130 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
17131 when tuning for 680x0 architecture @var{arch}. It also defines
17132 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
17133 option is used. If GCC is tuning for a range of architectures,
17134 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
17135 it defines the macros for every architecture in the range.
17137 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
17138 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
17139 of the arguments given above.
17145 Generate output for a 68000. This is the default
17146 when the compiler is configured for 68000-based systems.
17147 It is equivalent to @option{-march=68000}.
17149 Use this option for microcontrollers with a 68000 or EC000 core,
17150 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
17154 Generate output for a 68010. This is the default
17155 when the compiler is configured for 68010-based systems.
17156 It is equivalent to @option{-march=68010}.
17162 Generate output for a 68020. This is the default
17163 when the compiler is configured for 68020-based systems.
17164 It is equivalent to @option{-march=68020}.
17168 Generate output for a 68030. This is the default when the compiler is
17169 configured for 68030-based systems. It is equivalent to
17170 @option{-march=68030}.
17174 Generate output for a 68040. This is the default when the compiler is
17175 configured for 68040-based systems. It is equivalent to
17176 @option{-march=68040}.
17178 This option inhibits the use of 68881/68882 instructions that have to be
17179 emulated by software on the 68040. Use this option if your 68040 does not
17180 have code to emulate those instructions.
17184 Generate output for a 68060. This is the default when the compiler is
17185 configured for 68060-based systems. It is equivalent to
17186 @option{-march=68060}.
17188 This option inhibits the use of 68020 and 68881/68882 instructions that
17189 have to be emulated by software on the 68060. Use this option if your 68060
17190 does not have code to emulate those instructions.
17194 Generate output for a CPU32. This is the default
17195 when the compiler is configured for CPU32-based systems.
17196 It is equivalent to @option{-march=cpu32}.
17198 Use this option for microcontrollers with a
17199 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
17200 68336, 68340, 68341, 68349 and 68360.
17204 Generate output for a 520X ColdFire CPU@. This is the default
17205 when the compiler is configured for 520X-based systems.
17206 It is equivalent to @option{-mcpu=5206}, and is now deprecated
17207 in favor of that option.
17209 Use this option for microcontroller with a 5200 core, including
17210 the MCF5202, MCF5203, MCF5204 and MCF5206.
17214 Generate output for a 5206e ColdFire CPU@. The option is now
17215 deprecated in favor of the equivalent @option{-mcpu=5206e}.
17219 Generate output for a member of the ColdFire 528X family.
17220 The option is now deprecated in favor of the equivalent
17221 @option{-mcpu=528x}.
17225 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
17226 in favor of the equivalent @option{-mcpu=5307}.
17230 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
17231 in favor of the equivalent @option{-mcpu=5407}.
17235 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
17236 This includes use of hardware floating-point instructions.
17237 The option is equivalent to @option{-mcpu=547x}, and is now
17238 deprecated in favor of that option.
17242 Generate output for a 68040, without using any of the new instructions.
17243 This results in code that can run relatively efficiently on either a
17244 68020/68881 or a 68030 or a 68040. The generated code does use the
17245 68881 instructions that are emulated on the 68040.
17247 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
17251 Generate output for a 68060, without using any of the new instructions.
17252 This results in code that can run relatively efficiently on either a
17253 68020/68881 or a 68030 or a 68040. The generated code does use the
17254 68881 instructions that are emulated on the 68060.
17256 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
17260 @opindex mhard-float
17262 Generate floating-point instructions. This is the default for 68020
17263 and above, and for ColdFire devices that have an FPU@. It defines the
17264 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
17265 on ColdFire targets.
17268 @opindex msoft-float
17269 Do not generate floating-point instructions; use library calls instead.
17270 This is the default for 68000, 68010, and 68832 targets. It is also
17271 the default for ColdFire devices that have no FPU.
17277 Generate (do not generate) ColdFire hardware divide and remainder
17278 instructions. If @option{-march} is used without @option{-mcpu},
17279 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
17280 architectures. Otherwise, the default is taken from the target CPU
17281 (either the default CPU, or the one specified by @option{-mcpu}). For
17282 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
17283 @option{-mcpu=5206e}.
17285 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
17289 Consider type @code{int} to be 16 bits wide, like @code{short int}.
17290 Additionally, parameters passed on the stack are also aligned to a
17291 16-bit boundary even on targets whose API mandates promotion to 32-bit.
17295 Do not consider type @code{int} to be 16 bits wide. This is the default.
17298 @itemx -mno-bitfield
17299 @opindex mnobitfield
17300 @opindex mno-bitfield
17301 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
17302 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
17306 Do use the bit-field instructions. The @option{-m68020} option implies
17307 @option{-mbitfield}. This is the default if you use a configuration
17308 designed for a 68020.
17312 Use a different function-calling convention, in which functions
17313 that take a fixed number of arguments return with the @code{rtd}
17314 instruction, which pops their arguments while returning. This
17315 saves one instruction in the caller since there is no need to pop
17316 the arguments there.
17318 This calling convention is incompatible with the one normally
17319 used on Unix, so you cannot use it if you need to call libraries
17320 compiled with the Unix compiler.
17322 Also, you must provide function prototypes for all functions that
17323 take variable numbers of arguments (including @code{printf});
17324 otherwise incorrect code is generated for calls to those
17327 In addition, seriously incorrect code results if you call a
17328 function with too many arguments. (Normally, extra arguments are
17329 harmlessly ignored.)
17331 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
17332 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
17336 Do not use the calling conventions selected by @option{-mrtd}.
17337 This is the default.
17340 @itemx -mno-align-int
17341 @opindex malign-int
17342 @opindex mno-align-int
17343 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
17344 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
17345 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
17346 Aligning variables on 32-bit boundaries produces code that runs somewhat
17347 faster on processors with 32-bit busses at the expense of more memory.
17349 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
17350 aligns structures containing the above types differently than
17351 most published application binary interface specifications for the m68k.
17355 Use the pc-relative addressing mode of the 68000 directly, instead of
17356 using a global offset table. At present, this option implies @option{-fpic},
17357 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
17358 not presently supported with @option{-mpcrel}, though this could be supported for
17359 68020 and higher processors.
17361 @item -mno-strict-align
17362 @itemx -mstrict-align
17363 @opindex mno-strict-align
17364 @opindex mstrict-align
17365 Do not (do) assume that unaligned memory references are handled by
17369 Generate code that allows the data segment to be located in a different
17370 area of memory from the text segment. This allows for execute-in-place in
17371 an environment without virtual memory management. This option implies
17374 @item -mno-sep-data
17375 Generate code that assumes that the data segment follows the text segment.
17376 This is the default.
17378 @item -mid-shared-library
17379 Generate code that supports shared libraries via the library ID method.
17380 This allows for execute-in-place and shared libraries in an environment
17381 without virtual memory management. This option implies @option{-fPIC}.
17383 @item -mno-id-shared-library
17384 Generate code that doesn't assume ID-based shared libraries are being used.
17385 This is the default.
17387 @item -mshared-library-id=n
17388 Specifies the identification number of the ID-based shared library being
17389 compiled. Specifying a value of 0 generates more compact code; specifying
17390 other values forces the allocation of that number to the current
17391 library, but is no more space- or time-efficient than omitting this option.
17397 When generating position-independent code for ColdFire, generate code
17398 that works if the GOT has more than 8192 entries. This code is
17399 larger and slower than code generated without this option. On M680x0
17400 processors, this option is not needed; @option{-fPIC} suffices.
17402 GCC normally uses a single instruction to load values from the GOT@.
17403 While this is relatively efficient, it only works if the GOT
17404 is smaller than about 64k. Anything larger causes the linker
17405 to report an error such as:
17407 @cindex relocation truncated to fit (ColdFire)
17409 relocation truncated to fit: R_68K_GOT16O foobar
17412 If this happens, you should recompile your code with @option{-mxgot}.
17413 It should then work with very large GOTs. However, code generated with
17414 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
17415 the value of a global symbol.
17417 Note that some linkers, including newer versions of the GNU linker,
17418 can create multiple GOTs and sort GOT entries. If you have such a linker,
17419 you should only need to use @option{-mxgot} when compiling a single
17420 object file that accesses more than 8192 GOT entries. Very few do.
17422 These options have no effect unless GCC is generating
17423 position-independent code.
17427 @node MCore Options
17428 @subsection MCore Options
17429 @cindex MCore options
17431 These are the @samp{-m} options defined for the Motorola M*Core
17437 @itemx -mno-hardlit
17439 @opindex mno-hardlit
17440 Inline constants into the code stream if it can be done in two
17441 instructions or less.
17447 Use the divide instruction. (Enabled by default).
17449 @item -mrelax-immediate
17450 @itemx -mno-relax-immediate
17451 @opindex mrelax-immediate
17452 @opindex mno-relax-immediate
17453 Allow arbitrary-sized immediates in bit operations.
17455 @item -mwide-bitfields
17456 @itemx -mno-wide-bitfields
17457 @opindex mwide-bitfields
17458 @opindex mno-wide-bitfields
17459 Always treat bit-fields as @code{int}-sized.
17461 @item -m4byte-functions
17462 @itemx -mno-4byte-functions
17463 @opindex m4byte-functions
17464 @opindex mno-4byte-functions
17465 Force all functions to be aligned to a 4-byte boundary.
17467 @item -mcallgraph-data
17468 @itemx -mno-callgraph-data
17469 @opindex mcallgraph-data
17470 @opindex mno-callgraph-data
17471 Emit callgraph information.
17474 @itemx -mno-slow-bytes
17475 @opindex mslow-bytes
17476 @opindex mno-slow-bytes
17477 Prefer word access when reading byte quantities.
17479 @item -mlittle-endian
17480 @itemx -mbig-endian
17481 @opindex mlittle-endian
17482 @opindex mbig-endian
17483 Generate code for a little-endian target.
17489 Generate code for the 210 processor.
17493 Assume that runtime support has been provided and so omit the
17494 simulator library (@file{libsim.a)} from the linker command line.
17496 @item -mstack-increment=@var{size}
17497 @opindex mstack-increment
17498 Set the maximum amount for a single stack increment operation. Large
17499 values can increase the speed of programs that contain functions
17500 that need a large amount of stack space, but they can also trigger a
17501 segmentation fault if the stack is extended too much. The default
17507 @subsection MeP Options
17508 @cindex MeP options
17514 Enables the @code{abs} instruction, which is the absolute difference
17515 between two registers.
17519 Enables all the optional instructions---average, multiply, divide, bit
17520 operations, leading zero, absolute difference, min/max, clip, and
17526 Enables the @code{ave} instruction, which computes the average of two
17529 @item -mbased=@var{n}
17531 Variables of size @var{n} bytes or smaller are placed in the
17532 @code{.based} section by default. Based variables use the @code{$tp}
17533 register as a base register, and there is a 128-byte limit to the
17534 @code{.based} section.
17538 Enables the bit operation instructions---bit test (@code{btstm}), set
17539 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17540 test-and-set (@code{tas}).
17542 @item -mc=@var{name}
17544 Selects which section constant data is placed in. @var{name} may
17545 be @samp{tiny}, @samp{near}, or @samp{far}.
17549 Enables the @code{clip} instruction. Note that @option{-mclip} is not
17550 useful unless you also provide @option{-mminmax}.
17552 @item -mconfig=@var{name}
17554 Selects one of the built-in core configurations. Each MeP chip has
17555 one or more modules in it; each module has a core CPU and a variety of
17556 coprocessors, optional instructions, and peripherals. The
17557 @code{MeP-Integrator} tool, not part of GCC, provides these
17558 configurations through this option; using this option is the same as
17559 using all the corresponding command-line options. The default
17560 configuration is @samp{default}.
17564 Enables the coprocessor instructions. By default, this is a 32-bit
17565 coprocessor. Note that the coprocessor is normally enabled via the
17566 @option{-mconfig=} option.
17570 Enables the 32-bit coprocessor's instructions.
17574 Enables the 64-bit coprocessor's instructions.
17578 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17582 Causes constant variables to be placed in the @code{.near} section.
17586 Enables the @code{div} and @code{divu} instructions.
17590 Generate big-endian code.
17594 Generate little-endian code.
17596 @item -mio-volatile
17597 @opindex mio-volatile
17598 Tells the compiler that any variable marked with the @code{io}
17599 attribute is to be considered volatile.
17603 Causes variables to be assigned to the @code{.far} section by default.
17607 Enables the @code{leadz} (leading zero) instruction.
17611 Causes variables to be assigned to the @code{.near} section by default.
17615 Enables the @code{min} and @code{max} instructions.
17619 Enables the multiplication and multiply-accumulate instructions.
17623 Disables all the optional instructions enabled by @option{-mall-opts}.
17627 Enables the @code{repeat} and @code{erepeat} instructions, used for
17628 low-overhead looping.
17632 Causes all variables to default to the @code{.tiny} section. Note
17633 that there is a 65536-byte limit to this section. Accesses to these
17634 variables use the @code{%gp} base register.
17638 Enables the saturation instructions. Note that the compiler does not
17639 currently generate these itself, but this option is included for
17640 compatibility with other tools, like @code{as}.
17644 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17648 Link the simulator run-time libraries.
17652 Link the simulator runtime libraries, excluding built-in support
17653 for reset and exception vectors and tables.
17657 Causes all functions to default to the @code{.far} section. Without
17658 this option, functions default to the @code{.near} section.
17660 @item -mtiny=@var{n}
17662 Variables that are @var{n} bytes or smaller are allocated to the
17663 @code{.tiny} section. These variables use the @code{$gp} base
17664 register. The default for this option is 4, but note that there's a
17665 65536-byte limit to the @code{.tiny} section.
17669 @node MicroBlaze Options
17670 @subsection MicroBlaze Options
17671 @cindex MicroBlaze Options
17676 @opindex msoft-float
17677 Use software emulation for floating point (default).
17680 @opindex mhard-float
17681 Use hardware floating-point instructions.
17685 Do not optimize block moves, use @code{memcpy}.
17687 @item -mno-clearbss
17688 @opindex mno-clearbss
17689 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17691 @item -mcpu=@var{cpu-type}
17693 Use features of, and schedule code for, the given CPU.
17694 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17695 where @var{X} is a major version, @var{YY} is the minor version, and
17696 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17697 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17699 @item -mxl-soft-mul
17700 @opindex mxl-soft-mul
17701 Use software multiply emulation (default).
17703 @item -mxl-soft-div
17704 @opindex mxl-soft-div
17705 Use software emulation for divides (default).
17707 @item -mxl-barrel-shift
17708 @opindex mxl-barrel-shift
17709 Use the hardware barrel shifter.
17711 @item -mxl-pattern-compare
17712 @opindex mxl-pattern-compare
17713 Use pattern compare instructions.
17715 @item -msmall-divides
17716 @opindex msmall-divides
17717 Use table lookup optimization for small signed integer divisions.
17719 @item -mxl-stack-check
17720 @opindex mxl-stack-check
17721 This option is deprecated. Use @option{-fstack-check} instead.
17724 @opindex mxl-gp-opt
17725 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17727 @item -mxl-multiply-high
17728 @opindex mxl-multiply-high
17729 Use multiply high instructions for high part of 32x32 multiply.
17731 @item -mxl-float-convert
17732 @opindex mxl-float-convert
17733 Use hardware floating-point conversion instructions.
17735 @item -mxl-float-sqrt
17736 @opindex mxl-float-sqrt
17737 Use hardware floating-point square root instruction.
17740 @opindex mbig-endian
17741 Generate code for a big-endian target.
17743 @item -mlittle-endian
17744 @opindex mlittle-endian
17745 Generate code for a little-endian target.
17748 @opindex mxl-reorder
17749 Use reorder instructions (swap and byte reversed load/store).
17751 @item -mxl-mode-@var{app-model}
17752 Select application model @var{app-model}. Valid models are
17755 normal executable (default), uses startup code @file{crt0.o}.
17758 for use with Xilinx Microprocessor Debugger (XMD) based
17759 software intrusive debug agent called xmdstub. This uses startup file
17760 @file{crt1.o} and sets the start address of the program to 0x800.
17763 for applications that are loaded using a bootloader.
17764 This model uses startup file @file{crt2.o} which does not contain a processor
17765 reset vector handler. This is suitable for transferring control on a
17766 processor reset to the bootloader rather than the application.
17769 for applications that do not require any of the
17770 MicroBlaze vectors. This option may be useful for applications running
17771 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17774 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17775 @option{-mxl-mode-@var{app-model}}.
17780 @subsection MIPS Options
17781 @cindex MIPS options
17787 Generate big-endian code.
17791 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17794 @item -march=@var{arch}
17796 Generate code that runs on @var{arch}, which can be the name of a
17797 generic MIPS ISA, or the name of a particular processor.
17799 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17800 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17801 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17802 @samp{mips64r5} and @samp{mips64r6}.
17803 The processor names are:
17804 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17805 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17806 @samp{5kc}, @samp{5kf},
17808 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17809 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17810 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17811 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17812 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17813 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17815 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17816 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17819 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17820 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17821 @samp{rm7000}, @samp{rm9000},
17822 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17825 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17826 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17827 @samp{xlr} and @samp{xlp}.
17828 The special value @samp{from-abi} selects the
17829 most compatible architecture for the selected ABI (that is,
17830 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17832 The native Linux/GNU toolchain also supports the value @samp{native},
17833 which selects the best architecture option for the host processor.
17834 @option{-march=native} has no effect if GCC does not recognize
17837 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17838 (for example, @option{-march=r2k}). Prefixes are optional, and
17839 @samp{vr} may be written @samp{r}.
17841 Names of the form @samp{@var{n}f2_1} refer to processors with
17842 FPUs clocked at half the rate of the core, names of the form
17843 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17844 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17845 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17846 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17847 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17848 accepted as synonyms for @samp{@var{n}f1_1}.
17850 GCC defines two macros based on the value of this option. The first
17851 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17852 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17853 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17854 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17855 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17857 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17858 above. In other words, it has the full prefix and does not
17859 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17860 the macro names the resolved architecture (either @code{"mips1"} or
17861 @code{"mips3"}). It names the default architecture when no
17862 @option{-march} option is given.
17864 @item -mtune=@var{arch}
17866 Optimize for @var{arch}. Among other things, this option controls
17867 the way instructions are scheduled, and the perceived cost of arithmetic
17868 operations. The list of @var{arch} values is the same as for
17871 When this option is not used, GCC optimizes for the processor
17872 specified by @option{-march}. By using @option{-march} and
17873 @option{-mtune} together, it is possible to generate code that
17874 runs on a family of processors, but optimize the code for one
17875 particular member of that family.
17877 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17878 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17879 @option{-march} ones described above.
17883 Equivalent to @option{-march=mips1}.
17887 Equivalent to @option{-march=mips2}.
17891 Equivalent to @option{-march=mips3}.
17895 Equivalent to @option{-march=mips4}.
17899 Equivalent to @option{-march=mips32}.
17903 Equivalent to @option{-march=mips32r3}.
17907 Equivalent to @option{-march=mips32r5}.
17911 Equivalent to @option{-march=mips32r6}.
17915 Equivalent to @option{-march=mips64}.
17919 Equivalent to @option{-march=mips64r2}.
17923 Equivalent to @option{-march=mips64r3}.
17927 Equivalent to @option{-march=mips64r5}.
17931 Equivalent to @option{-march=mips64r6}.
17936 @opindex mno-mips16
17937 Generate (do not generate) MIPS16 code. If GCC is targeting a
17938 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17940 MIPS16 code generation can also be controlled on a per-function basis
17941 by means of @code{mips16} and @code{nomips16} attributes.
17942 @xref{Function Attributes}, for more information.
17944 @item -mflip-mips16
17945 @opindex mflip-mips16
17946 Generate MIPS16 code on alternating functions. This option is provided
17947 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17948 not intended for ordinary use in compiling user code.
17950 @item -minterlink-compressed
17951 @item -mno-interlink-compressed
17952 @opindex minterlink-compressed
17953 @opindex mno-interlink-compressed
17954 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17955 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17957 For example, code using the standard ISA encoding cannot jump directly
17958 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17959 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17960 knows that the target of the jump is not compressed.
17962 @item -minterlink-mips16
17963 @itemx -mno-interlink-mips16
17964 @opindex minterlink-mips16
17965 @opindex mno-interlink-mips16
17966 Aliases of @option{-minterlink-compressed} and
17967 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17968 and are retained for backwards compatibility.
17980 Generate code for the given ABI@.
17982 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17983 generates 64-bit code when you select a 64-bit architecture, but you
17984 can use @option{-mgp32} to get 32-bit code instead.
17986 For information about the O64 ABI, see
17987 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17989 GCC supports a variant of the o32 ABI in which floating-point registers
17990 are 64 rather than 32 bits wide. You can select this combination with
17991 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17992 and @code{mfhc1} instructions and is therefore only supported for
17993 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17995 The register assignments for arguments and return values remain the
17996 same, but each scalar value is passed in a single 64-bit register
17997 rather than a pair of 32-bit registers. For example, scalar
17998 floating-point values are returned in @samp{$f0} only, not a
17999 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
18000 remains the same in that the even-numbered double-precision registers
18003 Two additional variants of the o32 ABI are supported to enable
18004 a transition from 32-bit to 64-bit registers. These are FPXX
18005 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
18006 The FPXX extension mandates that all code must execute correctly
18007 when run using 32-bit or 64-bit registers. The code can be interlinked
18008 with either FP32 or FP64, but not both.
18009 The FP64A extension is similar to the FP64 extension but forbids the
18010 use of odd-numbered single-precision registers. This can be used
18011 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
18012 processors and allows both FP32 and FP64A code to interlink and
18013 run in the same process without changing FPU modes.
18016 @itemx -mno-abicalls
18018 @opindex mno-abicalls
18019 Generate (do not generate) code that is suitable for SVR4-style
18020 dynamic objects. @option{-mabicalls} is the default for SVR4-based
18025 Generate (do not generate) code that is fully position-independent,
18026 and that can therefore be linked into shared libraries. This option
18027 only affects @option{-mabicalls}.
18029 All @option{-mabicalls} code has traditionally been position-independent,
18030 regardless of options like @option{-fPIC} and @option{-fpic}. However,
18031 as an extension, the GNU toolchain allows executables to use absolute
18032 accesses for locally-binding symbols. It can also use shorter GP
18033 initialization sequences and generate direct calls to locally-defined
18034 functions. This mode is selected by @option{-mno-shared}.
18036 @option{-mno-shared} depends on binutils 2.16 or higher and generates
18037 objects that can only be linked by the GNU linker. However, the option
18038 does not affect the ABI of the final executable; it only affects the ABI
18039 of relocatable objects. Using @option{-mno-shared} generally makes
18040 executables both smaller and quicker.
18042 @option{-mshared} is the default.
18048 Assume (do not assume) that the static and dynamic linkers
18049 support PLTs and copy relocations. This option only affects
18050 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
18051 has no effect without @option{-msym32}.
18053 You can make @option{-mplt} the default by configuring
18054 GCC with @option{--with-mips-plt}. The default is
18055 @option{-mno-plt} otherwise.
18061 Lift (do not lift) the usual restrictions on the size of the global
18064 GCC normally uses a single instruction to load values from the GOT@.
18065 While this is relatively efficient, it only works if the GOT
18066 is smaller than about 64k. Anything larger causes the linker
18067 to report an error such as:
18069 @cindex relocation truncated to fit (MIPS)
18071 relocation truncated to fit: R_MIPS_GOT16 foobar
18074 If this happens, you should recompile your code with @option{-mxgot}.
18075 This works with very large GOTs, although the code is also
18076 less efficient, since it takes three instructions to fetch the
18077 value of a global symbol.
18079 Note that some linkers can create multiple GOTs. If you have such a
18080 linker, you should only need to use @option{-mxgot} when a single object
18081 file accesses more than 64k's worth of GOT entries. Very few do.
18083 These options have no effect unless GCC is generating position
18088 Assume that general-purpose registers are 32 bits wide.
18092 Assume that general-purpose registers are 64 bits wide.
18096 Assume that floating-point registers are 32 bits wide.
18100 Assume that floating-point registers are 64 bits wide.
18104 Do not assume the width of floating-point registers.
18107 @opindex mhard-float
18108 Use floating-point coprocessor instructions.
18111 @opindex msoft-float
18112 Do not use floating-point coprocessor instructions. Implement
18113 floating-point calculations using library calls instead.
18117 Equivalent to @option{-msoft-float}, but additionally asserts that the
18118 program being compiled does not perform any floating-point operations.
18119 This option is presently supported only by some bare-metal MIPS
18120 configurations, where it may select a special set of libraries
18121 that lack all floating-point support (including, for example, the
18122 floating-point @code{printf} formats).
18123 If code compiled with @option{-mno-float} accidentally contains
18124 floating-point operations, it is likely to suffer a link-time
18125 or run-time failure.
18127 @item -msingle-float
18128 @opindex msingle-float
18129 Assume that the floating-point coprocessor only supports single-precision
18132 @item -mdouble-float
18133 @opindex mdouble-float
18134 Assume that the floating-point coprocessor supports double-precision
18135 operations. This is the default.
18138 @itemx -mno-odd-spreg
18139 @opindex modd-spreg
18140 @opindex mno-odd-spreg
18141 Enable the use of odd-numbered single-precision floating-point registers
18142 for the o32 ABI. This is the default for processors that are known to
18143 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
18147 @itemx -mabs=legacy
18149 @opindex mabs=legacy
18150 These options control the treatment of the special not-a-number (NaN)
18151 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
18152 @code{neg.@i{fmt}} machine instructions.
18154 By default or when the @option{-mabs=legacy} is used the legacy
18155 treatment is selected. In this case these instructions are considered
18156 arithmetic and avoided where correct operation is required and the
18157 input operand might be a NaN. A longer sequence of instructions that
18158 manipulate the sign bit of floating-point datum manually is used
18159 instead unless the @option{-ffinite-math-only} option has also been
18162 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
18163 this case these instructions are considered non-arithmetic and therefore
18164 operating correctly in all cases, including in particular where the
18165 input operand is a NaN. These instructions are therefore always used
18166 for the respective operations.
18169 @itemx -mnan=legacy
18171 @opindex mnan=legacy
18172 These options control the encoding of the special not-a-number (NaN)
18173 IEEE 754 floating-point data.
18175 The @option{-mnan=legacy} option selects the legacy encoding. In this
18176 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
18177 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
18178 by the first bit of their trailing significand field being 1.
18180 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
18181 this case qNaNs are denoted by the first bit of their trailing
18182 significand field being 1, whereas sNaNs are denoted by the first bit of
18183 their trailing significand field being 0.
18185 The default is @option{-mnan=legacy} unless GCC has been configured with
18186 @option{--with-nan=2008}.
18192 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
18193 implement atomic memory built-in functions. When neither option is
18194 specified, GCC uses the instructions if the target architecture
18197 @option{-mllsc} is useful if the runtime environment can emulate the
18198 instructions and @option{-mno-llsc} can be useful when compiling for
18199 nonstandard ISAs. You can make either option the default by
18200 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
18201 respectively. @option{--with-llsc} is the default for some
18202 configurations; see the installation documentation for details.
18208 Use (do not use) revision 1 of the MIPS DSP ASE@.
18209 @xref{MIPS DSP Built-in Functions}. This option defines the
18210 preprocessor macro @code{__mips_dsp}. It also defines
18211 @code{__mips_dsp_rev} to 1.
18217 Use (do not use) revision 2 of the MIPS DSP ASE@.
18218 @xref{MIPS DSP Built-in Functions}. This option defines the
18219 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
18220 It also defines @code{__mips_dsp_rev} to 2.
18223 @itemx -mno-smartmips
18224 @opindex msmartmips
18225 @opindex mno-smartmips
18226 Use (do not use) the MIPS SmartMIPS ASE.
18228 @item -mpaired-single
18229 @itemx -mno-paired-single
18230 @opindex mpaired-single
18231 @opindex mno-paired-single
18232 Use (do not use) paired-single floating-point instructions.
18233 @xref{MIPS Paired-Single Support}. This option requires
18234 hardware floating-point support to be enabled.
18240 Use (do not use) MIPS Digital Media Extension instructions.
18241 This option can only be used when generating 64-bit code and requires
18242 hardware floating-point support to be enabled.
18247 @opindex mno-mips3d
18248 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
18249 The option @option{-mips3d} implies @option{-mpaired-single}.
18252 @itemx -mno-micromips
18253 @opindex mmicromips
18254 @opindex mno-mmicromips
18255 Generate (do not generate) microMIPS code.
18257 MicroMIPS code generation can also be controlled on a per-function basis
18258 by means of @code{micromips} and @code{nomicromips} attributes.
18259 @xref{Function Attributes}, for more information.
18265 Use (do not use) MT Multithreading instructions.
18271 Use (do not use) the MIPS MCU ASE instructions.
18277 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
18283 Use (do not use) the MIPS Virtualization Application Specific instructions.
18289 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
18293 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
18294 an explanation of the default and the way that the pointer size is
18299 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
18301 The default size of @code{int}s, @code{long}s and pointers depends on
18302 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
18303 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
18304 32-bit @code{long}s. Pointers are the same size as @code{long}s,
18305 or the same size as integer registers, whichever is smaller.
18311 Assume (do not assume) that all symbols have 32-bit values, regardless
18312 of the selected ABI@. This option is useful in combination with
18313 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
18314 to generate shorter and faster references to symbolic addresses.
18318 Put definitions of externally-visible data in a small data section
18319 if that data is no bigger than @var{num} bytes. GCC can then generate
18320 more efficient accesses to the data; see @option{-mgpopt} for details.
18322 The default @option{-G} option depends on the configuration.
18324 @item -mlocal-sdata
18325 @itemx -mno-local-sdata
18326 @opindex mlocal-sdata
18327 @opindex mno-local-sdata
18328 Extend (do not extend) the @option{-G} behavior to local data too,
18329 such as to static variables in C@. @option{-mlocal-sdata} is the
18330 default for all configurations.
18332 If the linker complains that an application is using too much small data,
18333 you might want to try rebuilding the less performance-critical parts with
18334 @option{-mno-local-sdata}. You might also want to build large
18335 libraries with @option{-mno-local-sdata}, so that the libraries leave
18336 more room for the main program.
18338 @item -mextern-sdata
18339 @itemx -mno-extern-sdata
18340 @opindex mextern-sdata
18341 @opindex mno-extern-sdata
18342 Assume (do not assume) that externally-defined data is in
18343 a small data section if the size of that data is within the @option{-G} limit.
18344 @option{-mextern-sdata} is the default for all configurations.
18346 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
18347 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
18348 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
18349 is placed in a small data section. If @var{Var} is defined by another
18350 module, you must either compile that module with a high-enough
18351 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
18352 definition. If @var{Var} is common, you must link the application
18353 with a high-enough @option{-G} setting.
18355 The easiest way of satisfying these restrictions is to compile
18356 and link every module with the same @option{-G} option. However,
18357 you may wish to build a library that supports several different
18358 small data limits. You can do this by compiling the library with
18359 the highest supported @option{-G} setting and additionally using
18360 @option{-mno-extern-sdata} to stop the library from making assumptions
18361 about externally-defined data.
18367 Use (do not use) GP-relative accesses for symbols that are known to be
18368 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
18369 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
18372 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
18373 might not hold the value of @code{_gp}. For example, if the code is
18374 part of a library that might be used in a boot monitor, programs that
18375 call boot monitor routines pass an unknown value in @code{$gp}.
18376 (In such situations, the boot monitor itself is usually compiled
18377 with @option{-G0}.)
18379 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
18380 @option{-mno-extern-sdata}.
18382 @item -membedded-data
18383 @itemx -mno-embedded-data
18384 @opindex membedded-data
18385 @opindex mno-embedded-data
18386 Allocate variables to the read-only data section first if possible, then
18387 next in the small data section if possible, otherwise in data. This gives
18388 slightly slower code than the default, but reduces the amount of RAM required
18389 when executing, and thus may be preferred for some embedded systems.
18391 @item -muninit-const-in-rodata
18392 @itemx -mno-uninit-const-in-rodata
18393 @opindex muninit-const-in-rodata
18394 @opindex mno-uninit-const-in-rodata
18395 Put uninitialized @code{const} variables in the read-only data section.
18396 This option is only meaningful in conjunction with @option{-membedded-data}.
18398 @item -mcode-readable=@var{setting}
18399 @opindex mcode-readable
18400 Specify whether GCC may generate code that reads from executable sections.
18401 There are three possible settings:
18404 @item -mcode-readable=yes
18405 Instructions may freely access executable sections. This is the
18408 @item -mcode-readable=pcrel
18409 MIPS16 PC-relative load instructions can access executable sections,
18410 but other instructions must not do so. This option is useful on 4KSc
18411 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
18412 It is also useful on processors that can be configured to have a dual
18413 instruction/data SRAM interface and that, like the M4K, automatically
18414 redirect PC-relative loads to the instruction RAM.
18416 @item -mcode-readable=no
18417 Instructions must not access executable sections. This option can be
18418 useful on targets that are configured to have a dual instruction/data
18419 SRAM interface but that (unlike the M4K) do not automatically redirect
18420 PC-relative loads to the instruction RAM.
18423 @item -msplit-addresses
18424 @itemx -mno-split-addresses
18425 @opindex msplit-addresses
18426 @opindex mno-split-addresses
18427 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
18428 relocation operators. This option has been superseded by
18429 @option{-mexplicit-relocs} but is retained for backwards compatibility.
18431 @item -mexplicit-relocs
18432 @itemx -mno-explicit-relocs
18433 @opindex mexplicit-relocs
18434 @opindex mno-explicit-relocs
18435 Use (do not use) assembler relocation operators when dealing with symbolic
18436 addresses. The alternative, selected by @option{-mno-explicit-relocs},
18437 is to use assembler macros instead.
18439 @option{-mexplicit-relocs} is the default if GCC was configured
18440 to use an assembler that supports relocation operators.
18442 @item -mcheck-zero-division
18443 @itemx -mno-check-zero-division
18444 @opindex mcheck-zero-division
18445 @opindex mno-check-zero-division
18446 Trap (do not trap) on integer division by zero.
18448 The default is @option{-mcheck-zero-division}.
18450 @item -mdivide-traps
18451 @itemx -mdivide-breaks
18452 @opindex mdivide-traps
18453 @opindex mdivide-breaks
18454 MIPS systems check for division by zero by generating either a
18455 conditional trap or a break instruction. Using traps results in
18456 smaller code, but is only supported on MIPS II and later. Also, some
18457 versions of the Linux kernel have a bug that prevents trap from
18458 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
18459 allow conditional traps on architectures that support them and
18460 @option{-mdivide-breaks} to force the use of breaks.
18462 The default is usually @option{-mdivide-traps}, but this can be
18463 overridden at configure time using @option{--with-divide=breaks}.
18464 Divide-by-zero checks can be completely disabled using
18465 @option{-mno-check-zero-division}.
18470 @opindex mno-memcpy
18471 Force (do not force) the use of @code{memcpy} for non-trivial block
18472 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
18473 most constant-sized copies.
18476 @itemx -mno-long-calls
18477 @opindex mlong-calls
18478 @opindex mno-long-calls
18479 Disable (do not disable) use of the @code{jal} instruction. Calling
18480 functions using @code{jal} is more efficient but requires the caller
18481 and callee to be in the same 256 megabyte segment.
18483 This option has no effect on abicalls code. The default is
18484 @option{-mno-long-calls}.
18490 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
18491 instructions, as provided by the R4650 ISA@.
18497 Enable (disable) use of the @code{madd} and @code{msub} integer
18498 instructions. The default is @option{-mimadd} on architectures
18499 that support @code{madd} and @code{msub} except for the 74k
18500 architecture where it was found to generate slower code.
18503 @itemx -mno-fused-madd
18504 @opindex mfused-madd
18505 @opindex mno-fused-madd
18506 Enable (disable) use of the floating-point multiply-accumulate
18507 instructions, when they are available. The default is
18508 @option{-mfused-madd}.
18510 On the R8000 CPU when multiply-accumulate instructions are used,
18511 the intermediate product is calculated to infinite precision
18512 and is not subject to the FCSR Flush to Zero bit. This may be
18513 undesirable in some circumstances. On other processors the result
18514 is numerically identical to the equivalent computation using
18515 separate multiply, add, subtract and negate instructions.
18519 Tell the MIPS assembler to not run its preprocessor over user
18520 assembler files (with a @samp{.s} suffix) when assembling them.
18525 @opindex mno-fix-24k
18526 Work around the 24K E48 (lost data on stores during refill) errata.
18527 The workarounds are implemented by the assembler rather than by GCC@.
18530 @itemx -mno-fix-r4000
18531 @opindex mfix-r4000
18532 @opindex mno-fix-r4000
18533 Work around certain R4000 CPU errata:
18536 A double-word or a variable shift may give an incorrect result if executed
18537 immediately after starting an integer division.
18539 A double-word or a variable shift may give an incorrect result if executed
18540 while an integer multiplication is in progress.
18542 An integer division may give an incorrect result if started in a delay slot
18543 of a taken branch or a jump.
18547 @itemx -mno-fix-r4400
18548 @opindex mfix-r4400
18549 @opindex mno-fix-r4400
18550 Work around certain R4400 CPU errata:
18553 A double-word or a variable shift may give an incorrect result if executed
18554 immediately after starting an integer division.
18558 @itemx -mno-fix-r10000
18559 @opindex mfix-r10000
18560 @opindex mno-fix-r10000
18561 Work around certain R10000 errata:
18564 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18565 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18568 This option can only be used if the target architecture supports
18569 branch-likely instructions. @option{-mfix-r10000} is the default when
18570 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18574 @itemx -mno-fix-rm7000
18575 @opindex mfix-rm7000
18576 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18577 workarounds are implemented by the assembler rather than by GCC@.
18580 @itemx -mno-fix-vr4120
18581 @opindex mfix-vr4120
18582 Work around certain VR4120 errata:
18585 @code{dmultu} does not always produce the correct result.
18587 @code{div} and @code{ddiv} do not always produce the correct result if one
18588 of the operands is negative.
18590 The workarounds for the division errata rely on special functions in
18591 @file{libgcc.a}. At present, these functions are only provided by
18592 the @code{mips64vr*-elf} configurations.
18594 Other VR4120 errata require a NOP to be inserted between certain pairs of
18595 instructions. These errata are handled by the assembler, not by GCC itself.
18598 @opindex mfix-vr4130
18599 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18600 workarounds are implemented by the assembler rather than by GCC,
18601 although GCC avoids using @code{mflo} and @code{mfhi} if the
18602 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18603 instructions are available instead.
18606 @itemx -mno-fix-sb1
18608 Work around certain SB-1 CPU core errata.
18609 (This flag currently works around the SB-1 revision 2
18610 ``F1'' and ``F2'' floating-point errata.)
18612 @item -mr10k-cache-barrier=@var{setting}
18613 @opindex mr10k-cache-barrier
18614 Specify whether GCC should insert cache barriers to avoid the
18615 side-effects of speculation on R10K processors.
18617 In common with many processors, the R10K tries to predict the outcome
18618 of a conditional branch and speculatively executes instructions from
18619 the ``taken'' branch. It later aborts these instructions if the
18620 predicted outcome is wrong. However, on the R10K, even aborted
18621 instructions can have side effects.
18623 This problem only affects kernel stores and, depending on the system,
18624 kernel loads. As an example, a speculatively-executed store may load
18625 the target memory into cache and mark the cache line as dirty, even if
18626 the store itself is later aborted. If a DMA operation writes to the
18627 same area of memory before the ``dirty'' line is flushed, the cached
18628 data overwrites the DMA-ed data. See the R10K processor manual
18629 for a full description, including other potential problems.
18631 One workaround is to insert cache barrier instructions before every memory
18632 access that might be speculatively executed and that might have side
18633 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18634 controls GCC's implementation of this workaround. It assumes that
18635 aborted accesses to any byte in the following regions does not have
18640 the memory occupied by the current function's stack frame;
18643 the memory occupied by an incoming stack argument;
18646 the memory occupied by an object with a link-time-constant address.
18649 It is the kernel's responsibility to ensure that speculative
18650 accesses to these regions are indeed safe.
18652 If the input program contains a function declaration such as:
18658 then the implementation of @code{foo} must allow @code{j foo} and
18659 @code{jal foo} to be executed speculatively. GCC honors this
18660 restriction for functions it compiles itself. It expects non-GCC
18661 functions (such as hand-written assembly code) to do the same.
18663 The option has three forms:
18666 @item -mr10k-cache-barrier=load-store
18667 Insert a cache barrier before a load or store that might be
18668 speculatively executed and that might have side effects even
18671 @item -mr10k-cache-barrier=store
18672 Insert a cache barrier before a store that might be speculatively
18673 executed and that might have side effects even if aborted.
18675 @item -mr10k-cache-barrier=none
18676 Disable the insertion of cache barriers. This is the default setting.
18679 @item -mflush-func=@var{func}
18680 @itemx -mno-flush-func
18681 @opindex mflush-func
18682 Specifies the function to call to flush the I and D caches, or to not
18683 call any such function. If called, the function must take the same
18684 arguments as the common @code{_flush_func}, that is, the address of the
18685 memory range for which the cache is being flushed, the size of the
18686 memory range, and the number 3 (to flush both caches). The default
18687 depends on the target GCC was configured for, but commonly is either
18688 @code{_flush_func} or @code{__cpu_flush}.
18690 @item mbranch-cost=@var{num}
18691 @opindex mbranch-cost
18692 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18693 This cost is only a heuristic and is not guaranteed to produce
18694 consistent results across releases. A zero cost redundantly selects
18695 the default, which is based on the @option{-mtune} setting.
18697 @item -mbranch-likely
18698 @itemx -mno-branch-likely
18699 @opindex mbranch-likely
18700 @opindex mno-branch-likely
18701 Enable or disable use of Branch Likely instructions, regardless of the
18702 default for the selected architecture. By default, Branch Likely
18703 instructions may be generated if they are supported by the selected
18704 architecture. An exception is for the MIPS32 and MIPS64 architectures
18705 and processors that implement those architectures; for those, Branch
18706 Likely instructions are not be generated by default because the MIPS32
18707 and MIPS64 architectures specifically deprecate their use.
18709 @item -mfp-exceptions
18710 @itemx -mno-fp-exceptions
18711 @opindex mfp-exceptions
18712 Specifies whether FP exceptions are enabled. This affects how
18713 FP instructions are scheduled for some processors.
18714 The default is that FP exceptions are
18717 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18718 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18721 @item -mvr4130-align
18722 @itemx -mno-vr4130-align
18723 @opindex mvr4130-align
18724 The VR4130 pipeline is two-way superscalar, but can only issue two
18725 instructions together if the first one is 8-byte aligned. When this
18726 option is enabled, GCC aligns pairs of instructions that it
18727 thinks should execute in parallel.
18729 This option only has an effect when optimizing for the VR4130.
18730 It normally makes code faster, but at the expense of making it bigger.
18731 It is enabled by default at optimization level @option{-O3}.
18736 Enable (disable) generation of @code{synci} instructions on
18737 architectures that support it. The @code{synci} instructions (if
18738 enabled) are generated when @code{__builtin___clear_cache} is
18741 This option defaults to @option{-mno-synci}, but the default can be
18742 overridden by configuring GCC with @option{--with-synci}.
18744 When compiling code for single processor systems, it is generally safe
18745 to use @code{synci}. However, on many multi-core (SMP) systems, it
18746 does not invalidate the instruction caches on all cores and may lead
18747 to undefined behavior.
18749 @item -mrelax-pic-calls
18750 @itemx -mno-relax-pic-calls
18751 @opindex mrelax-pic-calls
18752 Try to turn PIC calls that are normally dispatched via register
18753 @code{$25} into direct calls. This is only possible if the linker can
18754 resolve the destination at link-time and if the destination is within
18755 range for a direct call.
18757 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18758 an assembler and a linker that support the @code{.reloc} assembly
18759 directive and @option{-mexplicit-relocs} is in effect. With
18760 @option{-mno-explicit-relocs}, this optimization can be performed by the
18761 assembler and the linker alone without help from the compiler.
18763 @item -mmcount-ra-address
18764 @itemx -mno-mcount-ra-address
18765 @opindex mmcount-ra-address
18766 @opindex mno-mcount-ra-address
18767 Emit (do not emit) code that allows @code{_mcount} to modify the
18768 calling function's return address. When enabled, this option extends
18769 the usual @code{_mcount} interface with a new @var{ra-address}
18770 parameter, which has type @code{intptr_t *} and is passed in register
18771 @code{$12}. @code{_mcount} can then modify the return address by
18772 doing both of the following:
18775 Returning the new address in register @code{$31}.
18777 Storing the new address in @code{*@var{ra-address}},
18778 if @var{ra-address} is nonnull.
18781 The default is @option{-mno-mcount-ra-address}.
18786 @subsection MMIX Options
18787 @cindex MMIX Options
18789 These options are defined for the MMIX:
18793 @itemx -mno-libfuncs
18795 @opindex mno-libfuncs
18796 Specify that intrinsic library functions are being compiled, passing all
18797 values in registers, no matter the size.
18800 @itemx -mno-epsilon
18802 @opindex mno-epsilon
18803 Generate floating-point comparison instructions that compare with respect
18804 to the @code{rE} epsilon register.
18806 @item -mabi=mmixware
18808 @opindex mabi=mmixware
18810 Generate code that passes function parameters and return values that (in
18811 the called function) are seen as registers @code{$0} and up, as opposed to
18812 the GNU ABI which uses global registers @code{$231} and up.
18814 @item -mzero-extend
18815 @itemx -mno-zero-extend
18816 @opindex mzero-extend
18817 @opindex mno-zero-extend
18818 When reading data from memory in sizes shorter than 64 bits, use (do not
18819 use) zero-extending load instructions by default, rather than
18820 sign-extending ones.
18823 @itemx -mno-knuthdiv
18825 @opindex mno-knuthdiv
18826 Make the result of a division yielding a remainder have the same sign as
18827 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18828 remainder follows the sign of the dividend. Both methods are
18829 arithmetically valid, the latter being almost exclusively used.
18831 @item -mtoplevel-symbols
18832 @itemx -mno-toplevel-symbols
18833 @opindex mtoplevel-symbols
18834 @opindex mno-toplevel-symbols
18835 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18836 code can be used with the @code{PREFIX} assembly directive.
18840 Generate an executable in the ELF format, rather than the default
18841 @samp{mmo} format used by the @command{mmix} simulator.
18843 @item -mbranch-predict
18844 @itemx -mno-branch-predict
18845 @opindex mbranch-predict
18846 @opindex mno-branch-predict
18847 Use (do not use) the probable-branch instructions, when static branch
18848 prediction indicates a probable branch.
18850 @item -mbase-addresses
18851 @itemx -mno-base-addresses
18852 @opindex mbase-addresses
18853 @opindex mno-base-addresses
18854 Generate (do not generate) code that uses @emph{base addresses}. Using a
18855 base address automatically generates a request (handled by the assembler
18856 and the linker) for a constant to be set up in a global register. The
18857 register is used for one or more base address requests within the range 0
18858 to 255 from the value held in the register. The generally leads to short
18859 and fast code, but the number of different data items that can be
18860 addressed is limited. This means that a program that uses lots of static
18861 data may require @option{-mno-base-addresses}.
18863 @item -msingle-exit
18864 @itemx -mno-single-exit
18865 @opindex msingle-exit
18866 @opindex mno-single-exit
18867 Force (do not force) generated code to have a single exit point in each
18871 @node MN10300 Options
18872 @subsection MN10300 Options
18873 @cindex MN10300 options
18875 These @option{-m} options are defined for Matsushita MN10300 architectures:
18880 Generate code to avoid bugs in the multiply instructions for the MN10300
18881 processors. This is the default.
18883 @item -mno-mult-bug
18884 @opindex mno-mult-bug
18885 Do not generate code to avoid bugs in the multiply instructions for the
18886 MN10300 processors.
18890 Generate code using features specific to the AM33 processor.
18894 Do not generate code using features specific to the AM33 processor. This
18899 Generate code using features specific to the AM33/2.0 processor.
18903 Generate code using features specific to the AM34 processor.
18905 @item -mtune=@var{cpu-type}
18907 Use the timing characteristics of the indicated CPU type when
18908 scheduling instructions. This does not change the targeted processor
18909 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18910 @samp{am33-2} or @samp{am34}.
18912 @item -mreturn-pointer-on-d0
18913 @opindex mreturn-pointer-on-d0
18914 When generating a function that returns a pointer, return the pointer
18915 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18916 only in @code{a0}, and attempts to call such functions without a prototype
18917 result in errors. Note that this option is on by default; use
18918 @option{-mno-return-pointer-on-d0} to disable it.
18922 Do not link in the C run-time initialization object file.
18926 Indicate to the linker that it should perform a relaxation optimization pass
18927 to shorten branches, calls and absolute memory addresses. This option only
18928 has an effect when used on the command line for the final link step.
18930 This option makes symbolic debugging impossible.
18934 Allow the compiler to generate @emph{Long Instruction Word}
18935 instructions if the target is the @samp{AM33} or later. This is the
18936 default. This option defines the preprocessor macro @code{__LIW__}.
18940 Do not allow the compiler to generate @emph{Long Instruction Word}
18941 instructions. This option defines the preprocessor macro
18946 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18947 instructions if the target is the @samp{AM33} or later. This is the
18948 default. This option defines the preprocessor macro @code{__SETLB__}.
18952 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18953 instructions. This option defines the preprocessor macro
18954 @code{__NO_SETLB__}.
18958 @node Moxie Options
18959 @subsection Moxie Options
18960 @cindex Moxie Options
18966 Generate big-endian code. This is the default for @samp{moxie-*-*}
18971 Generate little-endian code.
18975 Generate mul.x and umul.x instructions. This is the default for
18976 @samp{moxiebox-*-*} configurations.
18980 Do not link in the C run-time initialization object file.
18984 @node MSP430 Options
18985 @subsection MSP430 Options
18986 @cindex MSP430 Options
18988 These options are defined for the MSP430:
18994 Force assembly output to always use hex constants. Normally such
18995 constants are signed decimals, but this option is available for
18996 testsuite and/or aesthetic purposes.
19000 Select the MCU to target. This is used to create a C preprocessor
19001 symbol based upon the MCU name, converted to upper case and pre- and
19002 post-fixed with @samp{__}. This in turn is used by the
19003 @file{msp430.h} header file to select an MCU-specific supplementary
19006 The option also sets the ISA to use. If the MCU name is one that is
19007 known to only support the 430 ISA then that is selected, otherwise the
19008 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
19009 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
19010 name selects the 430X ISA.
19012 In addition an MCU-specific linker script is added to the linker
19013 command line. The script's name is the name of the MCU with
19014 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
19015 command line defines the C preprocessor symbol @code{__XXX__} and
19016 cause the linker to search for a script called @file{xxx.ld}.
19018 This option is also passed on to the assembler.
19022 Specifies the ISA to use. Accepted values are @samp{msp430},
19023 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
19024 @option{-mmcu=} option should be used to select the ISA.
19028 Link to the simulator runtime libraries and linker script. Overrides
19029 any scripts that would be selected by the @option{-mmcu=} option.
19033 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
19037 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
19041 This option is passed to the assembler and linker, and allows the
19042 linker to perform certain optimizations that cannot be done until
19047 Describes the type of hardware multiply supported by the target.
19048 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
19049 for the original 16-bit-only multiply supported by early MCUs.
19050 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
19051 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
19052 A value of @samp{auto} can also be given. This tells GCC to deduce
19053 the hardware multiply support based upon the MCU name provided by the
19054 @option{-mmcu} option. If no @option{-mmcu} option is specified then
19055 @samp{32bit} hardware multiply support is assumed. @samp{auto} is the
19058 Hardware multiplies are normally performed by calling a library
19059 routine. This saves space in the generated code. When compiling at
19060 @option{-O3} or higher however the hardware multiplier is invoked
19061 inline. This makes for bigger, but faster code.
19063 The hardware multiply routines disable interrupts whilst running and
19064 restore the previous interrupt state when they finish. This makes
19065 them safe to use inside interrupt handlers as well as in normal code.
19069 Enable the use of a minimum runtime environment - no static
19070 initializers or constructors. This is intended for memory-constrained
19071 devices. The compiler includes special symbols in some objects
19072 that tell the linker and runtime which code fragments are required.
19076 @node NDS32 Options
19077 @subsection NDS32 Options
19078 @cindex NDS32 Options
19080 These options are defined for NDS32 implementations:
19085 @opindex mbig-endian
19086 Generate code in big-endian mode.
19088 @item -mlittle-endian
19089 @opindex mlittle-endian
19090 Generate code in little-endian mode.
19092 @item -mreduced-regs
19093 @opindex mreduced-regs
19094 Use reduced-set registers for register allocation.
19097 @opindex mfull-regs
19098 Use full-set registers for register allocation.
19102 Generate conditional move instructions.
19106 Do not generate conditional move instructions.
19110 Generate performance extension instructions.
19112 @item -mno-perf-ext
19113 @opindex mno-perf-ext
19114 Do not generate performance extension instructions.
19118 Generate v3 push25/pop25 instructions.
19121 @opindex mno-v3push
19122 Do not generate v3 push25/pop25 instructions.
19126 Generate 16-bit instructions.
19129 @opindex mno-16-bit
19130 Do not generate 16-bit instructions.
19133 @opindex mgp-direct
19134 Generate GP base instructions directly.
19136 @item -mno-gp-direct
19137 @opindex mno-gp-direct
19138 Do no generate GP base instructions directly.
19140 @item -misr-vector-size=@var{num}
19141 @opindex misr-vector-size
19142 Specify the size of each interrupt vector, which must be 4 or 16.
19144 @item -mcache-block-size=@var{num}
19145 @opindex mcache-block-size
19146 Specify the size of each cache block,
19147 which must be a power of 2 between 4 and 512.
19149 @item -march=@var{arch}
19151 Specify the name of the target architecture.
19153 @item -mforce-fp-as-gp
19154 @opindex mforce-fp-as-gp
19155 Prevent $fp being allocated during register allocation so that compiler
19156 is able to force performing fp-as-gp optimization.
19158 @item -mforbid-fp-as-gp
19159 @opindex mforbid-fp-as-gp
19160 Forbid using $fp to access static and global variables.
19161 This option strictly forbids fp-as-gp optimization
19162 regardless of @option{-mforce-fp-as-gp}.
19166 Use special directives to guide linker doing ex9 optimization.
19169 @opindex mctor-dtor
19170 Enable constructor/destructor feature.
19174 Guide linker to relax instructions.
19178 @node Nios II Options
19179 @subsection Nios II Options
19180 @cindex Nios II options
19181 @cindex Altera Nios II options
19183 These are the options defined for the Altera Nios II processor.
19189 @cindex smaller data references
19190 Put global and static objects less than or equal to @var{num} bytes
19191 into the small data or BSS sections instead of the normal data or BSS
19192 sections. The default value of @var{num} is 8.
19198 Generate (do not generate) GP-relative accesses for objects in the
19199 small data or BSS sections. The default is @option{-mgpopt} except
19200 when @option{-fpic} or @option{-fPIC} is specified to generate
19201 position-independent code. Note that the Nios II ABI does not permit
19202 GP-relative accesses from shared libraries.
19204 You may need to specify @option{-mno-gpopt} explicitly when building
19205 programs that include large amounts of small data, including large
19206 GOT data sections. In this case, the 16-bit offset for GP-relative
19207 addressing may not be large enough to allow access to the entire
19208 small data section.
19214 Generate little-endian (default) or big-endian (experimental) code,
19217 @item -mbypass-cache
19218 @itemx -mno-bypass-cache
19219 @opindex mno-bypass-cache
19220 @opindex mbypass-cache
19221 Force all load and store instructions to always bypass cache by
19222 using I/O variants of the instructions. The default is not to
19225 @item -mno-cache-volatile
19226 @itemx -mcache-volatile
19227 @opindex mcache-volatile
19228 @opindex mno-cache-volatile
19229 Volatile memory access bypass the cache using the I/O variants of
19230 the load and store instructions. The default is not to bypass the cache.
19232 @item -mno-fast-sw-div
19233 @itemx -mfast-sw-div
19234 @opindex mno-fast-sw-div
19235 @opindex mfast-sw-div
19236 Do not use table-based fast divide for small numbers. The default
19237 is to use the fast divide at @option{-O3} and above.
19241 @itemx -mno-hw-mulx
19245 @opindex mno-hw-mul
19247 @opindex mno-hw-mulx
19249 @opindex mno-hw-div
19251 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
19252 instructions by the compiler. The default is to emit @code{mul}
19253 and not emit @code{div} and @code{mulx}.
19255 @item -mcustom-@var{insn}=@var{N}
19256 @itemx -mno-custom-@var{insn}
19257 @opindex mcustom-@var{insn}
19258 @opindex mno-custom-@var{insn}
19259 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
19260 custom instruction with encoding @var{N} when generating code that uses
19261 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
19262 instruction 253 for single-precision floating-point add operations instead
19263 of the default behavior of using a library call.
19265 The following values of @var{insn} are supported. Except as otherwise
19266 noted, floating-point operations are expected to be implemented with
19267 normal IEEE 754 semantics and correspond directly to the C operators or the
19268 equivalent GCC built-in functions (@pxref{Other Builtins}).
19270 Single-precision floating point:
19273 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
19274 Binary arithmetic operations.
19280 Unary absolute value.
19282 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
19283 Comparison operations.
19285 @item @samp{fmins}, @samp{fmaxs}
19286 Floating-point minimum and maximum. These instructions are only
19287 generated if @option{-ffinite-math-only} is specified.
19289 @item @samp{fsqrts}
19290 Unary square root operation.
19292 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
19293 Floating-point trigonometric and exponential functions. These instructions
19294 are only generated if @option{-funsafe-math-optimizations} is also specified.
19298 Double-precision floating point:
19301 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
19302 Binary arithmetic operations.
19308 Unary absolute value.
19310 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
19311 Comparison operations.
19313 @item @samp{fmind}, @samp{fmaxd}
19314 Double-precision minimum and maximum. These instructions are only
19315 generated if @option{-ffinite-math-only} is specified.
19317 @item @samp{fsqrtd}
19318 Unary square root operation.
19320 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
19321 Double-precision trigonometric and exponential functions. These instructions
19322 are only generated if @option{-funsafe-math-optimizations} is also specified.
19328 @item @samp{fextsd}
19329 Conversion from single precision to double precision.
19331 @item @samp{ftruncds}
19332 Conversion from double precision to single precision.
19334 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
19335 Conversion from floating point to signed or unsigned integer types, with
19336 truncation towards zero.
19339 Conversion from single-precision floating point to signed integer,
19340 rounding to the nearest integer and ties away from zero.
19341 This corresponds to the @code{__builtin_lroundf} function when
19342 @option{-fno-math-errno} is used.
19344 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
19345 Conversion from signed or unsigned integer types to floating-point types.
19349 In addition, all of the following transfer instructions for internal
19350 registers X and Y must be provided to use any of the double-precision
19351 floating-point instructions. Custom instructions taking two
19352 double-precision source operands expect the first operand in the
19353 64-bit register X. The other operand (or only operand of a unary
19354 operation) is given to the custom arithmetic instruction with the
19355 least significant half in source register @var{src1} and the most
19356 significant half in @var{src2}. A custom instruction that returns a
19357 double-precision result returns the most significant 32 bits in the
19358 destination register and the other half in 32-bit register Y.
19359 GCC automatically generates the necessary code sequences to write
19360 register X and/or read register Y when double-precision floating-point
19361 instructions are used.
19366 Write @var{src1} into the least significant half of X and @var{src2} into
19367 the most significant half of X.
19370 Write @var{src1} into Y.
19372 @item @samp{frdxhi}, @samp{frdxlo}
19373 Read the most or least (respectively) significant half of X and store it in
19377 Read the value of Y and store it into @var{dest}.
19380 Note that you can gain more local control over generation of Nios II custom
19381 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
19382 and @code{target("no-custom-@var{insn}")} function attributes
19383 (@pxref{Function Attributes})
19384 or pragmas (@pxref{Function Specific Option Pragmas}).
19386 @item -mcustom-fpu-cfg=@var{name}
19387 @opindex mcustom-fpu-cfg
19389 This option enables a predefined, named set of custom instruction encodings
19390 (see @option{-mcustom-@var{insn}} above).
19391 Currently, the following sets are defined:
19393 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
19394 @gccoptlist{-mcustom-fmuls=252 @gol
19395 -mcustom-fadds=253 @gol
19396 -mcustom-fsubs=254 @gol
19397 -fsingle-precision-constant}
19399 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
19400 @gccoptlist{-mcustom-fmuls=252 @gol
19401 -mcustom-fadds=253 @gol
19402 -mcustom-fsubs=254 @gol
19403 -mcustom-fdivs=255 @gol
19404 -fsingle-precision-constant}
19406 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19407 @gccoptlist{-mcustom-floatus=243 @gol
19408 -mcustom-fixsi=244 @gol
19409 -mcustom-floatis=245 @gol
19410 -mcustom-fcmpgts=246 @gol
19411 -mcustom-fcmples=249 @gol
19412 -mcustom-fcmpeqs=250 @gol
19413 -mcustom-fcmpnes=251 @gol
19414 -mcustom-fmuls=252 @gol
19415 -mcustom-fadds=253 @gol
19416 -mcustom-fsubs=254 @gol
19417 -mcustom-fdivs=255 @gol
19418 -fsingle-precision-constant}
19420 Custom instruction assignments given by individual
19421 @option{-mcustom-@var{insn}=} options override those given by
19422 @option{-mcustom-fpu-cfg=}, regardless of the
19423 order of the options on the command line.
19425 Note that you can gain more local control over selection of a FPU
19426 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19427 function attribute (@pxref{Function Attributes})
19428 or pragma (@pxref{Function Specific Option Pragmas}).
19432 These additional @samp{-m} options are available for the Altera Nios II
19433 ELF (bare-metal) target:
19439 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19440 startup and termination code, and is typically used in conjunction with
19441 @option{-msys-crt0=} to specify the location of the alternate startup code
19442 provided by the HAL BSP.
19446 Link with a limited version of the C library, @option{-lsmallc}, rather than
19449 @item -msys-crt0=@var{startfile}
19451 @var{startfile} is the file name of the startfile (crt0) to use
19452 when linking. This option is only useful in conjunction with @option{-mhal}.
19454 @item -msys-lib=@var{systemlib}
19456 @var{systemlib} is the library name of the library that provides
19457 low-level system calls required by the C library,
19458 e.g. @code{read} and @code{write}.
19459 This option is typically used to link with a library provided by a HAL BSP.
19463 @node PDP-11 Options
19464 @subsection PDP-11 Options
19465 @cindex PDP-11 Options
19467 These options are defined for the PDP-11:
19472 Use hardware FPP floating point. This is the default. (FIS floating
19473 point on the PDP-11/40 is not supported.)
19476 @opindex msoft-float
19477 Do not use hardware floating point.
19481 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19485 Return floating-point results in memory. This is the default.
19489 Generate code for a PDP-11/40.
19493 Generate code for a PDP-11/45. This is the default.
19497 Generate code for a PDP-11/10.
19499 @item -mbcopy-builtin
19500 @opindex mbcopy-builtin
19501 Use inline @code{movmemhi} patterns for copying memory. This is the
19506 Do not use inline @code{movmemhi} patterns for copying memory.
19512 Use 16-bit @code{int}. This is the default.
19518 Use 32-bit @code{int}.
19521 @itemx -mno-float32
19523 @opindex mno-float32
19524 Use 64-bit @code{float}. This is the default.
19527 @itemx -mno-float64
19529 @opindex mno-float64
19530 Use 32-bit @code{float}.
19534 Use @code{abshi2} pattern. This is the default.
19538 Do not use @code{abshi2} pattern.
19540 @item -mbranch-expensive
19541 @opindex mbranch-expensive
19542 Pretend that branches are expensive. This is for experimenting with
19543 code generation only.
19545 @item -mbranch-cheap
19546 @opindex mbranch-cheap
19547 Do not pretend that branches are expensive. This is the default.
19551 Use Unix assembler syntax. This is the default when configured for
19552 @samp{pdp11-*-bsd}.
19556 Use DEC assembler syntax. This is the default when configured for any
19557 PDP-11 target other than @samp{pdp11-*-bsd}.
19560 @node picoChip Options
19561 @subsection picoChip Options
19562 @cindex picoChip options
19564 These @samp{-m} options are defined for picoChip implementations:
19568 @item -mae=@var{ae_type}
19570 Set the instruction set, register set, and instruction scheduling
19571 parameters for array element type @var{ae_type}. Supported values
19572 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19574 @option{-mae=ANY} selects a completely generic AE type. Code
19575 generated with this option runs on any of the other AE types. The
19576 code is not as efficient as it would be if compiled for a specific
19577 AE type, and some types of operation (e.g., multiplication) do not
19578 work properly on all types of AE.
19580 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19581 for compiled code, and is the default.
19583 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19584 option may suffer from poor performance of byte (char) manipulation,
19585 since the DSP AE does not provide hardware support for byte load/stores.
19587 @item -msymbol-as-address
19588 Enable the compiler to directly use a symbol name as an address in a
19589 load/store instruction, without first loading it into a
19590 register. Typically, the use of this option generates larger
19591 programs, which run faster than when the option isn't used. However, the
19592 results vary from program to program, so it is left as a user option,
19593 rather than being permanently enabled.
19595 @item -mno-inefficient-warnings
19596 Disables warnings about the generation of inefficient code. These
19597 warnings can be generated, for example, when compiling code that
19598 performs byte-level memory operations on the MAC AE type. The MAC AE has
19599 no hardware support for byte-level memory operations, so all byte
19600 load/stores must be synthesized from word load/store operations. This is
19601 inefficient and a warning is generated to indicate
19602 that you should rewrite the code to avoid byte operations, or to target
19603 an AE type that has the necessary hardware support. This option disables
19608 @node PowerPC Options
19609 @subsection PowerPC Options
19610 @cindex PowerPC options
19612 These are listed under @xref{RS/6000 and PowerPC Options}.
19615 @subsection RL78 Options
19616 @cindex RL78 Options
19622 Links in additional target libraries to support operation within a
19629 Specifies the type of hardware multiplication support to be used. The
19630 default is @samp{none}, which uses software multiplication functions.
19631 The @samp{g13} option is for the hardware multiply/divide peripheral
19632 only on the RL78/G13 targets. The @samp{rl78} option is for the
19633 standard hardware multiplication defined in the RL78 software manual.
19635 @item -m64bit-doubles
19636 @itemx -m32bit-doubles
19637 @opindex m64bit-doubles
19638 @opindex m32bit-doubles
19639 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19640 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19641 @option{-m32bit-doubles}.
19645 @node RS/6000 and PowerPC Options
19646 @subsection IBM RS/6000 and PowerPC Options
19647 @cindex RS/6000 and PowerPC Options
19648 @cindex IBM RS/6000 and PowerPC Options
19650 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19652 @item -mpowerpc-gpopt
19653 @itemx -mno-powerpc-gpopt
19654 @itemx -mpowerpc-gfxopt
19655 @itemx -mno-powerpc-gfxopt
19658 @itemx -mno-powerpc64
19662 @itemx -mno-popcntb
19664 @itemx -mno-popcntd
19673 @itemx -mno-hard-dfp
19674 @opindex mpowerpc-gpopt
19675 @opindex mno-powerpc-gpopt
19676 @opindex mpowerpc-gfxopt
19677 @opindex mno-powerpc-gfxopt
19678 @opindex mpowerpc64
19679 @opindex mno-powerpc64
19683 @opindex mno-popcntb
19685 @opindex mno-popcntd
19691 @opindex mno-mfpgpr
19693 @opindex mno-hard-dfp
19694 You use these options to specify which instructions are available on the
19695 processor you are using. The default value of these options is
19696 determined when configuring GCC@. Specifying the
19697 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19698 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19699 rather than the options listed above.
19701 Specifying @option{-mpowerpc-gpopt} allows
19702 GCC to use the optional PowerPC architecture instructions in the
19703 General Purpose group, including floating-point square root. Specifying
19704 @option{-mpowerpc-gfxopt} allows GCC to
19705 use the optional PowerPC architecture instructions in the Graphics
19706 group, including floating-point select.
19708 The @option{-mmfcrf} option allows GCC to generate the move from
19709 condition register field instruction implemented on the POWER4
19710 processor and other processors that support the PowerPC V2.01
19712 The @option{-mpopcntb} option allows GCC to generate the popcount and
19713 double-precision FP reciprocal estimate instruction implemented on the
19714 POWER5 processor and other processors that support the PowerPC V2.02
19716 The @option{-mpopcntd} option allows GCC to generate the popcount
19717 instruction implemented on the POWER7 processor and other processors
19718 that support the PowerPC V2.06 architecture.
19719 The @option{-mfprnd} option allows GCC to generate the FP round to
19720 integer instructions implemented on the POWER5+ processor and other
19721 processors that support the PowerPC V2.03 architecture.
19722 The @option{-mcmpb} option allows GCC to generate the compare bytes
19723 instruction implemented on the POWER6 processor and other processors
19724 that support the PowerPC V2.05 architecture.
19725 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19726 general-purpose register instructions implemented on the POWER6X
19727 processor and other processors that support the extended PowerPC V2.05
19729 The @option{-mhard-dfp} option allows GCC to generate the decimal
19730 floating-point instructions implemented on some POWER processors.
19732 The @option{-mpowerpc64} option allows GCC to generate the additional
19733 64-bit instructions that are found in the full PowerPC64 architecture
19734 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19735 @option{-mno-powerpc64}.
19737 @item -mcpu=@var{cpu_type}
19739 Set architecture type, register usage, and
19740 instruction scheduling parameters for machine type @var{cpu_type}.
19741 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19742 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19743 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19744 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19745 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19746 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19747 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19748 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19749 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19750 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
19751 @samp{powerpc64}, and @samp{rs64}.
19753 @option{-mcpu=powerpc}, and @option{-mcpu=powerpc64} specify pure 32-bit
19754 PowerPC and 64-bit PowerPC architecture machine
19755 types, with an appropriate, generic processor model assumed for
19756 scheduling purposes.
19758 The other options specify a specific processor. Code generated under
19759 those options runs best on that processor, and may not run at all on
19762 The @option{-mcpu} options automatically enable or disable the
19765 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19766 -mpopcntb -mpopcntd -mpowerpc64 @gol
19767 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19768 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19769 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19770 -mquad-memory -mquad-memory-atomic}
19772 The particular options set for any particular CPU varies between
19773 compiler versions, depending on what setting seems to produce optimal
19774 code for that CPU; it doesn't necessarily reflect the actual hardware's
19775 capabilities. If you wish to set an individual option to a particular
19776 value, you may specify it after the @option{-mcpu} option, like
19777 @option{-mcpu=970 -mno-altivec}.
19779 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19780 not enabled or disabled by the @option{-mcpu} option at present because
19781 AIX does not have full support for these options. You may still
19782 enable or disable them individually if you're sure it'll work in your
19785 @item -mtune=@var{cpu_type}
19787 Set the instruction scheduling parameters for machine type
19788 @var{cpu_type}, but do not set the architecture type or register usage,
19789 as @option{-mcpu=@var{cpu_type}} does. The same
19790 values for @var{cpu_type} are used for @option{-mtune} as for
19791 @option{-mcpu}. If both are specified, the code generated uses the
19792 architecture and registers set by @option{-mcpu}, but the
19793 scheduling parameters set by @option{-mtune}.
19795 @item -mcmodel=small
19796 @opindex mcmodel=small
19797 Generate PowerPC64 code for the small model: The TOC is limited to
19800 @item -mcmodel=medium
19801 @opindex mcmodel=medium
19802 Generate PowerPC64 code for the medium model: The TOC and other static
19803 data may be up to a total of 4G in size.
19805 @item -mcmodel=large
19806 @opindex mcmodel=large
19807 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19808 in size. Other data and code is only limited by the 64-bit address
19812 @itemx -mno-altivec
19814 @opindex mno-altivec
19815 Generate code that uses (does not use) AltiVec instructions, and also
19816 enable the use of built-in functions that allow more direct access to
19817 the AltiVec instruction set. You may also need to set
19818 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19821 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19822 @option{-maltivec=be}, the element order for Altivec intrinsics such
19823 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19824 match array element order corresponding to the endianness of the
19825 target. That is, element zero identifies the leftmost element in a
19826 vector register when targeting a big-endian platform, and identifies
19827 the rightmost element in a vector register when targeting a
19828 little-endian platform.
19831 @opindex maltivec=be
19832 Generate Altivec instructions using big-endian element order,
19833 regardless of whether the target is big- or little-endian. This is
19834 the default when targeting a big-endian platform.
19836 The element order is used to interpret element numbers in Altivec
19837 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19838 @code{vec_insert}. By default, these match array element order
19839 corresponding to the endianness for the target.
19842 @opindex maltivec=le
19843 Generate Altivec instructions using little-endian element order,
19844 regardless of whether the target is big- or little-endian. This is
19845 the default when targeting a little-endian platform. This option is
19846 currently ignored when targeting a big-endian platform.
19848 The element order is used to interpret element numbers in Altivec
19849 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19850 @code{vec_insert}. By default, these match array element order
19851 corresponding to the endianness for the target.
19856 @opindex mno-vrsave
19857 Generate VRSAVE instructions when generating AltiVec code.
19859 @item -mgen-cell-microcode
19860 @opindex mgen-cell-microcode
19861 Generate Cell microcode instructions.
19863 @item -mwarn-cell-microcode
19864 @opindex mwarn-cell-microcode
19865 Warn when a Cell microcode instruction is emitted. An example
19866 of a Cell microcode instruction is a variable shift.
19869 @opindex msecure-plt
19870 Generate code that allows @command{ld} and @command{ld.so}
19871 to build executables and shared
19872 libraries with non-executable @code{.plt} and @code{.got} sections.
19874 32-bit SYSV ABI option.
19878 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19880 requires @code{.plt} and @code{.got}
19881 sections that are both writable and executable.
19882 This is a PowerPC 32-bit SYSV ABI option.
19888 This switch enables or disables the generation of ISEL instructions.
19890 @item -misel=@var{yes/no}
19891 This switch has been deprecated. Use @option{-misel} and
19892 @option{-mno-isel} instead.
19898 This switch enables or disables the generation of SPE simd
19904 @opindex mno-paired
19905 This switch enables or disables the generation of PAIRED simd
19908 @item -mspe=@var{yes/no}
19909 This option has been deprecated. Use @option{-mspe} and
19910 @option{-mno-spe} instead.
19916 Generate code that uses (does not use) vector/scalar (VSX)
19917 instructions, and also enable the use of built-in functions that allow
19918 more direct access to the VSX instruction set.
19923 @opindex mno-crypto
19924 Enable the use (disable) of the built-in functions that allow direct
19925 access to the cryptographic instructions that were added in version
19926 2.07 of the PowerPC ISA.
19928 @item -mdirect-move
19929 @itemx -mno-direct-move
19930 @opindex mdirect-move
19931 @opindex mno-direct-move
19932 Generate code that uses (does not use) the instructions to move data
19933 between the general purpose registers and the vector/scalar (VSX)
19934 registers that were added in version 2.07 of the PowerPC ISA.
19936 @item -mpower8-fusion
19937 @itemx -mno-power8-fusion
19938 @opindex mpower8-fusion
19939 @opindex mno-power8-fusion
19940 Generate code that keeps (does not keeps) some integer operations
19941 adjacent so that the instructions can be fused together on power8 and
19944 @item -mpower8-vector
19945 @itemx -mno-power8-vector
19946 @opindex mpower8-vector
19947 @opindex mno-power8-vector
19948 Generate code that uses (does not use) the vector and scalar
19949 instructions that were added in version 2.07 of the PowerPC ISA. Also
19950 enable the use of built-in functions that allow more direct access to
19951 the vector instructions.
19953 @item -mquad-memory
19954 @itemx -mno-quad-memory
19955 @opindex mquad-memory
19956 @opindex mno-quad-memory
19957 Generate code that uses (does not use) the non-atomic quad word memory
19958 instructions. The @option{-mquad-memory} option requires use of
19961 @item -mquad-memory-atomic
19962 @itemx -mno-quad-memory-atomic
19963 @opindex mquad-memory-atomic
19964 @opindex mno-quad-memory-atomic
19965 Generate code that uses (does not use) the atomic quad word memory
19966 instructions. The @option{-mquad-memory-atomic} option requires use of
19969 @item -mupper-regs-df
19970 @itemx -mno-upper-regs-df
19971 @opindex mupper-regs-df
19972 @opindex mno-upper-regs-df
19973 Generate code that uses (does not use) the scalar double precision
19974 instructions that target all 64 registers in the vector/scalar
19975 floating point register set that were added in version 2.06 of the
19976 PowerPC ISA. The @option{-mupper-regs-df} turned on by default if you
19977 use either of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19978 @option{-mvsx} options.
19980 @item -mupper-regs-sf
19981 @itemx -mno-upper-regs-sf
19982 @opindex mupper-regs-sf
19983 @opindex mno-upper-regs-sf
19984 Generate code that uses (does not use) the scalar single precision
19985 instructions that target all 64 registers in the vector/scalar
19986 floating point register set that were added in version 2.07 of the
19987 PowerPC ISA. The @option{-mupper-regs-sf} turned on by default if you
19988 use either of the @option{-mcpu=power8}, or @option{-mpower8-vector}
19992 @itemx -mno-upper-regs
19993 @opindex mupper-regs
19994 @opindex mno-upper-regs
19995 Generate code that uses (does not use) the scalar
19996 instructions that target all 64 registers in the vector/scalar
19997 floating point register set, depending on the model of the machine.
19999 If the @option{-mno-upper-regs} option is used, it turns off both
20000 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
20002 @item -mfloat-gprs=@var{yes/single/double/no}
20003 @itemx -mfloat-gprs
20004 @opindex mfloat-gprs
20005 This switch enables or disables the generation of floating-point
20006 operations on the general-purpose registers for architectures that
20009 The argument @samp{yes} or @samp{single} enables the use of
20010 single-precision floating-point operations.
20012 The argument @samp{double} enables the use of single and
20013 double-precision floating-point operations.
20015 The argument @samp{no} disables floating-point operations on the
20016 general-purpose registers.
20018 This option is currently only available on the MPC854x.
20024 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
20025 targets (including GNU/Linux). The 32-bit environment sets int, long
20026 and pointer to 32 bits and generates code that runs on any PowerPC
20027 variant. The 64-bit environment sets int to 32 bits and long and
20028 pointer to 64 bits, and generates code for PowerPC64, as for
20029 @option{-mpowerpc64}.
20032 @itemx -mno-fp-in-toc
20033 @itemx -mno-sum-in-toc
20034 @itemx -mminimal-toc
20036 @opindex mno-fp-in-toc
20037 @opindex mno-sum-in-toc
20038 @opindex mminimal-toc
20039 Modify generation of the TOC (Table Of Contents), which is created for
20040 every executable file. The @option{-mfull-toc} option is selected by
20041 default. In that case, GCC allocates at least one TOC entry for
20042 each unique non-automatic variable reference in your program. GCC
20043 also places floating-point constants in the TOC@. However, only
20044 16,384 entries are available in the TOC@.
20046 If you receive a linker error message that saying you have overflowed
20047 the available TOC space, you can reduce the amount of TOC space used
20048 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
20049 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
20050 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
20051 generate code to calculate the sum of an address and a constant at
20052 run time instead of putting that sum into the TOC@. You may specify one
20053 or both of these options. Each causes GCC to produce very slightly
20054 slower and larger code at the expense of conserving TOC space.
20056 If you still run out of space in the TOC even when you specify both of
20057 these options, specify @option{-mminimal-toc} instead. This option causes
20058 GCC to make only one TOC entry for every file. When you specify this
20059 option, GCC produces code that is slower and larger but which
20060 uses extremely little TOC space. You may wish to use this option
20061 only on files that contain less frequently-executed code.
20067 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
20068 @code{long} type, and the infrastructure needed to support them.
20069 Specifying @option{-maix64} implies @option{-mpowerpc64},
20070 while @option{-maix32} disables the 64-bit ABI and
20071 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
20074 @itemx -mno-xl-compat
20075 @opindex mxl-compat
20076 @opindex mno-xl-compat
20077 Produce code that conforms more closely to IBM XL compiler semantics
20078 when using AIX-compatible ABI@. Pass floating-point arguments to
20079 prototyped functions beyond the register save area (RSA) on the stack
20080 in addition to argument FPRs. Do not assume that most significant
20081 double in 128-bit long double value is properly rounded when comparing
20082 values and converting to double. Use XL symbol names for long double
20085 The AIX calling convention was extended but not initially documented to
20086 handle an obscure K&R C case of calling a function that takes the
20087 address of its arguments with fewer arguments than declared. IBM XL
20088 compilers access floating-point arguments that do not fit in the
20089 RSA from the stack when a subroutine is compiled without
20090 optimization. Because always storing floating-point arguments on the
20091 stack is inefficient and rarely needed, this option is not enabled by
20092 default and only is necessary when calling subroutines compiled by IBM
20093 XL compilers without optimization.
20097 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
20098 application written to use message passing with special startup code to
20099 enable the application to run. The system must have PE installed in the
20100 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
20101 must be overridden with the @option{-specs=} option to specify the
20102 appropriate directory location. The Parallel Environment does not
20103 support threads, so the @option{-mpe} option and the @option{-pthread}
20104 option are incompatible.
20106 @item -malign-natural
20107 @itemx -malign-power
20108 @opindex malign-natural
20109 @opindex malign-power
20110 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
20111 @option{-malign-natural} overrides the ABI-defined alignment of larger
20112 types, such as floating-point doubles, on their natural size-based boundary.
20113 The option @option{-malign-power} instructs GCC to follow the ABI-specified
20114 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
20116 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
20120 @itemx -mhard-float
20121 @opindex msoft-float
20122 @opindex mhard-float
20123 Generate code that does not use (uses) the floating-point register set.
20124 Software floating-point emulation is provided if you use the
20125 @option{-msoft-float} option, and pass the option to GCC when linking.
20127 @item -msingle-float
20128 @itemx -mdouble-float
20129 @opindex msingle-float
20130 @opindex mdouble-float
20131 Generate code for single- or double-precision floating-point operations.
20132 @option{-mdouble-float} implies @option{-msingle-float}.
20135 @opindex msimple-fpu
20136 Do not generate @code{sqrt} and @code{div} instructions for hardware
20137 floating-point unit.
20139 @item -mfpu=@var{name}
20141 Specify type of floating-point unit. Valid values for @var{name} are
20142 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
20143 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
20144 @samp{sp_full} (equivalent to @option{-msingle-float}),
20145 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
20148 @opindex mxilinx-fpu
20149 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
20152 @itemx -mno-multiple
20154 @opindex mno-multiple
20155 Generate code that uses (does not use) the load multiple word
20156 instructions and the store multiple word instructions. These
20157 instructions are generated by default on POWER systems, and not
20158 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
20159 PowerPC systems, since those instructions do not work when the
20160 processor is in little-endian mode. The exceptions are PPC740 and
20161 PPC750 which permit these instructions in little-endian mode.
20166 @opindex mno-string
20167 Generate code that uses (does not use) the load string instructions
20168 and the store string word instructions to save multiple registers and
20169 do small block moves. These instructions are generated by default on
20170 POWER systems, and not generated on PowerPC systems. Do not use
20171 @option{-mstring} on little-endian PowerPC systems, since those
20172 instructions do not work when the processor is in little-endian mode.
20173 The exceptions are PPC740 and PPC750 which permit these instructions
20174 in little-endian mode.
20179 @opindex mno-update
20180 Generate code that uses (does not use) the load or store instructions
20181 that update the base register to the address of the calculated memory
20182 location. These instructions are generated by default. If you use
20183 @option{-mno-update}, there is a small window between the time that the
20184 stack pointer is updated and the address of the previous frame is
20185 stored, which means code that walks the stack frame across interrupts or
20186 signals may get corrupted data.
20188 @item -mavoid-indexed-addresses
20189 @itemx -mno-avoid-indexed-addresses
20190 @opindex mavoid-indexed-addresses
20191 @opindex mno-avoid-indexed-addresses
20192 Generate code that tries to avoid (not avoid) the use of indexed load
20193 or store instructions. These instructions can incur a performance
20194 penalty on Power6 processors in certain situations, such as when
20195 stepping through large arrays that cross a 16M boundary. This option
20196 is enabled by default when targeting Power6 and disabled otherwise.
20199 @itemx -mno-fused-madd
20200 @opindex mfused-madd
20201 @opindex mno-fused-madd
20202 Generate code that uses (does not use) the floating-point multiply and
20203 accumulate instructions. These instructions are generated by default
20204 if hardware floating point is used. The machine-dependent
20205 @option{-mfused-madd} option is now mapped to the machine-independent
20206 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20207 mapped to @option{-ffp-contract=off}.
20213 Generate code that uses (does not use) the half-word multiply and
20214 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
20215 These instructions are generated by default when targeting those
20222 Generate code that uses (does not use) the string-search @samp{dlmzb}
20223 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
20224 generated by default when targeting those processors.
20226 @item -mno-bit-align
20228 @opindex mno-bit-align
20229 @opindex mbit-align
20230 On System V.4 and embedded PowerPC systems do not (do) force structures
20231 and unions that contain bit-fields to be aligned to the base type of the
20234 For example, by default a structure containing nothing but 8
20235 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
20236 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
20237 the structure is aligned to a 1-byte boundary and is 1 byte in
20240 @item -mno-strict-align
20241 @itemx -mstrict-align
20242 @opindex mno-strict-align
20243 @opindex mstrict-align
20244 On System V.4 and embedded PowerPC systems do not (do) assume that
20245 unaligned memory references are handled by the system.
20247 @item -mrelocatable
20248 @itemx -mno-relocatable
20249 @opindex mrelocatable
20250 @opindex mno-relocatable
20251 Generate code that allows (does not allow) a static executable to be
20252 relocated to a different address at run time. A simple embedded
20253 PowerPC system loader should relocate the entire contents of
20254 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
20255 a table of 32-bit addresses generated by this option. For this to
20256 work, all objects linked together must be compiled with
20257 @option{-mrelocatable} or @option{-mrelocatable-lib}.
20258 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
20260 @item -mrelocatable-lib
20261 @itemx -mno-relocatable-lib
20262 @opindex mrelocatable-lib
20263 @opindex mno-relocatable-lib
20264 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
20265 @code{.fixup} section to allow static executables to be relocated at
20266 run time, but @option{-mrelocatable-lib} does not use the smaller stack
20267 alignment of @option{-mrelocatable}. Objects compiled with
20268 @option{-mrelocatable-lib} may be linked with objects compiled with
20269 any combination of the @option{-mrelocatable} options.
20275 On System V.4 and embedded PowerPC systems do not (do) assume that
20276 register 2 contains a pointer to a global area pointing to the addresses
20277 used in the program.
20280 @itemx -mlittle-endian
20282 @opindex mlittle-endian
20283 On System V.4 and embedded PowerPC systems compile code for the
20284 processor in little-endian mode. The @option{-mlittle-endian} option is
20285 the same as @option{-mlittle}.
20288 @itemx -mbig-endian
20290 @opindex mbig-endian
20291 On System V.4 and embedded PowerPC systems compile code for the
20292 processor in big-endian mode. The @option{-mbig-endian} option is
20293 the same as @option{-mbig}.
20295 @item -mdynamic-no-pic
20296 @opindex mdynamic-no-pic
20297 On Darwin and Mac OS X systems, compile code so that it is not
20298 relocatable, but that its external references are relocatable. The
20299 resulting code is suitable for applications, but not shared
20302 @item -msingle-pic-base
20303 @opindex msingle-pic-base
20304 Treat the register used for PIC addressing as read-only, rather than
20305 loading it in the prologue for each function. The runtime system is
20306 responsible for initializing this register with an appropriate value
20307 before execution begins.
20309 @item -mprioritize-restricted-insns=@var{priority}
20310 @opindex mprioritize-restricted-insns
20311 This option controls the priority that is assigned to
20312 dispatch-slot restricted instructions during the second scheduling
20313 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20314 or @samp{2} to assign no, highest, or second-highest (respectively)
20315 priority to dispatch-slot restricted
20318 @item -msched-costly-dep=@var{dependence_type}
20319 @opindex msched-costly-dep
20320 This option controls which dependences are considered costly
20321 by the target during instruction scheduling. The argument
20322 @var{dependence_type} takes one of the following values:
20326 No dependence is costly.
20329 All dependences are costly.
20331 @item @samp{true_store_to_load}
20332 A true dependence from store to load is costly.
20334 @item @samp{store_to_load}
20335 Any dependence from store to load is costly.
20338 Any dependence for which the latency is greater than or equal to
20339 @var{number} is costly.
20342 @item -minsert-sched-nops=@var{scheme}
20343 @opindex minsert-sched-nops
20344 This option controls which NOP insertion scheme is used during
20345 the second scheduling pass. The argument @var{scheme} takes one of the
20353 Pad with NOPs any dispatch group that has vacant issue slots,
20354 according to the scheduler's grouping.
20356 @item @samp{regroup_exact}
20357 Insert NOPs to force costly dependent insns into
20358 separate groups. Insert exactly as many NOPs as needed to force an insn
20359 to a new group, according to the estimated processor grouping.
20362 Insert NOPs to force costly dependent insns into
20363 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20367 @opindex mcall-sysv
20368 On System V.4 and embedded PowerPC systems compile code using calling
20369 conventions that adhere to the March 1995 draft of the System V
20370 Application Binary Interface, PowerPC processor supplement. This is the
20371 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20373 @item -mcall-sysv-eabi
20375 @opindex mcall-sysv-eabi
20376 @opindex mcall-eabi
20377 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20379 @item -mcall-sysv-noeabi
20380 @opindex mcall-sysv-noeabi
20381 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20383 @item -mcall-aixdesc
20385 On System V.4 and embedded PowerPC systems compile code for the AIX
20389 @opindex mcall-linux
20390 On System V.4 and embedded PowerPC systems compile code for the
20391 Linux-based GNU system.
20393 @item -mcall-freebsd
20394 @opindex mcall-freebsd
20395 On System V.4 and embedded PowerPC systems compile code for the
20396 FreeBSD operating system.
20398 @item -mcall-netbsd
20399 @opindex mcall-netbsd
20400 On System V.4 and embedded PowerPC systems compile code for the
20401 NetBSD operating system.
20403 @item -mcall-openbsd
20404 @opindex mcall-netbsd
20405 On System V.4 and embedded PowerPC systems compile code for the
20406 OpenBSD operating system.
20408 @item -maix-struct-return
20409 @opindex maix-struct-return
20410 Return all structures in memory (as specified by the AIX ABI)@.
20412 @item -msvr4-struct-return
20413 @opindex msvr4-struct-return
20414 Return structures smaller than 8 bytes in registers (as specified by the
20417 @item -mabi=@var{abi-type}
20419 Extend the current ABI with a particular extension, or remove such extension.
20420 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
20421 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
20422 @samp{elfv1}, @samp{elfv2}@.
20426 Extend the current ABI with SPE ABI extensions. This does not change
20427 the default ABI, instead it adds the SPE ABI extensions to the current
20431 @opindex mabi=no-spe
20432 Disable Book-E SPE ABI extensions for the current ABI@.
20434 @item -mabi=ibmlongdouble
20435 @opindex mabi=ibmlongdouble
20436 Change the current ABI to use IBM extended-precision long double.
20437 This is a PowerPC 32-bit SYSV ABI option.
20439 @item -mabi=ieeelongdouble
20440 @opindex mabi=ieeelongdouble
20441 Change the current ABI to use IEEE extended-precision long double.
20442 This is a PowerPC 32-bit Linux ABI option.
20445 @opindex mabi=elfv1
20446 Change the current ABI to use the ELFv1 ABI.
20447 This is the default ABI for big-endian PowerPC 64-bit Linux.
20448 Overriding the default ABI requires special system support and is
20449 likely to fail in spectacular ways.
20452 @opindex mabi=elfv2
20453 Change the current ABI to use the ELFv2 ABI.
20454 This is the default ABI for little-endian PowerPC 64-bit Linux.
20455 Overriding the default ABI requires special system support and is
20456 likely to fail in spectacular ways.
20459 @itemx -mno-prototype
20460 @opindex mprototype
20461 @opindex mno-prototype
20462 On System V.4 and embedded PowerPC systems assume that all calls to
20463 variable argument functions are properly prototyped. Otherwise, the
20464 compiler must insert an instruction before every non-prototyped call to
20465 set or clear bit 6 of the condition code register (@code{CR}) to
20466 indicate whether floating-point values are passed in the floating-point
20467 registers in case the function takes variable arguments. With
20468 @option{-mprototype}, only calls to prototyped variable argument functions
20469 set or clear the bit.
20473 On embedded PowerPC systems, assume that the startup module is called
20474 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20475 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20480 On embedded PowerPC systems, assume that the startup module is called
20481 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20486 On embedded PowerPC systems, assume that the startup module is called
20487 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20490 @item -myellowknife
20491 @opindex myellowknife
20492 On embedded PowerPC systems, assume that the startup module is called
20493 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20498 On System V.4 and embedded PowerPC systems, specify that you are
20499 compiling for a VxWorks system.
20503 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
20504 header to indicate that @samp{eabi} extended relocations are used.
20510 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20511 Embedded Applications Binary Interface (EABI), which is a set of
20512 modifications to the System V.4 specifications. Selecting @option{-meabi}
20513 means that the stack is aligned to an 8-byte boundary, a function
20514 @code{__eabi} is called from @code{main} to set up the EABI
20515 environment, and the @option{-msdata} option can use both @code{r2} and
20516 @code{r13} to point to two separate small data areas. Selecting
20517 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20518 no EABI initialization function is called from @code{main}, and the
20519 @option{-msdata} option only uses @code{r13} to point to a single
20520 small data area. The @option{-meabi} option is on by default if you
20521 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20524 @opindex msdata=eabi
20525 On System V.4 and embedded PowerPC systems, put small initialized
20526 @code{const} global and static data in the @code{.sdata2} section, which
20527 is pointed to by register @code{r2}. Put small initialized
20528 non-@code{const} global and static data in the @code{.sdata} section,
20529 which is pointed to by register @code{r13}. Put small uninitialized
20530 global and static data in the @code{.sbss} section, which is adjacent to
20531 the @code{.sdata} section. The @option{-msdata=eabi} option is
20532 incompatible with the @option{-mrelocatable} option. The
20533 @option{-msdata=eabi} option also sets the @option{-memb} option.
20536 @opindex msdata=sysv
20537 On System V.4 and embedded PowerPC systems, put small global and static
20538 data in the @code{.sdata} section, which is pointed to by register
20539 @code{r13}. Put small uninitialized global and static data in the
20540 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
20541 The @option{-msdata=sysv} option is incompatible with the
20542 @option{-mrelocatable} option.
20544 @item -msdata=default
20546 @opindex msdata=default
20548 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20549 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20550 same as @option{-msdata=sysv}.
20553 @opindex msdata=data
20554 On System V.4 and embedded PowerPC systems, put small global
20555 data in the @code{.sdata} section. Put small uninitialized global
20556 data in the @code{.sbss} section. Do not use register @code{r13}
20557 to address small data however. This is the default behavior unless
20558 other @option{-msdata} options are used.
20562 @opindex msdata=none
20564 On embedded PowerPC systems, put all initialized global and static data
20565 in the @code{.data} section, and all uninitialized data in the
20566 @code{.bss} section.
20568 @item -mblock-move-inline-limit=@var{num}
20569 @opindex mblock-move-inline-limit
20570 Inline all block moves (such as calls to @code{memcpy} or structure
20571 copies) less than or equal to @var{num} bytes. The minimum value for
20572 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20573 targets. The default value is target-specific.
20577 @cindex smaller data references (PowerPC)
20578 @cindex .sdata/.sdata2 references (PowerPC)
20579 On embedded PowerPC systems, put global and static items less than or
20580 equal to @var{num} bytes into the small data or BSS sections instead of
20581 the normal data or BSS section. By default, @var{num} is 8. The
20582 @option{-G @var{num}} switch is also passed to the linker.
20583 All modules should be compiled with the same @option{-G @var{num}} value.
20586 @itemx -mno-regnames
20588 @opindex mno-regnames
20589 On System V.4 and embedded PowerPC systems do (do not) emit register
20590 names in the assembly language output using symbolic forms.
20593 @itemx -mno-longcall
20595 @opindex mno-longcall
20596 By default assume that all calls are far away so that a longer and more
20597 expensive calling sequence is required. This is required for calls
20598 farther than 32 megabytes (33,554,432 bytes) from the current location.
20599 A short call is generated if the compiler knows
20600 the call cannot be that far away. This setting can be overridden by
20601 the @code{shortcall} function attribute, or by @code{#pragma
20604 Some linkers are capable of detecting out-of-range calls and generating
20605 glue code on the fly. On these systems, long calls are unnecessary and
20606 generate slower code. As of this writing, the AIX linker can do this,
20607 as can the GNU linker for PowerPC/64. It is planned to add this feature
20608 to the GNU linker for 32-bit PowerPC systems as well.
20610 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20611 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20612 addresses represent the callee and the branch island. The
20613 Darwin/PPC linker prefers the first address and generates a @code{bl
20614 callee} if the PPC @code{bl} instruction reaches the callee directly;
20615 otherwise, the linker generates @code{bl L42} to call the branch
20616 island. The branch island is appended to the body of the
20617 calling function; it computes the full 32-bit address of the callee
20620 On Mach-O (Darwin) systems, this option directs the compiler emit to
20621 the glue for every direct call, and the Darwin linker decides whether
20622 to use or discard it.
20624 In the future, GCC may ignore all longcall specifications
20625 when the linker is known to generate glue.
20627 @item -mtls-markers
20628 @itemx -mno-tls-markers
20629 @opindex mtls-markers
20630 @opindex mno-tls-markers
20631 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20632 specifying the function argument. The relocation allows the linker to
20633 reliably associate function call with argument setup instructions for
20634 TLS optimization, which in turn allows GCC to better schedule the
20639 Adds support for multithreading with the @dfn{pthreads} library.
20640 This option sets flags for both the preprocessor and linker.
20645 This option enables use of the reciprocal estimate and
20646 reciprocal square root estimate instructions with additional
20647 Newton-Raphson steps to increase precision instead of doing a divide or
20648 square root and divide for floating-point arguments. You should use
20649 the @option{-ffast-math} option when using @option{-mrecip} (or at
20650 least @option{-funsafe-math-optimizations},
20651 @option{-finite-math-only}, @option{-freciprocal-math} and
20652 @option{-fno-trapping-math}). Note that while the throughput of the
20653 sequence is generally higher than the throughput of the non-reciprocal
20654 instruction, the precision of the sequence can be decreased by up to 2
20655 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20658 @item -mrecip=@var{opt}
20659 @opindex mrecip=opt
20660 This option controls which reciprocal estimate instructions
20661 may be used. @var{opt} is a comma-separated list of options, which may
20662 be preceded by a @code{!} to invert the option:
20667 Enable all estimate instructions.
20670 Enable the default instructions, equivalent to @option{-mrecip}.
20673 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20676 Enable the reciprocal approximation instructions for both
20677 single and double precision.
20680 Enable the single-precision reciprocal approximation instructions.
20683 Enable the double-precision reciprocal approximation instructions.
20686 Enable the reciprocal square root approximation instructions for both
20687 single and double precision.
20690 Enable the single-precision reciprocal square root approximation instructions.
20693 Enable the double-precision reciprocal square root approximation instructions.
20697 So, for example, @option{-mrecip=all,!rsqrtd} enables
20698 all of the reciprocal estimate instructions, except for the
20699 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20700 which handle the double-precision reciprocal square root calculations.
20702 @item -mrecip-precision
20703 @itemx -mno-recip-precision
20704 @opindex mrecip-precision
20705 Assume (do not assume) that the reciprocal estimate instructions
20706 provide higher-precision estimates than is mandated by the PowerPC
20707 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20708 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20709 The double-precision square root estimate instructions are not generated by
20710 default on low-precision machines, since they do not provide an
20711 estimate that converges after three steps.
20713 @item -mveclibabi=@var{type}
20714 @opindex mveclibabi
20715 Specifies the ABI type to use for vectorizing intrinsics using an
20716 external library. The only type supported at present is @samp{mass},
20717 which specifies to use IBM's Mathematical Acceleration Subsystem
20718 (MASS) libraries for vectorizing intrinsics using external libraries.
20719 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20720 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20721 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20722 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20723 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20724 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20725 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20726 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20727 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20728 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20729 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20730 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20731 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20732 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20733 for power7. Both @option{-ftree-vectorize} and
20734 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20735 libraries must be specified at link time.
20740 Generate (do not generate) the @code{friz} instruction when the
20741 @option{-funsafe-math-optimizations} option is used to optimize
20742 rounding of floating-point values to 64-bit integer and back to floating
20743 point. The @code{friz} instruction does not return the same value if
20744 the floating-point number is too large to fit in an integer.
20746 @item -mpointers-to-nested-functions
20747 @itemx -mno-pointers-to-nested-functions
20748 @opindex mpointers-to-nested-functions
20749 Generate (do not generate) code to load up the static chain register
20750 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20751 systems where a function pointer points to a 3-word descriptor giving
20752 the function address, TOC value to be loaded in register @code{r2}, and
20753 static chain value to be loaded in register @code{r11}. The
20754 @option{-mpointers-to-nested-functions} is on by default. You cannot
20755 call through pointers to nested functions or pointers
20756 to functions compiled in other languages that use the static chain if
20757 you use the @option{-mno-pointers-to-nested-functions}.
20759 @item -msave-toc-indirect
20760 @itemx -mno-save-toc-indirect
20761 @opindex msave-toc-indirect
20762 Generate (do not generate) code to save the TOC value in the reserved
20763 stack location in the function prologue if the function calls through
20764 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20765 saved in the prologue, it is saved just before the call through the
20766 pointer. The @option{-mno-save-toc-indirect} option is the default.
20768 @item -mcompat-align-parm
20769 @itemx -mno-compat-align-parm
20770 @opindex mcompat-align-parm
20771 Generate (do not generate) code to pass structure parameters with a
20772 maximum alignment of 64 bits, for compatibility with older versions
20775 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20776 structure parameter on a 128-bit boundary when that structure contained
20777 a member requiring 128-bit alignment. This is corrected in more
20778 recent versions of GCC. This option may be used to generate code
20779 that is compatible with functions compiled with older versions of
20782 The @option{-mno-compat-align-parm} option is the default.
20786 @subsection RX Options
20789 These command-line options are defined for RX targets:
20792 @item -m64bit-doubles
20793 @itemx -m32bit-doubles
20794 @opindex m64bit-doubles
20795 @opindex m32bit-doubles
20796 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20797 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20798 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20799 works on 32-bit values, which is why the default is
20800 @option{-m32bit-doubles}.
20806 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20807 floating-point hardware. The default is enabled for the RX600
20808 series and disabled for the RX200 series.
20810 Floating-point instructions are only generated for 32-bit floating-point
20811 values, however, so the FPU hardware is not used for doubles if the
20812 @option{-m64bit-doubles} option is used.
20814 @emph{Note} If the @option{-fpu} option is enabled then
20815 @option{-funsafe-math-optimizations} is also enabled automatically.
20816 This is because the RX FPU instructions are themselves unsafe.
20818 @item -mcpu=@var{name}
20820 Selects the type of RX CPU to be targeted. Currently three types are
20821 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20822 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20824 The only difference between @samp{RX600} and @samp{RX610} is that the
20825 @samp{RX610} does not support the @code{MVTIPL} instruction.
20827 The @samp{RX200} series does not have a hardware floating-point unit
20828 and so @option{-nofpu} is enabled by default when this type is
20831 @item -mbig-endian-data
20832 @itemx -mlittle-endian-data
20833 @opindex mbig-endian-data
20834 @opindex mlittle-endian-data
20835 Store data (but not code) in the big-endian format. The default is
20836 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20839 @item -msmall-data-limit=@var{N}
20840 @opindex msmall-data-limit
20841 Specifies the maximum size in bytes of global and static variables
20842 which can be placed into the small data area. Using the small data
20843 area can lead to smaller and faster code, but the size of area is
20844 limited and it is up to the programmer to ensure that the area does
20845 not overflow. Also when the small data area is used one of the RX's
20846 registers (usually @code{r13}) is reserved for use pointing to this
20847 area, so it is no longer available for use by the compiler. This
20848 could result in slower and/or larger code if variables are pushed onto
20849 the stack instead of being held in this register.
20851 Note, common variables (variables that have not been initialized) and
20852 constants are not placed into the small data area as they are assigned
20853 to other sections in the output executable.
20855 The default value is zero, which disables this feature. Note, this
20856 feature is not enabled by default with higher optimization levels
20857 (@option{-O2} etc) because of the potentially detrimental effects of
20858 reserving a register. It is up to the programmer to experiment and
20859 discover whether this feature is of benefit to their program. See the
20860 description of the @option{-mpid} option for a description of how the
20861 actual register to hold the small data area pointer is chosen.
20867 Use the simulator runtime. The default is to use the libgloss
20868 board-specific runtime.
20870 @item -mas100-syntax
20871 @itemx -mno-as100-syntax
20872 @opindex mas100-syntax
20873 @opindex mno-as100-syntax
20874 When generating assembler output use a syntax that is compatible with
20875 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20876 assembler, but it has some restrictions so it is not generated by default.
20878 @item -mmax-constant-size=@var{N}
20879 @opindex mmax-constant-size
20880 Specifies the maximum size, in bytes, of a constant that can be used as
20881 an operand in a RX instruction. Although the RX instruction set does
20882 allow constants of up to 4 bytes in length to be used in instructions,
20883 a longer value equates to a longer instruction. Thus in some
20884 circumstances it can be beneficial to restrict the size of constants
20885 that are used in instructions. Constants that are too big are instead
20886 placed into a constant pool and referenced via register indirection.
20888 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20889 or 4 means that constants of any size are allowed.
20893 Enable linker relaxation. Linker relaxation is a process whereby the
20894 linker attempts to reduce the size of a program by finding shorter
20895 versions of various instructions. Disabled by default.
20897 @item -mint-register=@var{N}
20898 @opindex mint-register
20899 Specify the number of registers to reserve for fast interrupt handler
20900 functions. The value @var{N} can be between 0 and 4. A value of 1
20901 means that register @code{r13} is reserved for the exclusive use
20902 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20903 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20904 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20905 A value of 0, the default, does not reserve any registers.
20907 @item -msave-acc-in-interrupts
20908 @opindex msave-acc-in-interrupts
20909 Specifies that interrupt handler functions should preserve the
20910 accumulator register. This is only necessary if normal code might use
20911 the accumulator register, for example because it performs 64-bit
20912 multiplications. The default is to ignore the accumulator as this
20913 makes the interrupt handlers faster.
20919 Enables the generation of position independent data. When enabled any
20920 access to constant data is done via an offset from a base address
20921 held in a register. This allows the location of constant data to be
20922 determined at run time without requiring the executable to be
20923 relocated, which is a benefit to embedded applications with tight
20924 memory constraints. Data that can be modified is not affected by this
20927 Note, using this feature reserves a register, usually @code{r13}, for
20928 the constant data base address. This can result in slower and/or
20929 larger code, especially in complicated functions.
20931 The actual register chosen to hold the constant data base address
20932 depends upon whether the @option{-msmall-data-limit} and/or the
20933 @option{-mint-register} command-line options are enabled. Starting
20934 with register @code{r13} and proceeding downwards, registers are
20935 allocated first to satisfy the requirements of @option{-mint-register},
20936 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20937 is possible for the small data area register to be @code{r8} if both
20938 @option{-mint-register=4} and @option{-mpid} are specified on the
20941 By default this feature is not enabled. The default can be restored
20942 via the @option{-mno-pid} command-line option.
20944 @item -mno-warn-multiple-fast-interrupts
20945 @itemx -mwarn-multiple-fast-interrupts
20946 @opindex mno-warn-multiple-fast-interrupts
20947 @opindex mwarn-multiple-fast-interrupts
20948 Prevents GCC from issuing a warning message if it finds more than one
20949 fast interrupt handler when it is compiling a file. The default is to
20950 issue a warning for each extra fast interrupt handler found, as the RX
20951 only supports one such interrupt.
20955 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20956 has special significance to the RX port when used with the
20957 @code{interrupt} function attribute. This attribute indicates a
20958 function intended to process fast interrupts. GCC ensures
20959 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20960 and/or @code{r13} and only provided that the normal use of the
20961 corresponding registers have been restricted via the
20962 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20965 @node S/390 and zSeries Options
20966 @subsection S/390 and zSeries Options
20967 @cindex S/390 and zSeries Options
20969 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20973 @itemx -msoft-float
20974 @opindex mhard-float
20975 @opindex msoft-float
20976 Use (do not use) the hardware floating-point instructions and registers
20977 for floating-point operations. When @option{-msoft-float} is specified,
20978 functions in @file{libgcc.a} are used to perform floating-point
20979 operations. When @option{-mhard-float} is specified, the compiler
20980 generates IEEE floating-point instructions. This is the default.
20983 @itemx -mno-hard-dfp
20985 @opindex mno-hard-dfp
20986 Use (do not use) the hardware decimal-floating-point instructions for
20987 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20988 specified, functions in @file{libgcc.a} are used to perform
20989 decimal-floating-point operations. When @option{-mhard-dfp} is
20990 specified, the compiler generates decimal-floating-point hardware
20991 instructions. This is the default for @option{-march=z9-ec} or higher.
20993 @item -mlong-double-64
20994 @itemx -mlong-double-128
20995 @opindex mlong-double-64
20996 @opindex mlong-double-128
20997 These switches control the size of @code{long double} type. A size
20998 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20999 type. This is the default.
21002 @itemx -mno-backchain
21003 @opindex mbackchain
21004 @opindex mno-backchain
21005 Store (do not store) the address of the caller's frame as backchain pointer
21006 into the callee's stack frame.
21007 A backchain may be needed to allow debugging using tools that do not understand
21008 DWARF 2 call frame information.
21009 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
21010 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
21011 the backchain is placed into the topmost word of the 96/160 byte register
21014 In general, code compiled with @option{-mbackchain} is call-compatible with
21015 code compiled with @option{-mmo-backchain}; however, use of the backchain
21016 for debugging purposes usually requires that the whole binary is built with
21017 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
21018 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21019 to build a linux kernel use @option{-msoft-float}.
21021 The default is to not maintain the backchain.
21023 @item -mpacked-stack
21024 @itemx -mno-packed-stack
21025 @opindex mpacked-stack
21026 @opindex mno-packed-stack
21027 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
21028 specified, the compiler uses the all fields of the 96/160 byte register save
21029 area only for their default purpose; unused fields still take up stack space.
21030 When @option{-mpacked-stack} is specified, register save slots are densely
21031 packed at the top of the register save area; unused space is reused for other
21032 purposes, allowing for more efficient use of the available stack space.
21033 However, when @option{-mbackchain} is also in effect, the topmost word of
21034 the save area is always used to store the backchain, and the return address
21035 register is always saved two words below the backchain.
21037 As long as the stack frame backchain is not used, code generated with
21038 @option{-mpacked-stack} is call-compatible with code generated with
21039 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
21040 S/390 or zSeries generated code that uses the stack frame backchain at run
21041 time, not just for debugging purposes. Such code is not call-compatible
21042 with code compiled with @option{-mpacked-stack}. Also, note that the
21043 combination of @option{-mbackchain},
21044 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21045 to build a linux kernel use @option{-msoft-float}.
21047 The default is to not use the packed stack layout.
21050 @itemx -mno-small-exec
21051 @opindex msmall-exec
21052 @opindex mno-small-exec
21053 Generate (or do not generate) code using the @code{bras} instruction
21054 to do subroutine calls.
21055 This only works reliably if the total executable size does not
21056 exceed 64k. The default is to use the @code{basr} instruction instead,
21057 which does not have this limitation.
21063 When @option{-m31} is specified, generate code compliant to the
21064 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
21065 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
21066 particular to generate 64-bit instructions. For the @samp{s390}
21067 targets, the default is @option{-m31}, while the @samp{s390x}
21068 targets default to @option{-m64}.
21074 When @option{-mzarch} is specified, generate code using the
21075 instructions available on z/Architecture.
21076 When @option{-mesa} is specified, generate code using the
21077 instructions available on ESA/390. Note that @option{-mesa} is
21078 not possible with @option{-m64}.
21079 When generating code compliant to the GNU/Linux for S/390 ABI,
21080 the default is @option{-mesa}. When generating code compliant
21081 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
21087 Generate (or do not generate) code using the @code{mvcle} instruction
21088 to perform block moves. When @option{-mno-mvcle} is specified,
21089 use a @code{mvc} loop instead. This is the default unless optimizing for
21096 Print (or do not print) additional debug information when compiling.
21097 The default is to not print debug information.
21099 @item -march=@var{cpu-type}
21101 Generate code that runs on @var{cpu-type}, which is the name of a system
21102 representing a certain processor type. Possible values for
21103 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
21104 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
21105 When generating code using the instructions available on z/Architecture,
21106 the default is @option{-march=z900}. Otherwise, the default is
21107 @option{-march=g5}.
21109 @item -mtune=@var{cpu-type}
21111 Tune to @var{cpu-type} everything applicable about the generated code,
21112 except for the ABI and the set of available instructions.
21113 The list of @var{cpu-type} values is the same as for @option{-march}.
21114 The default is the value used for @option{-march}.
21117 @itemx -mno-tpf-trace
21118 @opindex mtpf-trace
21119 @opindex mno-tpf-trace
21120 Generate code that adds (does not add) in TPF OS specific branches to trace
21121 routines in the operating system. This option is off by default, even
21122 when compiling for the TPF OS@.
21125 @itemx -mno-fused-madd
21126 @opindex mfused-madd
21127 @opindex mno-fused-madd
21128 Generate code that uses (does not use) the floating-point multiply and
21129 accumulate instructions. These instructions are generated by default if
21130 hardware floating point is used.
21132 @item -mwarn-framesize=@var{framesize}
21133 @opindex mwarn-framesize
21134 Emit a warning if the current function exceeds the given frame size. Because
21135 this is a compile-time check it doesn't need to be a real problem when the program
21136 runs. It is intended to identify functions that most probably cause
21137 a stack overflow. It is useful to be used in an environment with limited stack
21138 size e.g.@: the linux kernel.
21140 @item -mwarn-dynamicstack
21141 @opindex mwarn-dynamicstack
21142 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
21143 arrays. This is generally a bad idea with a limited stack size.
21145 @item -mstack-guard=@var{stack-guard}
21146 @itemx -mstack-size=@var{stack-size}
21147 @opindex mstack-guard
21148 @opindex mstack-size
21149 If these options are provided the S/390 back end emits additional instructions in
21150 the function prologue that trigger a trap if the stack size is @var{stack-guard}
21151 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
21152 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
21153 the frame size of the compiled function is chosen.
21154 These options are intended to be used to help debugging stack overflow problems.
21155 The additionally emitted code causes only little overhead and hence can also be
21156 used in production-like systems without greater performance degradation. The given
21157 values have to be exact powers of 2 and @var{stack-size} has to be greater than
21158 @var{stack-guard} without exceeding 64k.
21159 In order to be efficient the extra code makes the assumption that the stack starts
21160 at an address aligned to the value given by @var{stack-size}.
21161 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
21163 @item -mhotpatch[=@var{halfwords}]
21164 @itemx -mno-hotpatch
21166 If the hotpatch option is enabled, a ``hot-patching'' function
21167 prologue is generated for all functions in the compilation unit.
21168 The funtion label is prepended with the given number of two-byte
21169 Nop instructions (@var{halfwords}, maximum 1000000) or 12 Nop
21170 instructions if no argument is present. Functions with a
21171 hot-patching prologue are never inlined automatically, and a
21172 hot-patching prologue is never generated for functions
21173 that are explicitly inline.
21175 This option can be overridden for individual functions with the
21176 @code{hotpatch} attribute.
21179 @node Score Options
21180 @subsection Score Options
21181 @cindex Score Options
21183 These options are defined for Score implementations:
21188 Compile code for big-endian mode. This is the default.
21192 Compile code for little-endian mode.
21196 Disable generation of @code{bcnz} instructions.
21200 Enable generation of unaligned load and store instructions.
21204 Enable the use of multiply-accumulate instructions. Disabled by default.
21208 Specify the SCORE5 as the target architecture.
21212 Specify the SCORE5U of the target architecture.
21216 Specify the SCORE7 as the target architecture. This is the default.
21220 Specify the SCORE7D as the target architecture.
21224 @subsection SH Options
21226 These @samp{-m} options are defined for the SH implementations:
21231 Generate code for the SH1.
21235 Generate code for the SH2.
21238 Generate code for the SH2e.
21242 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
21243 that the floating-point unit is not used.
21245 @item -m2a-single-only
21246 @opindex m2a-single-only
21247 Generate code for the SH2a-FPU, in such a way that no double-precision
21248 floating-point operations are used.
21251 @opindex m2a-single
21252 Generate code for the SH2a-FPU assuming the floating-point unit is in
21253 single-precision mode by default.
21257 Generate code for the SH2a-FPU assuming the floating-point unit is in
21258 double-precision mode by default.
21262 Generate code for the SH3.
21266 Generate code for the SH3e.
21270 Generate code for the SH4 without a floating-point unit.
21272 @item -m4-single-only
21273 @opindex m4-single-only
21274 Generate code for the SH4 with a floating-point unit that only
21275 supports single-precision arithmetic.
21279 Generate code for the SH4 assuming the floating-point unit is in
21280 single-precision mode by default.
21284 Generate code for the SH4.
21288 Generate code for SH4-100.
21290 @item -m4-100-nofpu
21291 @opindex m4-100-nofpu
21292 Generate code for SH4-100 in such a way that the
21293 floating-point unit is not used.
21295 @item -m4-100-single
21296 @opindex m4-100-single
21297 Generate code for SH4-100 assuming the floating-point unit is in
21298 single-precision mode by default.
21300 @item -m4-100-single-only
21301 @opindex m4-100-single-only
21302 Generate code for SH4-100 in such a way that no double-precision
21303 floating-point operations are used.
21307 Generate code for SH4-200.
21309 @item -m4-200-nofpu
21310 @opindex m4-200-nofpu
21311 Generate code for SH4-200 without in such a way that the
21312 floating-point unit is not used.
21314 @item -m4-200-single
21315 @opindex m4-200-single
21316 Generate code for SH4-200 assuming the floating-point unit is in
21317 single-precision mode by default.
21319 @item -m4-200-single-only
21320 @opindex m4-200-single-only
21321 Generate code for SH4-200 in such a way that no double-precision
21322 floating-point operations are used.
21326 Generate code for SH4-300.
21328 @item -m4-300-nofpu
21329 @opindex m4-300-nofpu
21330 Generate code for SH4-300 without in such a way that the
21331 floating-point unit is not used.
21333 @item -m4-300-single
21334 @opindex m4-300-single
21335 Generate code for SH4-300 in such a way that no double-precision
21336 floating-point operations are used.
21338 @item -m4-300-single-only
21339 @opindex m4-300-single-only
21340 Generate code for SH4-300 in such a way that no double-precision
21341 floating-point operations are used.
21345 Generate code for SH4-340 (no MMU, no FPU).
21349 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21354 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21355 floating-point unit is not used.
21357 @item -m4a-single-only
21358 @opindex m4a-single-only
21359 Generate code for the SH4a, in such a way that no double-precision
21360 floating-point operations are used.
21363 @opindex m4a-single
21364 Generate code for the SH4a assuming the floating-point unit is in
21365 single-precision mode by default.
21369 Generate code for the SH4a.
21373 Same as @option{-m4a-nofpu}, except that it implicitly passes
21374 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21375 instructions at the moment.
21378 @opindex m5-32media
21379 Generate 32-bit code for SHmedia.
21381 @item -m5-32media-nofpu
21382 @opindex m5-32media-nofpu
21383 Generate 32-bit code for SHmedia in such a way that the
21384 floating-point unit is not used.
21387 @opindex m5-64media
21388 Generate 64-bit code for SHmedia.
21390 @item -m5-64media-nofpu
21391 @opindex m5-64media-nofpu
21392 Generate 64-bit code for SHmedia in such a way that the
21393 floating-point unit is not used.
21396 @opindex m5-compact
21397 Generate code for SHcompact.
21399 @item -m5-compact-nofpu
21400 @opindex m5-compact-nofpu
21401 Generate code for SHcompact in such a way that the
21402 floating-point unit is not used.
21406 Compile code for the processor in big-endian mode.
21410 Compile code for the processor in little-endian mode.
21414 Align doubles at 64-bit boundaries. Note that this changes the calling
21415 conventions, and thus some functions from the standard C library do
21416 not work unless you recompile it first with @option{-mdalign}.
21420 Shorten some address references at link time, when possible; uses the
21421 linker option @option{-relax}.
21425 Use 32-bit offsets in @code{switch} tables. The default is to use
21430 Enable the use of bit manipulation instructions on SH2A.
21434 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21435 alignment constraints.
21439 Comply with the calling conventions defined by Renesas.
21442 @opindex mno-renesas
21443 Comply with the calling conventions defined for GCC before the Renesas
21444 conventions were available. This option is the default for all
21445 targets of the SH toolchain.
21448 @opindex mnomacsave
21449 Mark the @code{MAC} register as call-clobbered, even if
21450 @option{-mrenesas} is given.
21456 Control the IEEE compliance of floating-point comparisons, which affects the
21457 handling of cases where the result of a comparison is unordered. By default
21458 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21459 enabled @option{-mno-ieee} is implicitly set, which results in faster
21460 floating-point greater-equal and less-equal comparisons. The implcit settings
21461 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21463 @item -minline-ic_invalidate
21464 @opindex minline-ic_invalidate
21465 Inline code to invalidate instruction cache entries after setting up
21466 nested function trampolines.
21467 This option has no effect if @option{-musermode} is in effect and the selected
21468 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21470 If the selected code generation option does not allow the use of the @code{icbi}
21471 instruction, and @option{-musermode} is not in effect, the inlined code
21472 manipulates the instruction cache address array directly with an associative
21473 write. This not only requires privileged mode at run time, but it also
21474 fails if the cache line had been mapped via the TLB and has become unmapped.
21478 Dump instruction size and location in the assembly code.
21481 @opindex mpadstruct
21482 This option is deprecated. It pads structures to multiple of 4 bytes,
21483 which is incompatible with the SH ABI@.
21485 @item -matomic-model=@var{model}
21486 @opindex matomic-model=@var{model}
21487 Sets the model of atomic operations and additional parameters as a comma
21488 separated list. For details on the atomic built-in functions see
21489 @ref{__atomic Builtins}. The following models and parameters are supported:
21494 Disable compiler generated atomic sequences and emit library calls for atomic
21495 operations. This is the default if the target is not @code{sh*-*-linux*}.
21498 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21499 built-in functions. The generated atomic sequences require additional support
21500 from the interrupt/exception handling code of the system and are only suitable
21501 for SH3* and SH4* single-core systems. This option is enabled by default when
21502 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21503 this option also partially utilizes the hardware atomic instructions
21504 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21505 @samp{strict} is specified.
21508 Generate software atomic sequences that use a variable in the thread control
21509 block. This is a variation of the gUSA sequences which can also be used on
21510 SH1* and SH2* targets. The generated atomic sequences require additional
21511 support from the interrupt/exception handling code of the system and are only
21512 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21513 parameter has to be specified as well.
21516 Generate software atomic sequences that temporarily disable interrupts by
21517 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21518 in privileged mode and is only suitable for single-core systems. Additional
21519 support from the interrupt/exception handling code of the system is not
21520 required. This model is enabled by default when the target is
21521 @code{sh*-*-linux*} and SH1* or SH2*.
21524 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21525 instructions only. This is only available on SH4A and is suitable for
21526 multi-core systems. Since the hardware instructions support only 32 bit atomic
21527 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21528 Code compiled with this option is also compatible with other software
21529 atomic model interrupt/exception handling systems if executed on an SH4A
21530 system. Additional support from the interrupt/exception handling code of the
21531 system is not required for this model.
21534 This parameter specifies the offset in bytes of the variable in the thread
21535 control block structure that should be used by the generated atomic sequences
21536 when the @samp{soft-tcb} model has been selected. For other models this
21537 parameter is ignored. The specified value must be an integer multiple of four
21538 and in the range 0-1020.
21541 This parameter prevents mixed usage of multiple atomic models, even if they
21542 are compatible, and makes the compiler generate atomic sequences of the
21543 specified model only.
21549 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21550 Notice that depending on the particular hardware and software configuration
21551 this can degrade overall performance due to the operand cache line flushes
21552 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21553 processors the @code{tas.b} instruction must be used with caution since it
21554 can result in data corruption for certain cache configurations.
21557 @opindex mprefergot
21558 When generating position-independent code, emit function calls using
21559 the Global Offset Table instead of the Procedure Linkage Table.
21562 @itemx -mno-usermode
21564 @opindex mno-usermode
21565 Don't allow (allow) the compiler generating privileged mode code. Specifying
21566 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21567 inlined code would not work in user mode. @option{-musermode} is the default
21568 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21569 @option{-musermode} has no effect, since there is no user mode.
21571 @item -multcost=@var{number}
21572 @opindex multcost=@var{number}
21573 Set the cost to assume for a multiply insn.
21575 @item -mdiv=@var{strategy}
21576 @opindex mdiv=@var{strategy}
21577 Set the division strategy to be used for integer division operations.
21578 For SHmedia @var{strategy} can be one of:
21583 Performs the operation in floating point. This has a very high latency,
21584 but needs only a few instructions, so it might be a good choice if
21585 your code has enough easily-exploitable ILP to allow the compiler to
21586 schedule the floating-point instructions together with other instructions.
21587 Division by zero causes a floating-point exception.
21590 Uses integer operations to calculate the inverse of the divisor,
21591 and then multiplies the dividend with the inverse. This strategy allows
21592 CSE and hoisting of the inverse calculation. Division by zero calculates
21593 an unspecified result, but does not trap.
21596 A variant of @samp{inv} where, if no CSE or hoisting opportunities
21597 have been found, or if the entire operation has been hoisted to the same
21598 place, the last stages of the inverse calculation are intertwined with the
21599 final multiply to reduce the overall latency, at the expense of using a few
21600 more instructions, and thus offering fewer scheduling opportunities with
21604 Calls a library function that usually implements the @samp{inv:minlat}
21606 This gives high code density for @code{m5-*media-nofpu} compilations.
21609 Uses a different entry point of the same library function, where it
21610 assumes that a pointer to a lookup table has already been set up, which
21611 exposes the pointer load to CSE and code hoisting optimizations.
21616 Use the @samp{inv} algorithm for initial
21617 code generation, but if the code stays unoptimized, revert to the @samp{call},
21618 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
21619 potentially-trapping side effect of division by zero is carried by a
21620 separate instruction, so it is possible that all the integer instructions
21621 are hoisted out, but the marker for the side effect stays where it is.
21622 A recombination to floating-point operations or a call is not possible
21627 Variants of the @samp{inv:minlat} strategy. In the case
21628 that the inverse calculation is not separated from the multiply, they speed
21629 up division where the dividend fits into 20 bits (plus sign where applicable)
21630 by inserting a test to skip a number of operations in this case; this test
21631 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
21632 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
21636 For targets other than SHmedia @var{strategy} can be one of:
21641 Calls a library function that uses the single-step division instruction
21642 @code{div1} to perform the operation. Division by zero calculates an
21643 unspecified result and does not trap. This is the default except for SH4,
21644 SH2A and SHcompact.
21647 Calls a library function that performs the operation in double precision
21648 floating point. Division by zero causes a floating-point exception. This is
21649 the default for SHcompact with FPU. Specifying this for targets that do not
21650 have a double precision FPU defaults to @code{call-div1}.
21653 Calls a library function that uses a lookup table for small divisors and
21654 the @code{div1} instruction with case distinction for larger divisors. Division
21655 by zero calculates an unspecified result and does not trap. This is the default
21656 for SH4. Specifying this for targets that do not have dynamic shift
21657 instructions defaults to @code{call-div1}.
21661 When a division strategy has not been specified the default strategy is
21662 selected based on the current target. For SH2A the default strategy is to
21663 use the @code{divs} and @code{divu} instructions instead of library function
21666 @item -maccumulate-outgoing-args
21667 @opindex maccumulate-outgoing-args
21668 Reserve space once for outgoing arguments in the function prologue rather
21669 than around each call. Generally beneficial for performance and size. Also
21670 needed for unwinding to avoid changing the stack frame around conditional code.
21672 @item -mdivsi3_libfunc=@var{name}
21673 @opindex mdivsi3_libfunc=@var{name}
21674 Set the name of the library function used for 32-bit signed division to
21676 This only affects the name used in the @samp{call} and @samp{inv:call}
21677 division strategies, and the compiler still expects the same
21678 sets of input/output/clobbered registers as if this option were not present.
21680 @item -mfixed-range=@var{register-range}
21681 @opindex mfixed-range
21682 Generate code treating the given register range as fixed registers.
21683 A fixed register is one that the register allocator can not use. This is
21684 useful when compiling kernel code. A register range is specified as
21685 two registers separated by a dash. Multiple register ranges can be
21686 specified separated by a comma.
21688 @item -mindexed-addressing
21689 @opindex mindexed-addressing
21690 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
21691 This is only safe if the hardware and/or OS implement 32-bit wrap-around
21692 semantics for the indexed addressing mode. The architecture allows the
21693 implementation of processors with 64-bit MMU, which the OS could use to
21694 get 32-bit addressing, but since no current hardware implementation supports
21695 this or any other way to make the indexed addressing mode safe to use in
21696 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
21698 @item -mgettrcost=@var{number}
21699 @opindex mgettrcost=@var{number}
21700 Set the cost assumed for the @code{gettr} instruction to @var{number}.
21701 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
21705 Assume @code{pt*} instructions won't trap. This generally generates
21706 better-scheduled code, but is unsafe on current hardware.
21707 The current architecture
21708 definition says that @code{ptabs} and @code{ptrel} trap when the target
21710 This has the unintentional effect of making it unsafe to schedule these
21711 instructions before a branch, or hoist them out of a loop. For example,
21712 @code{__do_global_ctors}, a part of @file{libgcc}
21713 that runs constructors at program
21714 startup, calls functions in a list which is delimited by @minus{}1. With the
21715 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
21716 That means that all the constructors run a bit more quickly, but when
21717 the loop comes to the end of the list, the program crashes because @code{ptabs}
21718 loads @minus{}1 into a target register.
21720 Since this option is unsafe for any
21721 hardware implementing the current architecture specification, the default
21722 is @option{-mno-pt-fixed}. Unless specified explicitly with
21723 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
21724 this deters register allocation from using target registers for storing
21727 @item -minvalid-symbols
21728 @opindex minvalid-symbols
21729 Assume symbols might be invalid. Ordinary function symbols generated by
21730 the compiler are always valid to load with
21731 @code{movi}/@code{shori}/@code{ptabs} or
21732 @code{movi}/@code{shori}/@code{ptrel},
21733 but with assembler and/or linker tricks it is possible
21734 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
21735 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
21736 It prevents cross-basic-block CSE, hoisting and most scheduling
21737 of symbol loads. The default is @option{-mno-invalid-symbols}.
21739 @item -mbranch-cost=@var{num}
21740 @opindex mbranch-cost=@var{num}
21741 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21742 make the compiler try to generate more branch-free code if possible.
21743 If not specified the value is selected depending on the processor type that
21744 is being compiled for.
21747 @itemx -mno-zdcbranch
21748 @opindex mzdcbranch
21749 @opindex mno-zdcbranch
21750 Assume (do not assume) that zero displacement conditional branch instructions
21751 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21752 compiler prefers zero displacement branch code sequences. This is
21753 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21754 disabled by specifying @option{-mno-zdcbranch}.
21757 @itemx -mno-fused-madd
21758 @opindex mfused-madd
21759 @opindex mno-fused-madd
21760 Generate code that uses (does not use) the floating-point multiply and
21761 accumulate instructions. These instructions are generated by default
21762 if hardware floating point is used. The machine-dependent
21763 @option{-mfused-madd} option is now mapped to the machine-independent
21764 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21765 mapped to @option{-ffp-contract=off}.
21771 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21772 and cosine approximations. The option @option{-mfsca} must be used in
21773 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21774 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21775 approximations even if @option{-funsafe-math-optimizations} is in effect.
21781 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21782 reciprocal square root approximations. The option @option{-mfsrra} must be used
21783 in combination with @option{-funsafe-math-optimizations} and
21784 @option{-ffinite-math-only}. It is enabled by default when generating code for
21785 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21786 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21789 @item -mpretend-cmove
21790 @opindex mpretend-cmove
21791 Prefer zero-displacement conditional branches for conditional move instruction
21792 patterns. This can result in faster code on the SH4 processor.
21796 @node Solaris 2 Options
21797 @subsection Solaris 2 Options
21798 @cindex Solaris 2 options
21800 These @samp{-m} options are supported on Solaris 2:
21803 @item -mclear-hwcap
21804 @opindex mclear-hwcap
21805 @option{-mclear-hwcap} tells the compiler to remove the hardware
21806 capabilities generated by the Solaris assembler. This is only necessary
21807 when object files use ISA extensions not supported by the current
21808 machine, but check at runtime whether or not to use them.
21810 @item -mimpure-text
21811 @opindex mimpure-text
21812 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21813 the compiler to not pass @option{-z text} to the linker when linking a
21814 shared object. Using this option, you can link position-dependent
21815 code into a shared object.
21817 @option{-mimpure-text} suppresses the ``relocations remain against
21818 allocatable but non-writable sections'' linker error message.
21819 However, the necessary relocations trigger copy-on-write, and the
21820 shared object is not actually shared across processes. Instead of
21821 using @option{-mimpure-text}, you should compile all source code with
21822 @option{-fpic} or @option{-fPIC}.
21826 These switches are supported in addition to the above on Solaris 2:
21831 Add support for multithreading using the POSIX threads library. This
21832 option sets flags for both the preprocessor and linker. This option does
21833 not affect the thread safety of object code produced by the compiler or
21834 that of libraries supplied with it.
21838 This is a synonym for @option{-pthreads}.
21841 @node SPARC Options
21842 @subsection SPARC Options
21843 @cindex SPARC options
21845 These @samp{-m} options are supported on the SPARC:
21848 @item -mno-app-regs
21850 @opindex mno-app-regs
21852 Specify @option{-mapp-regs} to generate output using the global registers
21853 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21854 global register 1, each global register 2 through 4 is then treated as an
21855 allocable register that is clobbered by function calls. This is the default.
21857 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21858 specify @option{-mno-app-regs}. You should compile libraries and system
21859 software with this option.
21865 With @option{-mflat}, the compiler does not generate save/restore instructions
21866 and uses a ``flat'' or single register window model. This model is compatible
21867 with the regular register window model. The local registers and the input
21868 registers (0--5) are still treated as ``call-saved'' registers and are
21869 saved on the stack as needed.
21871 With @option{-mno-flat} (the default), the compiler generates save/restore
21872 instructions (except for leaf functions). This is the normal operating mode.
21875 @itemx -mhard-float
21877 @opindex mhard-float
21878 Generate output containing floating-point instructions. This is the
21882 @itemx -msoft-float
21884 @opindex msoft-float
21885 Generate output containing library calls for floating point.
21886 @strong{Warning:} the requisite libraries are not available for all SPARC
21887 targets. Normally the facilities of the machine's usual C compiler are
21888 used, but this cannot be done directly in cross-compilation. You must make
21889 your own arrangements to provide suitable library functions for
21890 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21891 @samp{sparclite-*-*} do provide software floating-point support.
21893 @option{-msoft-float} changes the calling convention in the output file;
21894 therefore, it is only useful if you compile @emph{all} of a program with
21895 this option. In particular, you need to compile @file{libgcc.a}, the
21896 library that comes with GCC, with @option{-msoft-float} in order for
21899 @item -mhard-quad-float
21900 @opindex mhard-quad-float
21901 Generate output containing quad-word (long double) floating-point
21904 @item -msoft-quad-float
21905 @opindex msoft-quad-float
21906 Generate output containing library calls for quad-word (long double)
21907 floating-point instructions. The functions called are those specified
21908 in the SPARC ABI@. This is the default.
21910 As of this writing, there are no SPARC implementations that have hardware
21911 support for the quad-word floating-point instructions. They all invoke
21912 a trap handler for one of these instructions, and then the trap handler
21913 emulates the effect of the instruction. Because of the trap handler overhead,
21914 this is much slower than calling the ABI library routines. Thus the
21915 @option{-msoft-quad-float} option is the default.
21917 @item -mno-unaligned-doubles
21918 @itemx -munaligned-doubles
21919 @opindex mno-unaligned-doubles
21920 @opindex munaligned-doubles
21921 Assume that doubles have 8-byte alignment. This is the default.
21923 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21924 alignment only if they are contained in another type, or if they have an
21925 absolute address. Otherwise, it assumes they have 4-byte alignment.
21926 Specifying this option avoids some rare compatibility problems with code
21927 generated by other compilers. It is not the default because it results
21928 in a performance loss, especially for floating-point code.
21931 @itemx -mno-user-mode
21932 @opindex muser-mode
21933 @opindex mno-user-mode
21934 Do not generate code that can only run in supervisor mode. This is relevant
21935 only for the @code{casa} instruction emitted for the LEON3 processor. The
21936 default is @option{-mno-user-mode}.
21938 @item -mno-faster-structs
21939 @itemx -mfaster-structs
21940 @opindex mno-faster-structs
21941 @opindex mfaster-structs
21942 With @option{-mfaster-structs}, the compiler assumes that structures
21943 should have 8-byte alignment. This enables the use of pairs of
21944 @code{ldd} and @code{std} instructions for copies in structure
21945 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21946 However, the use of this changed alignment directly violates the SPARC
21947 ABI@. Thus, it's intended only for use on targets where the developer
21948 acknowledges that their resulting code is not directly in line with
21949 the rules of the ABI@.
21951 @item -mcpu=@var{cpu_type}
21953 Set the instruction set, register set, and instruction scheduling parameters
21954 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21955 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21956 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21957 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21958 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21959 @samp{niagara3} and @samp{niagara4}.
21961 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21962 which selects the best architecture option for the host processor.
21963 @option{-mcpu=native} has no effect if GCC does not recognize
21966 Default instruction scheduling parameters are used for values that select
21967 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21968 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21970 Here is a list of each supported architecture and their supported
21978 supersparc, hypersparc, leon, leon3
21981 f930, f934, sparclite86x
21987 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21990 By default (unless configured otherwise), GCC generates code for the V7
21991 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21992 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21993 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21994 SPARCStation 1, 2, IPX etc.
21996 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21997 architecture. The only difference from V7 code is that the compiler emits
21998 the integer multiply and integer divide instructions which exist in SPARC-V8
21999 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
22000 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
22003 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
22004 the SPARC architecture. This adds the integer multiply, integer divide step
22005 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
22006 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
22007 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
22008 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
22009 MB86934 chip, which is the more recent SPARClite with FPU@.
22011 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
22012 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
22013 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
22014 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
22015 optimizes it for the TEMIC SPARClet chip.
22017 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
22018 architecture. This adds 64-bit integer and floating-point move instructions,
22019 3 additional floating-point condition code registers and conditional move
22020 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
22021 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
22022 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
22023 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
22024 @option{-mcpu=niagara}, the compiler additionally optimizes it for
22025 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
22026 additionally optimizes it for Sun UltraSPARC T2 chips. With
22027 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
22028 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
22029 additionally optimizes it for Sun UltraSPARC T4 chips.
22031 @item -mtune=@var{cpu_type}
22033 Set the instruction scheduling parameters for machine type
22034 @var{cpu_type}, but do not set the instruction set or register set that the
22035 option @option{-mcpu=@var{cpu_type}} does.
22037 The same values for @option{-mcpu=@var{cpu_type}} can be used for
22038 @option{-mtune=@var{cpu_type}}, but the only useful values are those
22039 that select a particular CPU implementation. Those are @samp{cypress},
22040 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
22041 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
22042 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
22043 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
22044 toolchains, @samp{native} can also be used.
22049 @opindex mno-v8plus
22050 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
22051 difference from the V8 ABI is that the global and out registers are
22052 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
22053 mode for all SPARC-V9 processors.
22059 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
22060 Visual Instruction Set extensions. The default is @option{-mno-vis}.
22066 With @option{-mvis2}, GCC generates code that takes advantage of
22067 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
22068 default is @option{-mvis2} when targeting a cpu that supports such
22069 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
22070 also sets @option{-mvis}.
22076 With @option{-mvis3}, GCC generates code that takes advantage of
22077 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
22078 default is @option{-mvis3} when targeting a cpu that supports such
22079 instructions, such as niagara-3 and later. Setting @option{-mvis3}
22080 also sets @option{-mvis2} and @option{-mvis}.
22085 @opindex mno-cbcond
22086 With @option{-mcbcond}, GCC generates code that takes advantage of
22087 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
22088 The default is @option{-mcbcond} when targeting a cpu that supports such
22089 instructions, such as niagara-4 and later.
22095 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
22096 population count instruction. The default is @option{-mpopc}
22097 when targeting a cpu that supports such instructions, such as Niagara-2 and
22104 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
22105 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
22106 when targeting a cpu that supports such instructions, such as Niagara-3 and
22110 @opindex mfix-at697f
22111 Enable the documented workaround for the single erratum of the Atmel AT697F
22112 processor (which corresponds to erratum #13 of the AT697E processor).
22115 @opindex mfix-ut699
22116 Enable the documented workarounds for the floating-point errata and the data
22117 cache nullify errata of the UT699 processor.
22120 These @samp{-m} options are supported in addition to the above
22121 on SPARC-V9 processors in 64-bit environments:
22128 Generate code for a 32-bit or 64-bit environment.
22129 The 32-bit environment sets int, long and pointer to 32 bits.
22130 The 64-bit environment sets int to 32 bits and long and pointer
22133 @item -mcmodel=@var{which}
22135 Set the code model to one of
22139 The Medium/Low code model: 64-bit addresses, programs
22140 must be linked in the low 32 bits of memory. Programs can be statically
22141 or dynamically linked.
22144 The Medium/Middle code model: 64-bit addresses, programs
22145 must be linked in the low 44 bits of memory, the text and data segments must
22146 be less than 2GB in size and the data segment must be located within 2GB of
22150 The Medium/Anywhere code model: 64-bit addresses, programs
22151 may be linked anywhere in memory, the text and data segments must be less
22152 than 2GB in size and the data segment must be located within 2GB of the
22156 The Medium/Anywhere code model for embedded systems:
22157 64-bit addresses, the text and data segments must be less than 2GB in
22158 size, both starting anywhere in memory (determined at link time). The
22159 global register %g4 points to the base of the data segment. Programs
22160 are statically linked and PIC is not supported.
22163 @item -mmemory-model=@var{mem-model}
22164 @opindex mmemory-model
22165 Set the memory model in force on the processor to one of
22169 The default memory model for the processor and operating system.
22172 Relaxed Memory Order
22175 Partial Store Order
22181 Sequential Consistency
22184 These memory models are formally defined in Appendix D of the Sparc V9
22185 architecture manual, as set in the processor's @code{PSTATE.MM} field.
22188 @itemx -mno-stack-bias
22189 @opindex mstack-bias
22190 @opindex mno-stack-bias
22191 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
22192 frame pointer if present, are offset by @minus{}2047 which must be added back
22193 when making stack frame references. This is the default in 64-bit mode.
22194 Otherwise, assume no such offset is present.
22198 @subsection SPU Options
22199 @cindex SPU options
22201 These @samp{-m} options are supported on the SPU:
22205 @itemx -merror-reloc
22206 @opindex mwarn-reloc
22207 @opindex merror-reloc
22209 The loader for SPU does not handle dynamic relocations. By default, GCC
22210 gives an error when it generates code that requires a dynamic
22211 relocation. @option{-mno-error-reloc} disables the error,
22212 @option{-mwarn-reloc} generates a warning instead.
22215 @itemx -munsafe-dma
22217 @opindex munsafe-dma
22219 Instructions that initiate or test completion of DMA must not be
22220 reordered with respect to loads and stores of the memory that is being
22222 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
22223 memory accesses, but that can lead to inefficient code in places where the
22224 memory is known to not change. Rather than mark the memory as volatile,
22225 you can use @option{-msafe-dma} to tell the compiler to treat
22226 the DMA instructions as potentially affecting all memory.
22228 @item -mbranch-hints
22229 @opindex mbranch-hints
22231 By default, GCC generates a branch hint instruction to avoid
22232 pipeline stalls for always-taken or probably-taken branches. A hint
22233 is not generated closer than 8 instructions away from its branch.
22234 There is little reason to disable them, except for debugging purposes,
22235 or to make an object a little bit smaller.
22239 @opindex msmall-mem
22240 @opindex mlarge-mem
22242 By default, GCC generates code assuming that addresses are never larger
22243 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
22244 a full 32-bit address.
22249 By default, GCC links against startup code that assumes the SPU-style
22250 main function interface (which has an unconventional parameter list).
22251 With @option{-mstdmain}, GCC links your program against startup
22252 code that assumes a C99-style interface to @code{main}, including a
22253 local copy of @code{argv} strings.
22255 @item -mfixed-range=@var{register-range}
22256 @opindex mfixed-range
22257 Generate code treating the given register range as fixed registers.
22258 A fixed register is one that the register allocator cannot use. This is
22259 useful when compiling kernel code. A register range is specified as
22260 two registers separated by a dash. Multiple register ranges can be
22261 specified separated by a comma.
22267 Compile code assuming that pointers to the PPU address space accessed
22268 via the @code{__ea} named address space qualifier are either 32 or 64
22269 bits wide. The default is 32 bits. As this is an ABI-changing option,
22270 all object code in an executable must be compiled with the same setting.
22272 @item -maddress-space-conversion
22273 @itemx -mno-address-space-conversion
22274 @opindex maddress-space-conversion
22275 @opindex mno-address-space-conversion
22276 Allow/disallow treating the @code{__ea} address space as superset
22277 of the generic address space. This enables explicit type casts
22278 between @code{__ea} and generic pointer as well as implicit
22279 conversions of generic pointers to @code{__ea} pointers. The
22280 default is to allow address space pointer conversions.
22282 @item -mcache-size=@var{cache-size}
22283 @opindex mcache-size
22284 This option controls the version of libgcc that the compiler links to an
22285 executable and selects a software-managed cache for accessing variables
22286 in the @code{__ea} address space with a particular cache size. Possible
22287 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
22288 and @samp{128}. The default cache size is 64KB.
22290 @item -matomic-updates
22291 @itemx -mno-atomic-updates
22292 @opindex matomic-updates
22293 @opindex mno-atomic-updates
22294 This option controls the version of libgcc that the compiler links to an
22295 executable and selects whether atomic updates to the software-managed
22296 cache of PPU-side variables are used. If you use atomic updates, changes
22297 to a PPU variable from SPU code using the @code{__ea} named address space
22298 qualifier do not interfere with changes to other PPU variables residing
22299 in the same cache line from PPU code. If you do not use atomic updates,
22300 such interference may occur; however, writing back cache lines is
22301 more efficient. The default behavior is to use atomic updates.
22304 @itemx -mdual-nops=@var{n}
22305 @opindex mdual-nops
22306 By default, GCC inserts nops to increase dual issue when it expects
22307 it to increase performance. @var{n} can be a value from 0 to 10. A
22308 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
22309 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
22311 @item -mhint-max-nops=@var{n}
22312 @opindex mhint-max-nops
22313 Maximum number of nops to insert for a branch hint. A branch hint must
22314 be at least 8 instructions away from the branch it is affecting. GCC
22315 inserts up to @var{n} nops to enforce this, otherwise it does not
22316 generate the branch hint.
22318 @item -mhint-max-distance=@var{n}
22319 @opindex mhint-max-distance
22320 The encoding of the branch hint instruction limits the hint to be within
22321 256 instructions of the branch it is affecting. By default, GCC makes
22322 sure it is within 125.
22325 @opindex msafe-hints
22326 Work around a hardware bug that causes the SPU to stall indefinitely.
22327 By default, GCC inserts the @code{hbrp} instruction to make sure
22328 this stall won't happen.
22332 @node System V Options
22333 @subsection Options for System V
22335 These additional options are available on System V Release 4 for
22336 compatibility with other compilers on those systems:
22341 Create a shared object.
22342 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22346 Identify the versions of each tool used by the compiler, in a
22347 @code{.ident} assembler directive in the output.
22351 Refrain from adding @code{.ident} directives to the output file (this is
22354 @item -YP,@var{dirs}
22356 Search the directories @var{dirs}, and no others, for libraries
22357 specified with @option{-l}.
22359 @item -Ym,@var{dir}
22361 Look in the directory @var{dir} to find the M4 preprocessor.
22362 The assembler uses this option.
22363 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22364 @c the generic assembler that comes with Solaris takes just -Ym.
22367 @node TILE-Gx Options
22368 @subsection TILE-Gx Options
22369 @cindex TILE-Gx options
22371 These @samp{-m} options are supported on the TILE-Gx:
22374 @item -mcmodel=small
22375 @opindex mcmodel=small
22376 Generate code for the small model. The distance for direct calls is
22377 limited to 500M in either direction. PC-relative addresses are 32
22378 bits. Absolute addresses support the full address range.
22380 @item -mcmodel=large
22381 @opindex mcmodel=large
22382 Generate code for the large model. There is no limitation on call
22383 distance, pc-relative addresses, or absolute addresses.
22385 @item -mcpu=@var{name}
22387 Selects the type of CPU to be targeted. Currently the only supported
22388 type is @samp{tilegx}.
22394 Generate code for a 32-bit or 64-bit environment. The 32-bit
22395 environment sets int, long, and pointer to 32 bits. The 64-bit
22396 environment sets int to 32 bits and long and pointer to 64 bits.
22399 @itemx -mlittle-endian
22400 @opindex mbig-endian
22401 @opindex mlittle-endian
22402 Generate code in big/little endian mode, respectively.
22405 @node TILEPro Options
22406 @subsection TILEPro Options
22407 @cindex TILEPro options
22409 These @samp{-m} options are supported on the TILEPro:
22412 @item -mcpu=@var{name}
22414 Selects the type of CPU to be targeted. Currently the only supported
22415 type is @samp{tilepro}.
22419 Generate code for a 32-bit environment, which sets int, long, and
22420 pointer to 32 bits. This is the only supported behavior so the flag
22421 is essentially ignored.
22425 @subsection V850 Options
22426 @cindex V850 Options
22428 These @samp{-m} options are defined for V850 implementations:
22432 @itemx -mno-long-calls
22433 @opindex mlong-calls
22434 @opindex mno-long-calls
22435 Treat all calls as being far away (near). If calls are assumed to be
22436 far away, the compiler always loads the function's address into a
22437 register, and calls indirect through the pointer.
22443 Do not optimize (do optimize) basic blocks that use the same index
22444 pointer 4 or more times to copy pointer into the @code{ep} register, and
22445 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22446 option is on by default if you optimize.
22448 @item -mno-prolog-function
22449 @itemx -mprolog-function
22450 @opindex mno-prolog-function
22451 @opindex mprolog-function
22452 Do not use (do use) external functions to save and restore registers
22453 at the prologue and epilogue of a function. The external functions
22454 are slower, but use less code space if more than one function saves
22455 the same number of registers. The @option{-mprolog-function} option
22456 is on by default if you optimize.
22460 Try to make the code as small as possible. At present, this just turns
22461 on the @option{-mep} and @option{-mprolog-function} options.
22463 @item -mtda=@var{n}
22465 Put static or global variables whose size is @var{n} bytes or less into
22466 the tiny data area that register @code{ep} points to. The tiny data
22467 area can hold up to 256 bytes in total (128 bytes for byte references).
22469 @item -msda=@var{n}
22471 Put static or global variables whose size is @var{n} bytes or less into
22472 the small data area that register @code{gp} points to. The small data
22473 area can hold up to 64 kilobytes.
22475 @item -mzda=@var{n}
22477 Put static or global variables whose size is @var{n} bytes or less into
22478 the first 32 kilobytes of memory.
22482 Specify that the target processor is the V850.
22486 Specify that the target processor is the V850E3V5. The preprocessor
22487 constant @code{__v850e3v5__} is defined if this option is used.
22491 Specify that the target processor is the V850E3V5. This is an alias for
22492 the @option{-mv850e3v5} option.
22496 Specify that the target processor is the V850E2V3. The preprocessor
22497 constant @code{__v850e2v3__} is defined if this option is used.
22501 Specify that the target processor is the V850E2. The preprocessor
22502 constant @code{__v850e2__} is defined if this option is used.
22506 Specify that the target processor is the V850E1. The preprocessor
22507 constants @code{__v850e1__} and @code{__v850e__} are defined if
22508 this option is used.
22512 Specify that the target processor is the V850ES. This is an alias for
22513 the @option{-mv850e1} option.
22517 Specify that the target processor is the V850E@. The preprocessor
22518 constant @code{__v850e__} is defined if this option is used.
22520 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22521 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22522 are defined then a default target processor is chosen and the
22523 relevant @samp{__v850*__} preprocessor constant is defined.
22525 The preprocessor constants @code{__v850} and @code{__v851__} are always
22526 defined, regardless of which processor variant is the target.
22528 @item -mdisable-callt
22529 @itemx -mno-disable-callt
22530 @opindex mdisable-callt
22531 @opindex mno-disable-callt
22532 This option suppresses generation of the @code{CALLT} instruction for the
22533 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22536 This option is enabled by default when the RH850 ABI is
22537 in use (see @option{-mrh850-abi}), and disabled by default when the
22538 GCC ABI is in use. If @code{CALLT} instructions are being generated
22539 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
22545 Pass on (or do not pass on) the @option{-mrelax} command line option
22549 @itemx -mno-long-jumps
22550 @opindex mlong-jumps
22551 @opindex mno-long-jumps
22552 Disable (or re-enable) the generation of PC-relative jump instructions.
22555 @itemx -mhard-float
22556 @opindex msoft-float
22557 @opindex mhard-float
22558 Disable (or re-enable) the generation of hardware floating point
22559 instructions. This option is only significant when the target
22560 architecture is @samp{V850E2V3} or higher. If hardware floating point
22561 instructions are being generated then the C preprocessor symbol
22562 @code{__FPU_OK__} is defined, otherwise the symbol
22563 @code{__NO_FPU__} is defined.
22567 Enables the use of the e3v5 LOOP instruction. The use of this
22568 instruction is not enabled by default when the e3v5 architecture is
22569 selected because its use is still experimental.
22573 @opindex mrh850-abi
22575 Enables support for the RH850 version of the V850 ABI. This is the
22576 default. With this version of the ABI the following rules apply:
22580 Integer sized structures and unions are returned via a memory pointer
22581 rather than a register.
22584 Large structures and unions (more than 8 bytes in size) are passed by
22588 Functions are aligned to 16-bit boundaries.
22591 The @option{-m8byte-align} command line option is supported.
22594 The @option{-mdisable-callt} command line option is enabled by
22595 default. The @option{-mno-disable-callt} command line option is not
22599 When this version of the ABI is enabled the C preprocessor symbol
22600 @code{__V850_RH850_ABI__} is defined.
22604 Enables support for the old GCC version of the V850 ABI. With this
22605 version of the ABI the following rules apply:
22609 Integer sized structures and unions are returned in register @code{r10}.
22612 Large structures and unions (more than 8 bytes in size) are passed by
22616 Functions are aligned to 32-bit boundaries, unless optimizing for
22620 The @option{-m8byte-align} command line option is not supported.
22623 The @option{-mdisable-callt} command line option is supported but not
22624 enabled by default.
22627 When this version of the ABI is enabled the C preprocessor symbol
22628 @code{__V850_GCC_ABI__} is defined.
22630 @item -m8byte-align
22631 @itemx -mno-8byte-align
22632 @opindex m8byte-align
22633 @opindex mno-8byte-align
22634 Enables support for @code{double} and @code{long long} types to be
22635 aligned on 8-byte boundaries. The default is to restrict the
22636 alignment of all objects to at most 4-bytes. When
22637 @option{-m8byte-align} is in effect the C preprocessor symbol
22638 @code{__V850_8BYTE_ALIGN__} is defined.
22641 @opindex mbig-switch
22642 Generate code suitable for big switch tables. Use this option only if
22643 the assembler/linker complain about out of range branches within a switch
22648 This option causes r2 and r5 to be used in the code generated by
22649 the compiler. This setting is the default.
22651 @item -mno-app-regs
22652 @opindex mno-app-regs
22653 This option causes r2 and r5 to be treated as fixed registers.
22658 @subsection VAX Options
22659 @cindex VAX options
22661 These @samp{-m} options are defined for the VAX:
22666 Do not output certain jump instructions (@code{aobleq} and so on)
22667 that the Unix assembler for the VAX cannot handle across long
22672 Do output those jump instructions, on the assumption that the
22673 GNU assembler is being used.
22677 Output code for G-format floating-point numbers instead of D-format.
22680 @node Visium Options
22681 @subsection Visium Options
22682 @cindex Visium options
22688 A program which performs file I/O and is destined to run on an MCM target
22689 should be linked with this option. It causes the libraries libc.a and
22690 libdebug.a to be linked. The program should be run on the target under
22691 the control of the GDB remote debugging stub.
22695 A program which performs file I/O and is destined to run on the simulator
22696 should be linked with option. This causes libraries libc.a and libsim.a to
22700 @itemx -mhard-float
22702 @opindex mhard-float
22703 Generate code containing floating-point instructions. This is the
22707 @itemx -msoft-float
22709 @opindex msoft-float
22710 Generate code containing library calls for floating-point.
22712 @option{-msoft-float} changes the calling convention in the output file;
22713 therefore, it is only useful if you compile @emph{all} of a program with
22714 this option. In particular, you need to compile @file{libgcc.a}, the
22715 library that comes with GCC, with @option{-msoft-float} in order for
22718 @item -mcpu=@var{cpu_type}
22720 Set the instruction set, register set, and instruction scheduling parameters
22721 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22722 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22724 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22726 By default (unless configured otherwise), GCC generates code for the GR5
22727 variant of the Visium architecture.
22729 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22730 architecture. The only difference from GR5 code is that the compiler will
22731 generate block move instructions.
22733 @item -mtune=@var{cpu_type}
22735 Set the instruction scheduling parameters for machine type @var{cpu_type},
22736 but do not set the instruction set or register set that the option
22737 @option{-mcpu=@var{cpu_type}} would.
22741 Generate code for the supervisor mode, where there are no restrictions on
22742 the access to general registers. This is the default.
22745 @opindex muser-mode
22746 Generate code for the user mode, where the access to some general registers
22747 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22748 mode; on the GR6, only registers r29 to r31 are affected.
22752 @subsection VMS Options
22754 These @samp{-m} options are defined for the VMS implementations:
22757 @item -mvms-return-codes
22758 @opindex mvms-return-codes
22759 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22760 condition (e.g.@ error) codes.
22762 @item -mdebug-main=@var{prefix}
22763 @opindex mdebug-main=@var{prefix}
22764 Flag the first routine whose name starts with @var{prefix} as the main
22765 routine for the debugger.
22769 Default to 64-bit memory allocation routines.
22771 @item -mpointer-size=@var{size}
22772 @opindex mpointer-size=@var{size}
22773 Set the default size of pointers. Possible options for @var{size} are
22774 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22775 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22776 The later option disables @code{pragma pointer_size}.
22779 @node VxWorks Options
22780 @subsection VxWorks Options
22781 @cindex VxWorks Options
22783 The options in this section are defined for all VxWorks targets.
22784 Options specific to the target hardware are listed with the other
22785 options for that target.
22790 GCC can generate code for both VxWorks kernels and real time processes
22791 (RTPs). This option switches from the former to the latter. It also
22792 defines the preprocessor macro @code{__RTP__}.
22795 @opindex non-static
22796 Link an RTP executable against shared libraries rather than static
22797 libraries. The options @option{-static} and @option{-shared} can
22798 also be used for RTPs (@pxref{Link Options}); @option{-static}
22805 These options are passed down to the linker. They are defined for
22806 compatibility with Diab.
22809 @opindex Xbind-lazy
22810 Enable lazy binding of function calls. This option is equivalent to
22811 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22815 Disable lazy binding of function calls. This option is the default and
22816 is defined for compatibility with Diab.
22819 @node x86-64 Options
22820 @subsection x86-64 Options
22821 @cindex x86-64 options
22823 These are listed under @xref{i386 and x86-64 Options}.
22825 @node Xstormy16 Options
22826 @subsection Xstormy16 Options
22827 @cindex Xstormy16 Options
22829 These options are defined for Xstormy16:
22834 Choose startup files and linker script suitable for the simulator.
22837 @node Xtensa Options
22838 @subsection Xtensa Options
22839 @cindex Xtensa Options
22841 These options are supported for Xtensa targets:
22845 @itemx -mno-const16
22847 @opindex mno-const16
22848 Enable or disable use of @code{CONST16} instructions for loading
22849 constant values. The @code{CONST16} instruction is currently not a
22850 standard option from Tensilica. When enabled, @code{CONST16}
22851 instructions are always used in place of the standard @code{L32R}
22852 instructions. The use of @code{CONST16} is enabled by default only if
22853 the @code{L32R} instruction is not available.
22856 @itemx -mno-fused-madd
22857 @opindex mfused-madd
22858 @opindex mno-fused-madd
22859 Enable or disable use of fused multiply/add and multiply/subtract
22860 instructions in the floating-point option. This has no effect if the
22861 floating-point option is not also enabled. Disabling fused multiply/add
22862 and multiply/subtract instructions forces the compiler to use separate
22863 instructions for the multiply and add/subtract operations. This may be
22864 desirable in some cases where strict IEEE 754-compliant results are
22865 required: the fused multiply add/subtract instructions do not round the
22866 intermediate result, thereby producing results with @emph{more} bits of
22867 precision than specified by the IEEE standard. Disabling fused multiply
22868 add/subtract instructions also ensures that the program output is not
22869 sensitive to the compiler's ability to combine multiply and add/subtract
22872 @item -mserialize-volatile
22873 @itemx -mno-serialize-volatile
22874 @opindex mserialize-volatile
22875 @opindex mno-serialize-volatile
22876 When this option is enabled, GCC inserts @code{MEMW} instructions before
22877 @code{volatile} memory references to guarantee sequential consistency.
22878 The default is @option{-mserialize-volatile}. Use
22879 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
22881 @item -mforce-no-pic
22882 @opindex mforce-no-pic
22883 For targets, like GNU/Linux, where all user-mode Xtensa code must be
22884 position-independent code (PIC), this option disables PIC for compiling
22887 @item -mtext-section-literals
22888 @itemx -mno-text-section-literals
22889 @opindex mtext-section-literals
22890 @opindex mno-text-section-literals
22891 These options control the treatment of literal pools. The default is
22892 @option{-mno-text-section-literals}, which places literals in a separate
22893 section in the output file. This allows the literal pool to be placed
22894 in a data RAM/ROM, and it also allows the linker to combine literal
22895 pools from separate object files to remove redundant literals and
22896 improve code size. With @option{-mtext-section-literals}, the literals
22897 are interspersed in the text section in order to keep them as close as
22898 possible to their references. This may be necessary for large assembly
22901 @item -mtarget-align
22902 @itemx -mno-target-align
22903 @opindex mtarget-align
22904 @opindex mno-target-align
22905 When this option is enabled, GCC instructs the assembler to
22906 automatically align instructions to reduce branch penalties at the
22907 expense of some code density. The assembler attempts to widen density
22908 instructions to align branch targets and the instructions following call
22909 instructions. If there are not enough preceding safe density
22910 instructions to align a target, no widening is performed. The
22911 default is @option{-mtarget-align}. These options do not affect the
22912 treatment of auto-aligned instructions like @code{LOOP}, which the
22913 assembler always aligns, either by widening density instructions or
22914 by inserting NOP instructions.
22917 @itemx -mno-longcalls
22918 @opindex mlongcalls
22919 @opindex mno-longcalls
22920 When this option is enabled, GCC instructs the assembler to translate
22921 direct calls to indirect calls unless it can determine that the target
22922 of a direct call is in the range allowed by the call instruction. This
22923 translation typically occurs for calls to functions in other source
22924 files. Specifically, the assembler translates a direct @code{CALL}
22925 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
22926 The default is @option{-mno-longcalls}. This option should be used in
22927 programs where the call target can potentially be out of range. This
22928 option is implemented in the assembler, not the compiler, so the
22929 assembly code generated by GCC still shows direct call
22930 instructions---look at the disassembled object code to see the actual
22931 instructions. Note that the assembler uses an indirect call for
22932 every cross-file call, not just those that really are out of range.
22935 @node zSeries Options
22936 @subsection zSeries Options
22937 @cindex zSeries options
22939 These are listed under @xref{S/390 and zSeries Options}.
22941 @node Code Gen Options
22942 @section Options for Code Generation Conventions
22943 @cindex code generation conventions
22944 @cindex options, code generation
22945 @cindex run-time options
22947 These machine-independent options control the interface conventions
22948 used in code generation.
22950 Most of them have both positive and negative forms; the negative form
22951 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
22952 one of the forms is listed---the one that is not the default. You
22953 can figure out the other form by either removing @samp{no-} or adding
22957 @item -fbounds-check
22958 @opindex fbounds-check
22959 For front ends that support it, generate additional code to check that
22960 indices used to access arrays are within the declared range. This is
22961 currently only supported by the Java and Fortran front ends, where
22962 this option defaults to true and false respectively.
22964 @item -fstack-reuse=@var{reuse-level}
22965 @opindex fstack_reuse
22966 This option controls stack space reuse for user declared local/auto variables
22967 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
22968 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
22969 local variables and temporaries, @samp{named_vars} enables the reuse only for
22970 user defined local variables with names, and @samp{none} disables stack reuse
22971 completely. The default value is @samp{all}. The option is needed when the
22972 program extends the lifetime of a scoped local variable or a compiler generated
22973 temporary beyond the end point defined by the language. When a lifetime of
22974 a variable ends, and if the variable lives in memory, the optimizing compiler
22975 has the freedom to reuse its stack space with other temporaries or scoped
22976 local variables whose live range does not overlap with it. Legacy code extending
22977 local lifetime is likely to break with the stack reuse optimization.
22996 if (*p == 10) // out of scope use of local1
23007 A(int k) : i(k), j(k) @{ @}
23014 void foo(const A& ar)
23021 foo(A(10)); // temp object's lifetime ends when foo returns
23027 ap->i+= 10; // ap references out of scope temp whose space
23028 // is reused with a. What is the value of ap->i?
23033 The lifetime of a compiler generated temporary is well defined by the C++
23034 standard. When a lifetime of a temporary ends, and if the temporary lives
23035 in memory, the optimizing compiler has the freedom to reuse its stack
23036 space with other temporaries or scoped local variables whose live range
23037 does not overlap with it. However some of the legacy code relies on
23038 the behavior of older compilers in which temporaries' stack space is
23039 not reused, the aggressive stack reuse can lead to runtime errors. This
23040 option is used to control the temporary stack reuse optimization.
23044 This option generates traps for signed overflow on addition, subtraction,
23045 multiplication operations.
23049 This option instructs the compiler to assume that signed arithmetic
23050 overflow of addition, subtraction and multiplication wraps around
23051 using twos-complement representation. This flag enables some optimizations
23052 and disables others. This option is enabled by default for the Java
23053 front end, as required by the Java language specification.
23056 @opindex fexceptions
23057 Enable exception handling. Generates extra code needed to propagate
23058 exceptions. For some targets, this implies GCC generates frame
23059 unwind information for all functions, which can produce significant data
23060 size overhead, although it does not affect execution. If you do not
23061 specify this option, GCC enables it by default for languages like
23062 C++ that normally require exception handling, and disables it for
23063 languages like C that do not normally require it. However, you may need
23064 to enable this option when compiling C code that needs to interoperate
23065 properly with exception handlers written in C++. You may also wish to
23066 disable this option if you are compiling older C++ programs that don't
23067 use exception handling.
23069 @item -fnon-call-exceptions
23070 @opindex fnon-call-exceptions
23071 Generate code that allows trapping instructions to throw exceptions.
23072 Note that this requires platform-specific runtime support that does
23073 not exist everywhere. Moreover, it only allows @emph{trapping}
23074 instructions to throw exceptions, i.e.@: memory references or floating-point
23075 instructions. It does not allow exceptions to be thrown from
23076 arbitrary signal handlers such as @code{SIGALRM}.
23078 @item -fdelete-dead-exceptions
23079 @opindex fdelete-dead-exceptions
23080 Consider that instructions that may throw exceptions but don't otherwise
23081 contribute to the execution of the program can be optimized away.
23082 This option is enabled by default for the Ada front end, as permitted by
23083 the Ada language specification.
23084 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23086 @item -funwind-tables
23087 @opindex funwind-tables
23088 Similar to @option{-fexceptions}, except that it just generates any needed
23089 static data, but does not affect the generated code in any other way.
23090 You normally do not need to enable this option; instead, a language processor
23091 that needs this handling enables it on your behalf.
23093 @item -fasynchronous-unwind-tables
23094 @opindex fasynchronous-unwind-tables
23095 Generate unwind table in DWARF 2 format, if supported by target machine. The
23096 table is exact at each instruction boundary, so it can be used for stack
23097 unwinding from asynchronous events (such as debugger or garbage collector).
23099 @item -fno-gnu-unique
23100 @opindex fno-gnu-unique
23101 On systems with recent GNU assembler and C library, the C++ compiler
23102 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23103 of template static data members and static local variables in inline
23104 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23105 is necessary to avoid problems with a library used by two different
23106 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23107 therefore disagreeing with the other one about the binding of the
23108 symbol. But this causes @code{dlclose} to be ignored for affected
23109 DSOs; if your program relies on reinitialization of a DSO via
23110 @code{dlclose} and @code{dlopen}, you can use
23111 @option{-fno-gnu-unique}.
23113 @item -fpcc-struct-return
23114 @opindex fpcc-struct-return
23115 Return ``short'' @code{struct} and @code{union} values in memory like
23116 longer ones, rather than in registers. This convention is less
23117 efficient, but it has the advantage of allowing intercallability between
23118 GCC-compiled files and files compiled with other compilers, particularly
23119 the Portable C Compiler (pcc).
23121 The precise convention for returning structures in memory depends
23122 on the target configuration macros.
23124 Short structures and unions are those whose size and alignment match
23125 that of some integer type.
23127 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23128 switch is not binary compatible with code compiled with the
23129 @option{-freg-struct-return} switch.
23130 Use it to conform to a non-default application binary interface.
23132 @item -freg-struct-return
23133 @opindex freg-struct-return
23134 Return @code{struct} and @code{union} values in registers when possible.
23135 This is more efficient for small structures than
23136 @option{-fpcc-struct-return}.
23138 If you specify neither @option{-fpcc-struct-return} nor
23139 @option{-freg-struct-return}, GCC defaults to whichever convention is
23140 standard for the target. If there is no standard convention, GCC
23141 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23142 the principal compiler. In those cases, we can choose the standard, and
23143 we chose the more efficient register return alternative.
23145 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23146 switch is not binary compatible with code compiled with the
23147 @option{-fpcc-struct-return} switch.
23148 Use it to conform to a non-default application binary interface.
23150 @item -fshort-enums
23151 @opindex fshort-enums
23152 Allocate to an @code{enum} type only as many bytes as it needs for the
23153 declared range of possible values. Specifically, the @code{enum} type
23154 is equivalent to the smallest integer type that has enough room.
23156 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23157 code that is not binary compatible with code generated without that switch.
23158 Use it to conform to a non-default application binary interface.
23160 @item -fshort-double
23161 @opindex fshort-double
23162 Use the same size for @code{double} as for @code{float}.
23164 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23165 code that is not binary compatible with code generated without that switch.
23166 Use it to conform to a non-default application binary interface.
23168 @item -fshort-wchar
23169 @opindex fshort-wchar
23170 Override the underlying type for @code{wchar_t} to be @code{short
23171 unsigned int} instead of the default for the target. This option is
23172 useful for building programs to run under WINE@.
23174 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23175 code that is not binary compatible with code generated without that switch.
23176 Use it to conform to a non-default application binary interface.
23179 @opindex fno-common
23180 In C code, controls the placement of uninitialized global variables.
23181 Unix C compilers have traditionally permitted multiple definitions of
23182 such variables in different compilation units by placing the variables
23184 This is the behavior specified by @option{-fcommon}, and is the default
23185 for GCC on most targets.
23186 On the other hand, this behavior is not required by ISO C, and on some
23187 targets may carry a speed or code size penalty on variable references.
23188 The @option{-fno-common} option specifies that the compiler should place
23189 uninitialized global variables in the data section of the object file,
23190 rather than generating them as common blocks.
23191 This has the effect that if the same variable is declared
23192 (without @code{extern}) in two different compilations,
23193 you get a multiple-definition error when you link them.
23194 In this case, you must compile with @option{-fcommon} instead.
23195 Compiling with @option{-fno-common} is useful on targets for which
23196 it provides better performance, or if you wish to verify that the
23197 program will work on other systems that always treat uninitialized
23198 variable declarations this way.
23202 Ignore the @code{#ident} directive.
23204 @item -finhibit-size-directive
23205 @opindex finhibit-size-directive
23206 Don't output a @code{.size} assembler directive, or anything else that
23207 would cause trouble if the function is split in the middle, and the
23208 two halves are placed at locations far apart in memory. This option is
23209 used when compiling @file{crtstuff.c}; you should not need to use it
23212 @item -fverbose-asm
23213 @opindex fverbose-asm
23214 Put extra commentary information in the generated assembly code to
23215 make it more readable. This option is generally only of use to those
23216 who actually need to read the generated assembly code (perhaps while
23217 debugging the compiler itself).
23219 @option{-fno-verbose-asm}, the default, causes the
23220 extra information to be omitted and is useful when comparing two assembler
23223 @item -frecord-gcc-switches
23224 @opindex frecord-gcc-switches
23225 This switch causes the command line used to invoke the
23226 compiler to be recorded into the object file that is being created.
23227 This switch is only implemented on some targets and the exact format
23228 of the recording is target and binary file format dependent, but it
23229 usually takes the form of a section containing ASCII text. This
23230 switch is related to the @option{-fverbose-asm} switch, but that
23231 switch only records information in the assembler output file as
23232 comments, so it never reaches the object file.
23233 See also @option{-grecord-gcc-switches} for another
23234 way of storing compiler options into the object file.
23238 @cindex global offset table
23240 Generate position-independent code (PIC) suitable for use in a shared
23241 library, if supported for the target machine. Such code accesses all
23242 constant addresses through a global offset table (GOT)@. The dynamic
23243 loader resolves the GOT entries when the program starts (the dynamic
23244 loader is not part of GCC; it is part of the operating system). If
23245 the GOT size for the linked executable exceeds a machine-specific
23246 maximum size, you get an error message from the linker indicating that
23247 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23248 instead. (These maximums are 8k on the SPARC and 32k
23249 on the m68k and RS/6000. The 386 has no such limit.)
23251 Position-independent code requires special support, and therefore works
23252 only on certain machines. For the 386, GCC supports PIC for System V
23253 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23254 position-independent.
23256 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23261 If supported for the target machine, emit position-independent code,
23262 suitable for dynamic linking and avoiding any limit on the size of the
23263 global offset table. This option makes a difference on the m68k,
23264 PowerPC and SPARC@.
23266 Position-independent code requires special support, and therefore works
23267 only on certain machines.
23269 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23276 These options are similar to @option{-fpic} and @option{-fPIC}, but
23277 generated position independent code can be only linked into executables.
23278 Usually these options are used when @option{-pie} GCC option is
23279 used during linking.
23281 @option{-fpie} and @option{-fPIE} both define the macros
23282 @code{__pie__} and @code{__PIE__}. The macros have the value 1
23283 for @option{-fpie} and 2 for @option{-fPIE}.
23285 @item -fno-jump-tables
23286 @opindex fno-jump-tables
23287 Do not use jump tables for switch statements even where it would be
23288 more efficient than other code generation strategies. This option is
23289 of use in conjunction with @option{-fpic} or @option{-fPIC} for
23290 building code that forms part of a dynamic linker and cannot
23291 reference the address of a jump table. On some targets, jump tables
23292 do not require a GOT and this option is not needed.
23294 @item -ffixed-@var{reg}
23296 Treat the register named @var{reg} as a fixed register; generated code
23297 should never refer to it (except perhaps as a stack pointer, frame
23298 pointer or in some other fixed role).
23300 @var{reg} must be the name of a register. The register names accepted
23301 are machine-specific and are defined in the @code{REGISTER_NAMES}
23302 macro in the machine description macro file.
23304 This flag does not have a negative form, because it specifies a
23307 @item -fcall-used-@var{reg}
23308 @opindex fcall-used
23309 Treat the register named @var{reg} as an allocable register that is
23310 clobbered by function calls. It may be allocated for temporaries or
23311 variables that do not live across a call. Functions compiled this way
23312 do not save and restore the register @var{reg}.
23314 It is an error to use this flag with the frame pointer or stack pointer.
23315 Use of this flag for other registers that have fixed pervasive roles in
23316 the machine's execution model produces disastrous results.
23318 This flag does not have a negative form, because it specifies a
23321 @item -fcall-saved-@var{reg}
23322 @opindex fcall-saved
23323 Treat the register named @var{reg} as an allocable register saved by
23324 functions. It may be allocated even for temporaries or variables that
23325 live across a call. Functions compiled this way save and restore
23326 the register @var{reg} if they use it.
23328 It is an error to use this flag with the frame pointer or stack pointer.
23329 Use of this flag for other registers that have fixed pervasive roles in
23330 the machine's execution model produces disastrous results.
23332 A different sort of disaster results from the use of this flag for
23333 a register in which function values may be returned.
23335 This flag does not have a negative form, because it specifies a
23338 @item -fpack-struct[=@var{n}]
23339 @opindex fpack-struct
23340 Without a value specified, pack all structure members together without
23341 holes. When a value is specified (which must be a small power of two), pack
23342 structure members according to this value, representing the maximum
23343 alignment (that is, objects with default alignment requirements larger than
23344 this are output potentially unaligned at the next fitting location.
23346 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
23347 code that is not binary compatible with code generated without that switch.
23348 Additionally, it makes the code suboptimal.
23349 Use it to conform to a non-default application binary interface.
23351 @item -finstrument-functions
23352 @opindex finstrument-functions
23353 Generate instrumentation calls for entry and exit to functions. Just
23354 after function entry and just before function exit, the following
23355 profiling functions are called with the address of the current
23356 function and its call site. (On some platforms,
23357 @code{__builtin_return_address} does not work beyond the current
23358 function, so the call site information may not be available to the
23359 profiling functions otherwise.)
23362 void __cyg_profile_func_enter (void *this_fn,
23364 void __cyg_profile_func_exit (void *this_fn,
23368 The first argument is the address of the start of the current function,
23369 which may be looked up exactly in the symbol table.
23371 This instrumentation is also done for functions expanded inline in other
23372 functions. The profiling calls indicate where, conceptually, the
23373 inline function is entered and exited. This means that addressable
23374 versions of such functions must be available. If all your uses of a
23375 function are expanded inline, this may mean an additional expansion of
23376 code size. If you use @code{extern inline} in your C code, an
23377 addressable version of such functions must be provided. (This is
23378 normally the case anyway, but if you get lucky and the optimizer always
23379 expands the functions inline, you might have gotten away without
23380 providing static copies.)
23382 A function may be given the attribute @code{no_instrument_function}, in
23383 which case this instrumentation is not done. This can be used, for
23384 example, for the profiling functions listed above, high-priority
23385 interrupt routines, and any functions from which the profiling functions
23386 cannot safely be called (perhaps signal handlers, if the profiling
23387 routines generate output or allocate memory).
23389 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23390 @opindex finstrument-functions-exclude-file-list
23392 Set the list of functions that are excluded from instrumentation (see
23393 the description of @option{-finstrument-functions}). If the file that
23394 contains a function definition matches with one of @var{file}, then
23395 that function is not instrumented. The match is done on substrings:
23396 if the @var{file} parameter is a substring of the file name, it is
23397 considered to be a match.
23402 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
23406 excludes any inline function defined in files whose pathnames
23407 contain @file{/bits/stl} or @file{include/sys}.
23409 If, for some reason, you want to include letter @samp{,} in one of
23410 @var{sym}, write @samp{\,}. For example,
23411 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
23412 (note the single quote surrounding the option).
23414 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
23415 @opindex finstrument-functions-exclude-function-list
23417 This is similar to @option{-finstrument-functions-exclude-file-list},
23418 but this option sets the list of function names to be excluded from
23419 instrumentation. The function name to be matched is its user-visible
23420 name, such as @code{vector<int> blah(const vector<int> &)}, not the
23421 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
23422 match is done on substrings: if the @var{sym} parameter is a substring
23423 of the function name, it is considered to be a match. For C99 and C++
23424 extended identifiers, the function name must be given in UTF-8, not
23425 using universal character names.
23427 @item -fstack-check
23428 @opindex fstack-check
23429 Generate code to verify that you do not go beyond the boundary of the
23430 stack. You should specify this flag if you are running in an
23431 environment with multiple threads, but you only rarely need to specify it in
23432 a single-threaded environment since stack overflow is automatically
23433 detected on nearly all systems if there is only one stack.
23435 Note that this switch does not actually cause checking to be done; the
23436 operating system or the language runtime must do that. The switch causes
23437 generation of code to ensure that they see the stack being extended.
23439 You can additionally specify a string parameter: @samp{no} means no
23440 checking, @samp{generic} means force the use of old-style checking,
23441 @samp{specific} means use the best checking method and is equivalent
23442 to bare @option{-fstack-check}.
23444 Old-style checking is a generic mechanism that requires no specific
23445 target support in the compiler but comes with the following drawbacks:
23449 Modified allocation strategy for large objects: they are always
23450 allocated dynamically if their size exceeds a fixed threshold.
23453 Fixed limit on the size of the static frame of functions: when it is
23454 topped by a particular function, stack checking is not reliable and
23455 a warning is issued by the compiler.
23458 Inefficiency: because of both the modified allocation strategy and the
23459 generic implementation, code performance is hampered.
23462 Note that old-style stack checking is also the fallback method for
23463 @samp{specific} if no target support has been added in the compiler.
23465 @item -fstack-limit-register=@var{reg}
23466 @itemx -fstack-limit-symbol=@var{sym}
23467 @itemx -fno-stack-limit
23468 @opindex fstack-limit-register
23469 @opindex fstack-limit-symbol
23470 @opindex fno-stack-limit
23471 Generate code to ensure that the stack does not grow beyond a certain value,
23472 either the value of a register or the address of a symbol. If a larger
23473 stack is required, a signal is raised at run time. For most targets,
23474 the signal is raised before the stack overruns the boundary, so
23475 it is possible to catch the signal without taking special precautions.
23477 For instance, if the stack starts at absolute address @samp{0x80000000}
23478 and grows downwards, you can use the flags
23479 @option{-fstack-limit-symbol=__stack_limit} and
23480 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
23481 of 128KB@. Note that this may only work with the GNU linker.
23483 @item -fsplit-stack
23484 @opindex fsplit-stack
23485 Generate code to automatically split the stack before it overflows.
23486 The resulting program has a discontiguous stack which can only
23487 overflow if the program is unable to allocate any more memory. This
23488 is most useful when running threaded programs, as it is no longer
23489 necessary to calculate a good stack size to use for each thread. This
23490 is currently only implemented for the i386 and x86_64 back ends running
23493 When code compiled with @option{-fsplit-stack} calls code compiled
23494 without @option{-fsplit-stack}, there may not be much stack space
23495 available for the latter code to run. If compiling all code,
23496 including library code, with @option{-fsplit-stack} is not an option,
23497 then the linker can fix up these calls so that the code compiled
23498 without @option{-fsplit-stack} always has a large stack. Support for
23499 this is implemented in the gold linker in GNU binutils release 2.21
23502 @item -fleading-underscore
23503 @opindex fleading-underscore
23504 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
23505 change the way C symbols are represented in the object file. One use
23506 is to help link with legacy assembly code.
23508 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
23509 generate code that is not binary compatible with code generated without that
23510 switch. Use it to conform to a non-default application binary interface.
23511 Not all targets provide complete support for this switch.
23513 @item -ftls-model=@var{model}
23514 @opindex ftls-model
23515 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
23516 The @var{model} argument should be one of @samp{global-dynamic},
23517 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
23518 Note that the choice is subject to optimization: the compiler may use
23519 a more efficient model for symbols not visible outside of the translation
23520 unit, or if @option{-fpic} is not given on the command line.
23522 The default without @option{-fpic} is @samp{initial-exec}; with
23523 @option{-fpic} the default is @samp{global-dynamic}.
23525 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
23526 @opindex fvisibility
23527 Set the default ELF image symbol visibility to the specified option---all
23528 symbols are marked with this unless overridden within the code.
23529 Using this feature can very substantially improve linking and
23530 load times of shared object libraries, produce more optimized
23531 code, provide near-perfect API export and prevent symbol clashes.
23532 It is @strong{strongly} recommended that you use this in any shared objects
23535 Despite the nomenclature, @samp{default} always means public; i.e.,
23536 available to be linked against from outside the shared object.
23537 @samp{protected} and @samp{internal} are pretty useless in real-world
23538 usage so the only other commonly used option is @samp{hidden}.
23539 The default if @option{-fvisibility} isn't specified is
23540 @samp{default}, i.e., make every
23541 symbol public---this causes the same behavior as previous versions of
23544 A good explanation of the benefits offered by ensuring ELF
23545 symbols have the correct visibility is given by ``How To Write
23546 Shared Libraries'' by Ulrich Drepper (which can be found at
23547 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
23548 solution made possible by this option to marking things hidden when
23549 the default is public is to make the default hidden and mark things
23550 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
23551 and @code{__attribute__ ((visibility("default")))} instead of
23552 @code{__declspec(dllexport)} you get almost identical semantics with
23553 identical syntax. This is a great boon to those working with
23554 cross-platform projects.
23556 For those adding visibility support to existing code, you may find
23557 @code{#pragma GCC visibility} of use. This works by you enclosing
23558 the declarations you wish to set visibility for with (for example)
23559 @code{#pragma GCC visibility push(hidden)} and
23560 @code{#pragma GCC visibility pop}.
23561 Bear in mind that symbol visibility should be viewed @strong{as
23562 part of the API interface contract} and thus all new code should
23563 always specify visibility when it is not the default; i.e., declarations
23564 only for use within the local DSO should @strong{always} be marked explicitly
23565 as hidden as so to avoid PLT indirection overheads---making this
23566 abundantly clear also aids readability and self-documentation of the code.
23567 Note that due to ISO C++ specification requirements, @code{operator new} and
23568 @code{operator delete} must always be of default visibility.
23570 Be aware that headers from outside your project, in particular system
23571 headers and headers from any other library you use, may not be
23572 expecting to be compiled with visibility other than the default. You
23573 may need to explicitly say @code{#pragma GCC visibility push(default)}
23574 before including any such headers.
23576 @code{extern} declarations are not affected by @option{-fvisibility}, so
23577 a lot of code can be recompiled with @option{-fvisibility=hidden} with
23578 no modifications. However, this means that calls to @code{extern}
23579 functions with no explicit visibility use the PLT, so it is more
23580 effective to use @code{__attribute ((visibility))} and/or
23581 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
23582 declarations should be treated as hidden.
23584 Note that @option{-fvisibility} does affect C++ vague linkage
23585 entities. This means that, for instance, an exception class that is
23586 be thrown between DSOs must be explicitly marked with default
23587 visibility so that the @samp{type_info} nodes are unified between
23590 An overview of these techniques, their benefits and how to use them
23591 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
23593 @item -fstrict-volatile-bitfields
23594 @opindex fstrict-volatile-bitfields
23595 This option should be used if accesses to volatile bit-fields (or other
23596 structure fields, although the compiler usually honors those types
23597 anyway) should use a single access of the width of the
23598 field's type, aligned to a natural alignment if possible. For
23599 example, targets with memory-mapped peripheral registers might require
23600 all such accesses to be 16 bits wide; with this flag you can
23601 declare all peripheral bit-fields as @code{unsigned short} (assuming short
23602 is 16 bits on these targets) to force GCC to use 16-bit accesses
23603 instead of, perhaps, a more efficient 32-bit access.
23605 If this option is disabled, the compiler uses the most efficient
23606 instruction. In the previous example, that might be a 32-bit load
23607 instruction, even though that accesses bytes that do not contain
23608 any portion of the bit-field, or memory-mapped registers unrelated to
23609 the one being updated.
23611 In some cases, such as when the @code{packed} attribute is applied to a
23612 structure field, it may not be possible to access the field with a single
23613 read or write that is correctly aligned for the target machine. In this
23614 case GCC falls back to generating multiple accesses rather than code that
23615 will fault or truncate the result at run time.
23617 Note: Due to restrictions of the C/C++11 memory model, write accesses are
23618 not allowed to touch non bit-field members. It is therefore recommended
23619 to define all bits of the field's type as bit-field members.
23621 The default value of this option is determined by the application binary
23622 interface for the target processor.
23624 @item -fsync-libcalls
23625 @opindex fsync-libcalls
23626 This option controls whether any out-of-line instance of the @code{__sync}
23627 family of functions may be used to implement the C++11 @code{__atomic}
23628 family of functions.
23630 The default value of this option is enabled, thus the only useful form
23631 of the option is @option{-fno-sync-libcalls}. This option is used in
23632 the implementation of the @file{libatomic} runtime library.
23638 @node Environment Variables
23639 @section Environment Variables Affecting GCC
23640 @cindex environment variables
23642 @c man begin ENVIRONMENT
23643 This section describes several environment variables that affect how GCC
23644 operates. Some of them work by specifying directories or prefixes to use
23645 when searching for various kinds of files. Some are used to specify other
23646 aspects of the compilation environment.
23648 Note that you can also specify places to search using options such as
23649 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
23650 take precedence over places specified using environment variables, which
23651 in turn take precedence over those specified by the configuration of GCC@.
23652 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
23653 GNU Compiler Collection (GCC) Internals}.
23658 @c @itemx LC_COLLATE
23660 @c @itemx LC_MONETARY
23661 @c @itemx LC_NUMERIC
23666 @c @findex LC_COLLATE
23667 @findex LC_MESSAGES
23668 @c @findex LC_MONETARY
23669 @c @findex LC_NUMERIC
23673 These environment variables control the way that GCC uses
23674 localization information which allows GCC to work with different
23675 national conventions. GCC inspects the locale categories
23676 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
23677 so. These locale categories can be set to any value supported by your
23678 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
23679 Kingdom encoded in UTF-8.
23681 The @env{LC_CTYPE} environment variable specifies character
23682 classification. GCC uses it to determine the character boundaries in
23683 a string; this is needed for some multibyte encodings that contain quote
23684 and escape characters that are otherwise interpreted as a string
23687 The @env{LC_MESSAGES} environment variable specifies the language to
23688 use in diagnostic messages.
23690 If the @env{LC_ALL} environment variable is set, it overrides the value
23691 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
23692 and @env{LC_MESSAGES} default to the value of the @env{LANG}
23693 environment variable. If none of these variables are set, GCC
23694 defaults to traditional C English behavior.
23698 If @env{TMPDIR} is set, it specifies the directory to use for temporary
23699 files. GCC uses temporary files to hold the output of one stage of
23700 compilation which is to be used as input to the next stage: for example,
23701 the output of the preprocessor, which is the input to the compiler
23704 @item GCC_COMPARE_DEBUG
23705 @findex GCC_COMPARE_DEBUG
23706 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
23707 @option{-fcompare-debug} to the compiler driver. See the documentation
23708 of this option for more details.
23710 @item GCC_EXEC_PREFIX
23711 @findex GCC_EXEC_PREFIX
23712 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
23713 names of the subprograms executed by the compiler. No slash is added
23714 when this prefix is combined with the name of a subprogram, but you can
23715 specify a prefix that ends with a slash if you wish.
23717 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
23718 an appropriate prefix to use based on the pathname it is invoked with.
23720 If GCC cannot find the subprogram using the specified prefix, it
23721 tries looking in the usual places for the subprogram.
23723 The default value of @env{GCC_EXEC_PREFIX} is
23724 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
23725 the installed compiler. In many cases @var{prefix} is the value
23726 of @code{prefix} when you ran the @file{configure} script.
23728 Other prefixes specified with @option{-B} take precedence over this prefix.
23730 This prefix is also used for finding files such as @file{crt0.o} that are
23733 In addition, the prefix is used in an unusual way in finding the
23734 directories to search for header files. For each of the standard
23735 directories whose name normally begins with @samp{/usr/local/lib/gcc}
23736 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
23737 replacing that beginning with the specified prefix to produce an
23738 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
23739 @file{foo/bar} just before it searches the standard directory
23740 @file{/usr/local/lib/bar}.
23741 If a standard directory begins with the configured
23742 @var{prefix} then the value of @var{prefix} is replaced by
23743 @env{GCC_EXEC_PREFIX} when looking for header files.
23745 @item COMPILER_PATH
23746 @findex COMPILER_PATH
23747 The value of @env{COMPILER_PATH} is a colon-separated list of
23748 directories, much like @env{PATH}. GCC tries the directories thus
23749 specified when searching for subprograms, if it can't find the
23750 subprograms using @env{GCC_EXEC_PREFIX}.
23753 @findex LIBRARY_PATH
23754 The value of @env{LIBRARY_PATH} is a colon-separated list of
23755 directories, much like @env{PATH}. When configured as a native compiler,
23756 GCC tries the directories thus specified when searching for special
23757 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
23758 using GCC also uses these directories when searching for ordinary
23759 libraries for the @option{-l} option (but directories specified with
23760 @option{-L} come first).
23764 @cindex locale definition
23765 This variable is used to pass locale information to the compiler. One way in
23766 which this information is used is to determine the character set to be used
23767 when character literals, string literals and comments are parsed in C and C++.
23768 When the compiler is configured to allow multibyte characters,
23769 the following values for @env{LANG} are recognized:
23773 Recognize JIS characters.
23775 Recognize SJIS characters.
23777 Recognize EUCJP characters.
23780 If @env{LANG} is not defined, or if it has some other value, then the
23781 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
23782 recognize and translate multibyte characters.
23786 Some additional environment variables affect the behavior of the
23789 @include cppenv.texi
23793 @node Precompiled Headers
23794 @section Using Precompiled Headers
23795 @cindex precompiled headers
23796 @cindex speed of compilation
23798 Often large projects have many header files that are included in every
23799 source file. The time the compiler takes to process these header files
23800 over and over again can account for nearly all of the time required to
23801 build the project. To make builds faster, GCC allows you to
23802 @dfn{precompile} a header file.
23804 To create a precompiled header file, simply compile it as you would any
23805 other file, if necessary using the @option{-x} option to make the driver
23806 treat it as a C or C++ header file. You may want to use a
23807 tool like @command{make} to keep the precompiled header up-to-date when
23808 the headers it contains change.
23810 A precompiled header file is searched for when @code{#include} is
23811 seen in the compilation. As it searches for the included file
23812 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
23813 compiler looks for a precompiled header in each directory just before it
23814 looks for the include file in that directory. The name searched for is
23815 the name specified in the @code{#include} with @samp{.gch} appended. If
23816 the precompiled header file can't be used, it is ignored.
23818 For instance, if you have @code{#include "all.h"}, and you have
23819 @file{all.h.gch} in the same directory as @file{all.h}, then the
23820 precompiled header file is used if possible, and the original
23821 header is used otherwise.
23823 Alternatively, you might decide to put the precompiled header file in a
23824 directory and use @option{-I} to ensure that directory is searched
23825 before (or instead of) the directory containing the original header.
23826 Then, if you want to check that the precompiled header file is always
23827 used, you can put a file of the same name as the original header in this
23828 directory containing an @code{#error} command.
23830 This also works with @option{-include}. So yet another way to use
23831 precompiled headers, good for projects not designed with precompiled
23832 header files in mind, is to simply take most of the header files used by
23833 a project, include them from another header file, precompile that header
23834 file, and @option{-include} the precompiled header. If the header files
23835 have guards against multiple inclusion, they are skipped because
23836 they've already been included (in the precompiled header).
23838 If you need to precompile the same header file for different
23839 languages, targets, or compiler options, you can instead make a
23840 @emph{directory} named like @file{all.h.gch}, and put each precompiled
23841 header in the directory, perhaps using @option{-o}. It doesn't matter
23842 what you call the files in the directory; every precompiled header in
23843 the directory is considered. The first precompiled header
23844 encountered in the directory that is valid for this compilation is
23845 used; they're searched in no particular order.
23847 There are many other possibilities, limited only by your imagination,
23848 good sense, and the constraints of your build system.
23850 A precompiled header file can be used only when these conditions apply:
23854 Only one precompiled header can be used in a particular compilation.
23857 A precompiled header can't be used once the first C token is seen. You
23858 can have preprocessor directives before a precompiled header; you cannot
23859 include a precompiled header from inside another header.
23862 The precompiled header file must be produced for the same language as
23863 the current compilation. You can't use a C precompiled header for a C++
23867 The precompiled header file must have been produced by the same compiler
23868 binary as the current compilation is using.
23871 Any macros defined before the precompiled header is included must
23872 either be defined in the same way as when the precompiled header was
23873 generated, or must not affect the precompiled header, which usually
23874 means that they don't appear in the precompiled header at all.
23876 The @option{-D} option is one way to define a macro before a
23877 precompiled header is included; using a @code{#define} can also do it.
23878 There are also some options that define macros implicitly, like
23879 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
23882 @item If debugging information is output when using the precompiled
23883 header, using @option{-g} or similar, the same kind of debugging information
23884 must have been output when building the precompiled header. However,
23885 a precompiled header built using @option{-g} can be used in a compilation
23886 when no debugging information is being output.
23888 @item The same @option{-m} options must generally be used when building
23889 and using the precompiled header. @xref{Submodel Options},
23890 for any cases where this rule is relaxed.
23892 @item Each of the following options must be the same when building and using
23893 the precompiled header:
23895 @gccoptlist{-fexceptions}
23898 Some other command-line options starting with @option{-f},
23899 @option{-p}, or @option{-O} must be defined in the same way as when
23900 the precompiled header was generated. At present, it's not clear
23901 which options are safe to change and which are not; the safest choice
23902 is to use exactly the same options when generating and using the
23903 precompiled header. The following are known to be safe:
23905 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
23906 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
23907 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
23912 For all of these except the last, the compiler automatically
23913 ignores the precompiled header if the conditions aren't met. If you
23914 find an option combination that doesn't work and doesn't cause the
23915 precompiled header to be ignored, please consider filing a bug report,
23918 If you do use differing options when generating and using the
23919 precompiled header, the actual behavior is a mixture of the
23920 behavior for the options. For instance, if you use @option{-g} to
23921 generate the precompiled header but not when using it, you may or may
23922 not get debugging information for routines in the precompiled header.