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 -fopenacc -fopenmp -fopenmp-simd @gol
172 -fms-extensions -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 -fipa-cp-alignment @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 @gol
432 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
433 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
434 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
435 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
436 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
437 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
438 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
439 -ftree-loop-if-convert-stores -ftree-loop-im @gol
440 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
441 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
442 -ftree-loop-vectorize @gol
443 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
444 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
445 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
446 -ftree-vectorize -ftree-vrp @gol
447 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
448 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
449 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
450 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
451 --param @var{name}=@var{value}
452 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
454 @item Preprocessor Options
455 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
456 @gccoptlist{-A@var{question}=@var{answer} @gol
457 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
458 -C -dD -dI -dM -dN @gol
459 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
460 -idirafter @var{dir} @gol
461 -include @var{file} -imacros @var{file} @gol
462 -iprefix @var{file} -iwithprefix @var{dir} @gol
463 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
464 -imultilib @var{dir} -isysroot @var{dir} @gol
465 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
466 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
467 -remap -trigraphs -undef -U@var{macro} @gol
468 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
470 @item Assembler Option
471 @xref{Assembler Options,,Passing Options to the Assembler}.
472 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
475 @xref{Link Options,,Options for Linking}.
476 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
477 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
478 -s -static -static-libgcc -static-libstdc++ @gol
479 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
480 -shared -shared-libgcc -symbolic @gol
481 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
482 -u @var{symbol} -z @var{keyword}}
484 @item Directory Options
485 @xref{Directory Options,,Options for Directory Search}.
486 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
487 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
488 --sysroot=@var{dir} --no-sysroot-suffix}
490 @item Machine Dependent Options
491 @xref{Submodel Options,,Hardware Models and Configurations}.
492 @c This list is ordered alphanumerically by subsection name.
493 @c Try and put the significant identifier (CPU or system) first,
494 @c so users have a clue at guessing where the ones they want will be.
496 @emph{AArch64 Options}
497 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
498 -mgeneral-regs-only @gol
499 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
501 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
502 -mtls-dialect=desc -mtls-dialect=traditional @gol
503 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
504 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
506 @emph{Adapteva Epiphany Options}
507 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
508 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
509 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
510 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
511 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
512 -msplit-vecmove-early -m1reg-@var{reg}}
515 @gccoptlist{-mbarrel-shifter @gol
516 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
517 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
518 -mea -mno-mpy -mmul32x16 -mmul64 @gol
519 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
520 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
521 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
522 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
523 -mucb-mcount -mvolatile-cache @gol
524 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
525 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
526 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
527 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
528 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
529 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
532 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
533 -mabi=@var{name} @gol
534 -mapcs-stack-check -mno-apcs-stack-check @gol
535 -mapcs-float -mno-apcs-float @gol
536 -mapcs-reentrant -mno-apcs-reentrant @gol
537 -msched-prolog -mno-sched-prolog @gol
538 -mlittle-endian -mbig-endian @gol
539 -mfloat-abi=@var{name} @gol
540 -mfp16-format=@var{name}
541 -mthumb-interwork -mno-thumb-interwork @gol
542 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
543 -mstructure-size-boundary=@var{n} @gol
544 -mabort-on-noreturn @gol
545 -mlong-calls -mno-long-calls @gol
546 -msingle-pic-base -mno-single-pic-base @gol
547 -mpic-register=@var{reg} @gol
548 -mnop-fun-dllimport @gol
549 -mpoke-function-name @gol
551 -mtpcs-frame -mtpcs-leaf-frame @gol
552 -mcaller-super-interworking -mcallee-super-interworking @gol
553 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
554 -mword-relocations @gol
555 -mfix-cortex-m3-ldrd @gol
556 -munaligned-access @gol
557 -mneon-for-64bits @gol
558 -mslow-flash-data @gol
559 -masm-syntax-unified @gol
563 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
564 -mcall-prologues -mint8 -mno-interrupts -mrelax @gol
565 -mstrict-X -mtiny-stack -Waddr-space-convert}
567 @emph{Blackfin Options}
568 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
569 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
570 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
571 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
572 -mno-id-shared-library -mshared-library-id=@var{n} @gol
573 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
574 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
575 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
579 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
580 -msim -msdata=@var{sdata-type}}
583 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
584 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
585 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
586 -mstack-align -mdata-align -mconst-align @gol
587 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
588 -melf -maout -melinux -mlinux -sim -sim2 @gol
589 -mmul-bug-workaround -mno-mul-bug-workaround}
592 @gccoptlist{-mmac @gol
593 -mcr16cplus -mcr16c @gol
594 -msim -mint32 -mbit-ops
595 -mdata-model=@var{model}}
597 @emph{Darwin Options}
598 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
599 -arch_only -bind_at_load -bundle -bundle_loader @gol
600 -client_name -compatibility_version -current_version @gol
602 -dependency-file -dylib_file -dylinker_install_name @gol
603 -dynamic -dynamiclib -exported_symbols_list @gol
604 -filelist -flat_namespace -force_cpusubtype_ALL @gol
605 -force_flat_namespace -headerpad_max_install_names @gol
607 -image_base -init -install_name -keep_private_externs @gol
608 -multi_module -multiply_defined -multiply_defined_unused @gol
609 -noall_load -no_dead_strip_inits_and_terms @gol
610 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
611 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
612 -private_bundle -read_only_relocs -sectalign @gol
613 -sectobjectsymbols -whyload -seg1addr @gol
614 -sectcreate -sectobjectsymbols -sectorder @gol
615 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
616 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
617 -segprot -segs_read_only_addr -segs_read_write_addr @gol
618 -single_module -static -sub_library -sub_umbrella @gol
619 -twolevel_namespace -umbrella -undefined @gol
620 -unexported_symbols_list -weak_reference_mismatches @gol
621 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
622 -mkernel -mone-byte-bool}
624 @emph{DEC Alpha Options}
625 @gccoptlist{-mno-fp-regs -msoft-float @gol
626 -mieee -mieee-with-inexact -mieee-conformant @gol
627 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
628 -mtrap-precision=@var{mode} -mbuild-constants @gol
629 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
630 -mbwx -mmax -mfix -mcix @gol
631 -mfloat-vax -mfloat-ieee @gol
632 -mexplicit-relocs -msmall-data -mlarge-data @gol
633 -msmall-text -mlarge-text @gol
634 -mmemory-latency=@var{time}}
637 @gccoptlist{-msmall-model -mno-lsim}
640 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
641 -mhard-float -msoft-float @gol
642 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
643 -mdouble -mno-double @gol
644 -mmedia -mno-media -mmuladd -mno-muladd @gol
645 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
646 -mlinked-fp -mlong-calls -malign-labels @gol
647 -mlibrary-pic -macc-4 -macc-8 @gol
648 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
649 -moptimize-membar -mno-optimize-membar @gol
650 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
651 -mvliw-branch -mno-vliw-branch @gol
652 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
653 -mno-nested-cond-exec -mtomcat-stats @gol
657 @emph{GNU/Linux Options}
658 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
659 -tno-android-cc -tno-android-ld}
661 @emph{H8/300 Options}
662 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
665 @gccoptlist{-march=@var{architecture-type} @gol
666 -mdisable-fpregs -mdisable-indexing @gol
667 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
668 -mfixed-range=@var{register-range} @gol
669 -mjump-in-delay -mlinker-opt -mlong-calls @gol
670 -mlong-load-store -mno-disable-fpregs @gol
671 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
672 -mno-jump-in-delay -mno-long-load-store @gol
673 -mno-portable-runtime -mno-soft-float @gol
674 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
675 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
676 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
677 -munix=@var{unix-std} -nolibdld -static -threads}
680 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
681 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
682 -mconstant-gp -mauto-pic -mfused-madd @gol
683 -minline-float-divide-min-latency @gol
684 -minline-float-divide-max-throughput @gol
685 -mno-inline-float-divide @gol
686 -minline-int-divide-min-latency @gol
687 -minline-int-divide-max-throughput @gol
688 -mno-inline-int-divide @gol
689 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
690 -mno-inline-sqrt @gol
691 -mdwarf2-asm -mearly-stop-bits @gol
692 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
693 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
694 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
695 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
696 -msched-spec-ldc -msched-spec-control-ldc @gol
697 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
698 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
699 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
700 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
703 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
704 -msign-extend-enabled -muser-enabled}
706 @emph{M32R/D Options}
707 @gccoptlist{-m32r2 -m32rx -m32r @gol
709 -malign-loops -mno-align-loops @gol
710 -missue-rate=@var{number} @gol
711 -mbranch-cost=@var{number} @gol
712 -mmodel=@var{code-size-model-type} @gol
713 -msdata=@var{sdata-type} @gol
714 -mno-flush-func -mflush-func=@var{name} @gol
715 -mno-flush-trap -mflush-trap=@var{number} @gol
719 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
721 @emph{M680x0 Options}
722 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
723 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
724 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
725 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
726 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
727 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
728 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
729 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
733 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
734 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
735 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
736 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
737 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
740 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
741 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
742 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
743 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
746 @emph{MicroBlaze Options}
747 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
748 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
749 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
750 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
751 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
754 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
755 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
756 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
757 -mips16 -mno-mips16 -mflip-mips16 @gol
758 -minterlink-compressed -mno-interlink-compressed @gol
759 -minterlink-mips16 -mno-interlink-mips16 @gol
760 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
761 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
762 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
763 -mno-float -msingle-float -mdouble-float @gol
764 -modd-spreg -mno-odd-spreg @gol
765 -mabs=@var{mode} -mnan=@var{encoding} @gol
766 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
769 -mvirt -mno-virt @gol
771 -mmicromips -mno-micromips @gol
772 -mfpu=@var{fpu-type} @gol
773 -msmartmips -mno-smartmips @gol
774 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
775 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
776 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
777 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
778 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
779 -membedded-data -mno-embedded-data @gol
780 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
781 -mcode-readable=@var{setting} @gol
782 -msplit-addresses -mno-split-addresses @gol
783 -mexplicit-relocs -mno-explicit-relocs @gol
784 -mcheck-zero-division -mno-check-zero-division @gol
785 -mdivide-traps -mdivide-breaks @gol
786 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
787 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
788 -mfix-24k -mno-fix-24k @gol
789 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
790 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
791 -mfix-vr4120 -mno-fix-vr4120 @gol
792 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
793 -mflush-func=@var{func} -mno-flush-func @gol
794 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
795 -mfp-exceptions -mno-fp-exceptions @gol
796 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
797 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
800 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
801 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
802 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
803 -mno-base-addresses -msingle-exit -mno-single-exit}
805 @emph{MN10300 Options}
806 @gccoptlist{-mmult-bug -mno-mult-bug @gol
807 -mno-am33 -mam33 -mam33-2 -mam34 @gol
808 -mtune=@var{cpu-type} @gol
809 -mreturn-pointer-on-d0 @gol
810 -mno-crt0 -mrelax -mliw -msetlb}
813 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
815 @emph{MSP430 Options}
816 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
820 @gccoptlist{-mbig-endian -mlittle-endian @gol
821 -mreduced-regs -mfull-regs @gol
822 -mcmov -mno-cmov @gol
823 -mperf-ext -mno-perf-ext @gol
824 -mv3push -mno-v3push @gol
825 -m16bit -mno-16bit @gol
826 -misr-vector-size=@var{num} @gol
827 -mcache-block-size=@var{num} @gol
828 -march=@var{arch} @gol
829 -mcmodel=@var{code-model} @gol
832 @emph{Nios II Options}
833 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
835 -mno-bypass-cache -mbypass-cache @gol
836 -mno-cache-volatile -mcache-volatile @gol
837 -mno-fast-sw-div -mfast-sw-div @gol
838 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
839 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
840 -mcustom-fpu-cfg=@var{name} @gol
841 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
843 @emph{Nvidia PTX Options}
844 @gccoptlist{-m32 -m64 -mmainkernel}
846 @emph{PDP-11 Options}
847 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
848 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
849 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
850 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
851 -mbranch-expensive -mbranch-cheap @gol
852 -munix-asm -mdec-asm}
854 @emph{picoChip Options}
855 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
856 -msymbol-as-address -mno-inefficient-warnings}
858 @emph{PowerPC Options}
859 See RS/6000 and PowerPC Options.
862 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78 @gol
863 -m64bit-doubles -m32bit-doubles}
865 @emph{RS/6000 and PowerPC Options}
866 @gccoptlist{-mcpu=@var{cpu-type} @gol
867 -mtune=@var{cpu-type} @gol
868 -mcmodel=@var{code-model} @gol
870 -maltivec -mno-altivec @gol
871 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
872 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
873 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
874 -mfprnd -mno-fprnd @gol
875 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
876 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
877 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
878 -malign-power -malign-natural @gol
879 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
880 -msingle-float -mdouble-float -msimple-fpu @gol
881 -mstring -mno-string -mupdate -mno-update @gol
882 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
883 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
884 -mstrict-align -mno-strict-align -mrelocatable @gol
885 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
886 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
887 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
888 -mprioritize-restricted-insns=@var{priority} @gol
889 -msched-costly-dep=@var{dependence_type} @gol
890 -minsert-sched-nops=@var{scheme} @gol
891 -mcall-sysv -mcall-netbsd @gol
892 -maix-struct-return -msvr4-struct-return @gol
893 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
894 -mblock-move-inline-limit=@var{num} @gol
895 -misel -mno-isel @gol
896 -misel=yes -misel=no @gol
898 -mspe=yes -mspe=no @gol
900 -mgen-cell-microcode -mwarn-cell-microcode @gol
901 -mvrsave -mno-vrsave @gol
902 -mmulhw -mno-mulhw @gol
903 -mdlmzb -mno-dlmzb @gol
904 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
905 -mprototype -mno-prototype @gol
906 -msim -mmvme -mads -myellowknife -memb -msdata @gol
907 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
908 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
909 -mno-recip-precision @gol
910 -mveclibabi=@var{type} -mfriz -mno-friz @gol
911 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
912 -msave-toc-indirect -mno-save-toc-indirect @gol
913 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
914 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
915 -mquad-memory -mno-quad-memory @gol
916 -mquad-memory-atomic -mno-quad-memory-atomic @gol
917 -mcompat-align-parm -mno-compat-align-parm @gol
918 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
919 -mupper-regs -mno-upper-regs}
922 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
924 -mbig-endian-data -mlittle-endian-data @gol
927 -mas100-syntax -mno-as100-syntax@gol
929 -mmax-constant-size=@gol
932 -mno-warn-multiple-fast-interrupts@gol
933 -msave-acc-in-interrupts}
935 @emph{S/390 and zSeries Options}
936 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
937 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
938 -mlong-double-64 -mlong-double-128 @gol
939 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
940 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
941 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
942 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
943 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
944 -mhotpatch=@var{halfwords},@var{halfwords}}
947 @gccoptlist{-meb -mel @gol
951 -mscore5 -mscore5u -mscore7 -mscore7d}
954 @gccoptlist{-m1 -m2 -m2e @gol
955 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
957 -m4-nofpu -m4-single-only -m4-single -m4 @gol
958 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
959 -m5-64media -m5-64media-nofpu @gol
960 -m5-32media -m5-32media-nofpu @gol
961 -m5-compact -m5-compact-nofpu @gol
962 -mb -ml -mdalign -mrelax @gol
963 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
964 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
965 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
966 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
967 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
968 -maccumulate-outgoing-args -minvalid-symbols @gol
969 -matomic-model=@var{atomic-model} @gol
970 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
971 -mcbranch-force-delay-slot @gol
972 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
973 -mpretend-cmove -mtas}
975 @emph{Solaris 2 Options}
976 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
980 @gccoptlist{-mcpu=@var{cpu-type} @gol
981 -mtune=@var{cpu-type} @gol
982 -mcmodel=@var{code-model} @gol
983 -mmemory-model=@var{mem-model} @gol
984 -m32 -m64 -mapp-regs -mno-app-regs @gol
985 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
986 -mfpu -mno-fpu -mhard-float -msoft-float @gol
987 -mhard-quad-float -msoft-quad-float @gol
988 -mstack-bias -mno-stack-bias @gol
989 -munaligned-doubles -mno-unaligned-doubles @gol
990 -muser-mode -mno-user-mode @gol
991 -mv8plus -mno-v8plus -mvis -mno-vis @gol
992 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
993 -mcbcond -mno-cbcond @gol
994 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
995 -mfix-at697f -mfix-ut699}
998 @gccoptlist{-mwarn-reloc -merror-reloc @gol
999 -msafe-dma -munsafe-dma @gol
1001 -msmall-mem -mlarge-mem -mstdmain @gol
1002 -mfixed-range=@var{register-range} @gol
1004 -maddress-space-conversion -mno-address-space-conversion @gol
1005 -mcache-size=@var{cache-size} @gol
1006 -matomic-updates -mno-atomic-updates}
1008 @emph{System V Options}
1009 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1011 @emph{TILE-Gx Options}
1012 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1013 -mcmodel=@var{code-model}}
1015 @emph{TILEPro Options}
1016 @gccoptlist{-mcpu=@var{cpu} -m32}
1019 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1020 -mprolog-function -mno-prolog-function -mspace @gol
1021 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1022 -mapp-regs -mno-app-regs @gol
1023 -mdisable-callt -mno-disable-callt @gol
1024 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1025 -mv850e -mv850 -mv850e3v5 @gol
1036 @gccoptlist{-mg -mgnu -munix}
1038 @emph{Visium Options}
1039 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1040 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1043 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1044 -mpointer-size=@var{size}}
1046 @emph{VxWorks Options}
1047 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1048 -Xbind-lazy -Xbind-now}
1051 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1052 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1053 -mfpmath=@var{unit} @gol
1054 -masm=@var{dialect} -mno-fancy-math-387 @gol
1055 -mno-fp-ret-in-387 -msoft-float @gol
1056 -mno-wide-multiply -mrtd -malign-double @gol
1057 -mpreferred-stack-boundary=@var{num} @gol
1058 -mincoming-stack-boundary=@var{num} @gol
1059 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1060 -mrecip -mrecip=@var{opt} @gol
1061 -mvzeroupper -mprefer-avx128 @gol
1062 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1063 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
1064 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
1065 -mclflushopt -mxsavec -mxsaves @gol
1066 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1067 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mthreads @gol
1068 -mno-align-stringops -minline-all-stringops @gol
1069 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1070 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1071 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1072 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1073 -mregparm=@var{num} -msseregparm @gol
1074 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1075 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1076 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1077 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1078 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
1079 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1080 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1081 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
1083 @emph{x86 Windows Options}
1084 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1085 -mnop-fun-dllimport -mthread @gol
1086 -municode -mwin32 -mwindows -fno-set-stack-executable}
1088 @emph{Xstormy16 Options}
1091 @emph{Xtensa Options}
1092 @gccoptlist{-mconst16 -mno-const16 @gol
1093 -mfused-madd -mno-fused-madd @gol
1095 -mserialize-volatile -mno-serialize-volatile @gol
1096 -mtext-section-literals -mno-text-section-literals @gol
1097 -mtarget-align -mno-target-align @gol
1098 -mlongcalls -mno-longcalls}
1100 @emph{zSeries Options}
1101 See S/390 and zSeries Options.
1103 @item Code Generation Options
1104 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1105 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1106 -ffixed-@var{reg} -fexceptions @gol
1107 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1108 -fasynchronous-unwind-tables @gol
1109 -fno-gnu-unique @gol
1110 -finhibit-size-directive -finstrument-functions @gol
1111 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1112 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1113 -fno-common -fno-ident @gol
1114 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1115 -fno-jump-tables @gol
1116 -frecord-gcc-switches @gol
1117 -freg-struct-return -fshort-enums @gol
1118 -fshort-double -fshort-wchar @gol
1119 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1120 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1121 -fno-stack-limit -fsplit-stack @gol
1122 -fleading-underscore -ftls-model=@var{model} @gol
1123 -fstack-reuse=@var{reuse_level} @gol
1124 -ftrapv -fwrapv -fbounds-check @gol
1125 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1126 -fstrict-volatile-bitfields -fsync-libcalls}
1130 @node Overall Options
1131 @section Options Controlling the Kind of Output
1133 Compilation can involve up to four stages: preprocessing, compilation
1134 proper, assembly and linking, always in that order. GCC is capable of
1135 preprocessing and compiling several files either into several
1136 assembler input files, or into one assembler input file; then each
1137 assembler input file produces an object file, and linking combines all
1138 the object files (those newly compiled, and those specified as input)
1139 into an executable file.
1141 @cindex file name suffix
1142 For any given input file, the file name suffix determines what kind of
1143 compilation is done:
1147 C source code that must be preprocessed.
1150 C source code that should not be preprocessed.
1153 C++ source code that should not be preprocessed.
1156 Objective-C source code. Note that you must link with the @file{libobjc}
1157 library to make an Objective-C program work.
1160 Objective-C source code that should not be preprocessed.
1164 Objective-C++ source code. Note that you must link with the @file{libobjc}
1165 library to make an Objective-C++ program work. Note that @samp{.M} refers
1166 to a literal capital M@.
1168 @item @var{file}.mii
1169 Objective-C++ source code that should not be preprocessed.
1172 C, C++, Objective-C or Objective-C++ header file to be turned into a
1173 precompiled header (default), or C, C++ header file to be turned into an
1174 Ada spec (via the @option{-fdump-ada-spec} switch).
1177 @itemx @var{file}.cp
1178 @itemx @var{file}.cxx
1179 @itemx @var{file}.cpp
1180 @itemx @var{file}.CPP
1181 @itemx @var{file}.c++
1183 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1184 the last two letters must both be literally @samp{x}. Likewise,
1185 @samp{.C} refers to a literal capital C@.
1189 Objective-C++ source code that must be preprocessed.
1191 @item @var{file}.mii
1192 Objective-C++ source code that should not be preprocessed.
1196 @itemx @var{file}.hp
1197 @itemx @var{file}.hxx
1198 @itemx @var{file}.hpp
1199 @itemx @var{file}.HPP
1200 @itemx @var{file}.h++
1201 @itemx @var{file}.tcc
1202 C++ header file to be turned into a precompiled header or Ada spec.
1205 @itemx @var{file}.for
1206 @itemx @var{file}.ftn
1207 Fixed form Fortran source code that should not be preprocessed.
1210 @itemx @var{file}.FOR
1211 @itemx @var{file}.fpp
1212 @itemx @var{file}.FPP
1213 @itemx @var{file}.FTN
1214 Fixed form Fortran source code that must be preprocessed (with the traditional
1217 @item @var{file}.f90
1218 @itemx @var{file}.f95
1219 @itemx @var{file}.f03
1220 @itemx @var{file}.f08
1221 Free form Fortran source code that should not be preprocessed.
1223 @item @var{file}.F90
1224 @itemx @var{file}.F95
1225 @itemx @var{file}.F03
1226 @itemx @var{file}.F08
1227 Free form Fortran source code that must be preprocessed (with the
1228 traditional preprocessor).
1233 @c FIXME: Descriptions of Java file types.
1239 @item @var{file}.ads
1240 Ada source code file that contains a library unit declaration (a
1241 declaration of a package, subprogram, or generic, or a generic
1242 instantiation), or a library unit renaming declaration (a package,
1243 generic, or subprogram renaming declaration). Such files are also
1246 @item @var{file}.adb
1247 Ada source code file containing a library unit body (a subprogram or
1248 package body). Such files are also called @dfn{bodies}.
1250 @c GCC also knows about some suffixes for languages not yet included:
1261 @itemx @var{file}.sx
1262 Assembler code that must be preprocessed.
1265 An object file to be fed straight into linking.
1266 Any file name with no recognized suffix is treated this way.
1270 You can specify the input language explicitly with the @option{-x} option:
1273 @item -x @var{language}
1274 Specify explicitly the @var{language} for the following input files
1275 (rather than letting the compiler choose a default based on the file
1276 name suffix). This option applies to all following input files until
1277 the next @option{-x} option. Possible values for @var{language} are:
1279 c c-header cpp-output
1280 c++ c++-header c++-cpp-output
1281 objective-c objective-c-header objective-c-cpp-output
1282 objective-c++ objective-c++-header objective-c++-cpp-output
1283 assembler assembler-with-cpp
1285 f77 f77-cpp-input f95 f95-cpp-input
1291 Turn off any specification of a language, so that subsequent files are
1292 handled according to their file name suffixes (as they are if @option{-x}
1293 has not been used at all).
1295 @item -pass-exit-codes
1296 @opindex pass-exit-codes
1297 Normally the @command{gcc} program exits with the code of 1 if any
1298 phase of the compiler returns a non-success return code. If you specify
1299 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1300 the numerically highest error produced by any phase returning an error
1301 indication. The C, C++, and Fortran front ends return 4 if an internal
1302 compiler error is encountered.
1305 If you only want some of the stages of compilation, you can use
1306 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1307 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1308 @command{gcc} is to stop. Note that some combinations (for example,
1309 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1314 Compile or assemble the source files, but do not link. The linking
1315 stage simply is not done. The ultimate output is in the form of an
1316 object file for each source file.
1318 By default, the object file name for a source file is made by replacing
1319 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1321 Unrecognized input files, not requiring compilation or assembly, are
1326 Stop after the stage of compilation proper; do not assemble. The output
1327 is in the form of an assembler code file for each non-assembler input
1330 By default, the assembler file name for a source file is made by
1331 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1333 Input files that don't require compilation are ignored.
1337 Stop after the preprocessing stage; do not run the compiler proper. The
1338 output is in the form of preprocessed source code, which is sent to the
1341 Input files that don't require preprocessing are ignored.
1343 @cindex output file option
1346 Place output in file @var{file}. This applies to whatever
1347 sort of output is being produced, whether it be an executable file,
1348 an object file, an assembler file or preprocessed C code.
1350 If @option{-o} is not specified, the default is to put an executable
1351 file in @file{a.out}, the object file for
1352 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1353 assembler file in @file{@var{source}.s}, a precompiled header file in
1354 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1359 Print (on standard error output) the commands executed to run the stages
1360 of compilation. Also print the version number of the compiler driver
1361 program and of the preprocessor and the compiler proper.
1365 Like @option{-v} except the commands are not executed and arguments
1366 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1367 This is useful for shell scripts to capture the driver-generated command lines.
1371 Use pipes rather than temporary files for communication between the
1372 various stages of compilation. This fails to work on some systems where
1373 the assembler is unable to read from a pipe; but the GNU assembler has
1378 Print (on the standard output) a description of the command-line options
1379 understood by @command{gcc}. If the @option{-v} option is also specified
1380 then @option{--help} is also passed on to the various processes
1381 invoked by @command{gcc}, so that they can display the command-line options
1382 they accept. If the @option{-Wextra} option has also been specified
1383 (prior to the @option{--help} option), then command-line options that
1384 have no documentation associated with them are also displayed.
1387 @opindex target-help
1388 Print (on the standard output) a description of target-specific command-line
1389 options for each tool. For some targets extra target-specific
1390 information may also be printed.
1392 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1393 Print (on the standard output) a description of the command-line
1394 options understood by the compiler that fit into all specified classes
1395 and qualifiers. These are the supported classes:
1398 @item @samp{optimizers}
1399 Display all of the optimization options supported by the
1402 @item @samp{warnings}
1403 Display all of the options controlling warning messages
1404 produced by the compiler.
1407 Display target-specific options. Unlike the
1408 @option{--target-help} option however, target-specific options of the
1409 linker and assembler are not displayed. This is because those
1410 tools do not currently support the extended @option{--help=} syntax.
1413 Display the values recognized by the @option{--param}
1416 @item @var{language}
1417 Display the options supported for @var{language}, where
1418 @var{language} is the name of one of the languages supported in this
1422 Display the options that are common to all languages.
1425 These are the supported qualifiers:
1428 @item @samp{undocumented}
1429 Display only those options that are undocumented.
1432 Display options taking an argument that appears after an equal
1433 sign in the same continuous piece of text, such as:
1434 @samp{--help=target}.
1436 @item @samp{separate}
1437 Display options taking an argument that appears as a separate word
1438 following the original option, such as: @samp{-o output-file}.
1441 Thus for example to display all the undocumented target-specific
1442 switches supported by the compiler, use:
1445 --help=target,undocumented
1448 The sense of a qualifier can be inverted by prefixing it with the
1449 @samp{^} character, so for example to display all binary warning
1450 options (i.e., ones that are either on or off and that do not take an
1451 argument) that have a description, use:
1454 --help=warnings,^joined,^undocumented
1457 The argument to @option{--help=} should not consist solely of inverted
1460 Combining several classes is possible, although this usually
1461 restricts the output so much that there is nothing to display. One
1462 case where it does work, however, is when one of the classes is
1463 @var{target}. For example, to display all the target-specific
1464 optimization options, use:
1467 --help=target,optimizers
1470 The @option{--help=} option can be repeated on the command line. Each
1471 successive use displays its requested class of options, skipping
1472 those that have already been displayed.
1474 If the @option{-Q} option appears on the command line before the
1475 @option{--help=} option, then the descriptive text displayed by
1476 @option{--help=} is changed. Instead of describing the displayed
1477 options, an indication is given as to whether the option is enabled,
1478 disabled or set to a specific value (assuming that the compiler
1479 knows this at the point where the @option{--help=} option is used).
1481 Here is a truncated example from the ARM port of @command{gcc}:
1484 % gcc -Q -mabi=2 --help=target -c
1485 The following options are target specific:
1487 -mabort-on-noreturn [disabled]
1491 The output is sensitive to the effects of previous command-line
1492 options, so for example it is possible to find out which optimizations
1493 are enabled at @option{-O2} by using:
1496 -Q -O2 --help=optimizers
1499 Alternatively you can discover which binary optimizations are enabled
1500 by @option{-O3} by using:
1503 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1504 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1505 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1508 @item -no-canonical-prefixes
1509 @opindex no-canonical-prefixes
1510 Do not expand any symbolic links, resolve references to @samp{/../}
1511 or @samp{/./}, or make the path absolute when generating a relative
1516 Display the version number and copyrights of the invoked GCC@.
1520 Invoke all subcommands under a wrapper program. The name of the
1521 wrapper program and its parameters are passed as a comma separated
1525 gcc -c t.c -wrapper gdb,--args
1529 This invokes all subprograms of @command{gcc} under
1530 @samp{gdb --args}, thus the invocation of @command{cc1} is
1531 @samp{gdb --args cc1 @dots{}}.
1533 @item -fplugin=@var{name}.so
1535 Load the plugin code in file @var{name}.so, assumed to be a
1536 shared object to be dlopen'd by the compiler. The base name of
1537 the shared object file is used to identify the plugin for the
1538 purposes of argument parsing (See
1539 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1540 Each plugin should define the callback functions specified in the
1543 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1544 @opindex fplugin-arg
1545 Define an argument called @var{key} with a value of @var{value}
1546 for the plugin called @var{name}.
1548 @item -fdump-ada-spec@r{[}-slim@r{]}
1549 @opindex fdump-ada-spec
1550 For C and C++ source and include files, generate corresponding Ada specs.
1551 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1552 GNAT User's Guide}, which provides detailed documentation on this feature.
1554 @item -fada-spec-parent=@var{unit}
1555 @opindex fada-spec-parent
1556 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1557 Ada specs as child units of parent @var{unit}.
1559 @item -fdump-go-spec=@var{file}
1560 @opindex fdump-go-spec
1561 For input files in any language, generate corresponding Go
1562 declarations in @var{file}. This generates Go @code{const},
1563 @code{type}, @code{var}, and @code{func} declarations which may be a
1564 useful way to start writing a Go interface to code written in some
1567 @include @value{srcdir}/../libiberty/at-file.texi
1571 @section Compiling C++ Programs
1573 @cindex suffixes for C++ source
1574 @cindex C++ source file suffixes
1575 C++ source files conventionally use one of the suffixes @samp{.C},
1576 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1577 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1578 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1579 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1580 files with these names and compiles them as C++ programs even if you
1581 call the compiler the same way as for compiling C programs (usually
1582 with the name @command{gcc}).
1586 However, the use of @command{gcc} does not add the C++ library.
1587 @command{g++} is a program that calls GCC and automatically specifies linking
1588 against the C++ library. It treats @samp{.c},
1589 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1590 files unless @option{-x} is used. This program is also useful when
1591 precompiling a C header file with a @samp{.h} extension for use in C++
1592 compilations. On many systems, @command{g++} is also installed with
1593 the name @command{c++}.
1595 @cindex invoking @command{g++}
1596 When you compile C++ programs, you may specify many of the same
1597 command-line options that you use for compiling programs in any
1598 language; or command-line options meaningful for C and related
1599 languages; or options that are meaningful only for C++ programs.
1600 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1601 explanations of options for languages related to C@.
1602 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1603 explanations of options that are meaningful only for C++ programs.
1605 @node C Dialect Options
1606 @section Options Controlling C Dialect
1607 @cindex dialect options
1608 @cindex language dialect options
1609 @cindex options, dialect
1611 The following options control the dialect of C (or languages derived
1612 from C, such as C++, Objective-C and Objective-C++) that the compiler
1616 @cindex ANSI support
1620 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1621 equivalent to @option{-std=c++98}.
1623 This turns off certain features of GCC that are incompatible with ISO
1624 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1625 such as the @code{asm} and @code{typeof} keywords, and
1626 predefined macros such as @code{unix} and @code{vax} that identify the
1627 type of system you are using. It also enables the undesirable and
1628 rarely used ISO trigraph feature. For the C compiler,
1629 it disables recognition of C++ style @samp{//} comments as well as
1630 the @code{inline} keyword.
1632 The alternate keywords @code{__asm__}, @code{__extension__},
1633 @code{__inline__} and @code{__typeof__} continue to work despite
1634 @option{-ansi}. You would not want to use them in an ISO C program, of
1635 course, but it is useful to put them in header files that might be included
1636 in compilations done with @option{-ansi}. Alternate predefined macros
1637 such as @code{__unix__} and @code{__vax__} are also available, with or
1638 without @option{-ansi}.
1640 The @option{-ansi} option does not cause non-ISO programs to be
1641 rejected gratuitously. For that, @option{-Wpedantic} is required in
1642 addition to @option{-ansi}. @xref{Warning Options}.
1644 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1645 option is used. Some header files may notice this macro and refrain
1646 from declaring certain functions or defining certain macros that the
1647 ISO standard doesn't call for; this is to avoid interfering with any
1648 programs that might use these names for other things.
1650 Functions that are normally built in but do not have semantics
1651 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1652 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1653 built-in functions provided by GCC}, for details of the functions
1658 Determine the language standard. @xref{Standards,,Language Standards
1659 Supported by GCC}, for details of these standard versions. This option
1660 is currently only supported when compiling C or C++.
1662 The compiler can accept several base standards, such as @samp{c90} or
1663 @samp{c++98}, and GNU dialects of those standards, such as
1664 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1665 compiler accepts all programs following that standard plus those
1666 using GNU extensions that do not contradict it. For example,
1667 @option{-std=c90} turns off certain features of GCC that are
1668 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1669 keywords, but not other GNU extensions that do not have a meaning in
1670 ISO C90, such as omitting the middle term of a @code{?:}
1671 expression. On the other hand, when a GNU dialect of a standard is
1672 specified, all features supported by the compiler are enabled, even when
1673 those features change the meaning of the base standard. As a result, some
1674 strict-conforming programs may be rejected. The particular standard
1675 is used by @option{-Wpedantic} to identify which features are GNU
1676 extensions given that version of the standard. For example
1677 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1678 comments, while @option{-std=gnu99 -Wpedantic} does not.
1680 A value for this option must be provided; possible values are
1686 Support all ISO C90 programs (certain GNU extensions that conflict
1687 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1689 @item iso9899:199409
1690 ISO C90 as modified in amendment 1.
1696 ISO C99. This standard is substantially completely supported, modulo
1697 bugs and floating-point issues
1698 (mainly but not entirely relating to optional C99 features from
1699 Annexes F and G). See
1700 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1701 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1706 ISO C11, the 2011 revision of the ISO C standard. This standard is
1707 substantially completely supported, modulo bugs, floating-point issues
1708 (mainly but not entirely relating to optional C11 features from
1709 Annexes F and G) and the optional Annexes K (Bounds-checking
1710 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1714 GNU dialect of ISO C90 (including some C99 features).
1718 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1722 GNU dialect of ISO C11. This is the default for C code.
1723 The name @samp{gnu1x} is deprecated.
1727 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1728 additional defect reports. Same as @option{-ansi} for C++ code.
1732 GNU dialect of @option{-std=c++98}. This is the default for
1737 The 2011 ISO C++ standard plus amendments.
1738 The name @samp{c++0x} is deprecated.
1742 GNU dialect of @option{-std=c++11}.
1743 The name @samp{gnu++0x} is deprecated.
1747 The 2014 ISO C++ standard plus amendments.
1748 The name @samp{c++1y} is deprecated.
1752 GNU dialect of @option{-std=c++14}.
1753 The name @samp{gnu++1y} is deprecated.
1756 The next revision of the ISO C++ standard, tentatively planned for
1757 2017. Support is highly experimental, and will almost certainly
1758 change in incompatible ways in future releases.
1761 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1762 and will almost certainly change in incompatible ways in future
1766 @item -fgnu89-inline
1767 @opindex fgnu89-inline
1768 The option @option{-fgnu89-inline} tells GCC to use the traditional
1769 GNU semantics for @code{inline} functions when in C99 mode.
1770 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1771 is accepted and ignored by GCC versions 4.1.3 up to but not including
1772 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1773 C99 mode. Using this option is roughly equivalent to adding the
1774 @code{gnu_inline} function attribute to all inline functions
1775 (@pxref{Function Attributes}).
1777 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1778 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1779 specifies the default behavior). This option was first supported in
1780 GCC 4.3. This option is not supported in @option{-std=c90} or
1781 @option{-std=gnu90} mode.
1783 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1784 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1785 in effect for @code{inline} functions. @xref{Common Predefined
1786 Macros,,,cpp,The C Preprocessor}.
1788 @item -aux-info @var{filename}
1790 Output to the given filename prototyped declarations for all functions
1791 declared and/or defined in a translation unit, including those in header
1792 files. This option is silently ignored in any language other than C@.
1794 Besides declarations, the file indicates, in comments, the origin of
1795 each declaration (source file and line), whether the declaration was
1796 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1797 @samp{O} for old, respectively, in the first character after the line
1798 number and the colon), and whether it came from a declaration or a
1799 definition (@samp{C} or @samp{F}, respectively, in the following
1800 character). In the case of function definitions, a K&R-style list of
1801 arguments followed by their declarations is also provided, inside
1802 comments, after the declaration.
1804 @item -fallow-parameterless-variadic-functions
1805 @opindex fallow-parameterless-variadic-functions
1806 Accept variadic functions without named parameters.
1808 Although it is possible to define such a function, this is not very
1809 useful as it is not possible to read the arguments. This is only
1810 supported for C as this construct is allowed by C++.
1814 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1815 keyword, so that code can use these words as identifiers. You can use
1816 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1817 instead. @option{-ansi} implies @option{-fno-asm}.
1819 In C++, this switch only affects the @code{typeof} keyword, since
1820 @code{asm} and @code{inline} are standard keywords. You may want to
1821 use the @option{-fno-gnu-keywords} flag instead, which has the same
1822 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1823 switch only affects the @code{asm} and @code{typeof} keywords, since
1824 @code{inline} is a standard keyword in ISO C99.
1827 @itemx -fno-builtin-@var{function}
1828 @opindex fno-builtin
1829 @cindex built-in functions
1830 Don't recognize built-in functions that do not begin with
1831 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1832 functions provided by GCC}, for details of the functions affected,
1833 including those which are not built-in functions when @option{-ansi} or
1834 @option{-std} options for strict ISO C conformance are used because they
1835 do not have an ISO standard meaning.
1837 GCC normally generates special code to handle certain built-in functions
1838 more efficiently; for instance, calls to @code{alloca} may become single
1839 instructions which adjust the stack directly, and calls to @code{memcpy}
1840 may become inline copy loops. The resulting code is often both smaller
1841 and faster, but since the function calls no longer appear as such, you
1842 cannot set a breakpoint on those calls, nor can you change the behavior
1843 of the functions by linking with a different library. In addition,
1844 when a function is recognized as a built-in function, GCC may use
1845 information about that function to warn about problems with calls to
1846 that function, or to generate more efficient code, even if the
1847 resulting code still contains calls to that function. For example,
1848 warnings are given with @option{-Wformat} for bad calls to
1849 @code{printf} when @code{printf} is built in and @code{strlen} is
1850 known not to modify global memory.
1852 With the @option{-fno-builtin-@var{function}} option
1853 only the built-in function @var{function} is
1854 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1855 function is named that is not built-in in this version of GCC, this
1856 option is ignored. There is no corresponding
1857 @option{-fbuiltin-@var{function}} option; if you wish to enable
1858 built-in functions selectively when using @option{-fno-builtin} or
1859 @option{-ffreestanding}, you may define macros such as:
1862 #define abs(n) __builtin_abs ((n))
1863 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1868 @cindex hosted environment
1870 Assert that compilation targets a hosted environment. This implies
1871 @option{-fbuiltin}. A hosted environment is one in which the
1872 entire standard library is available, and in which @code{main} has a return
1873 type of @code{int}. Examples are nearly everything except a kernel.
1874 This is equivalent to @option{-fno-freestanding}.
1876 @item -ffreestanding
1877 @opindex ffreestanding
1878 @cindex hosted environment
1880 Assert that compilation targets a freestanding environment. This
1881 implies @option{-fno-builtin}. A freestanding environment
1882 is one in which the standard library may not exist, and program startup may
1883 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1884 This is equivalent to @option{-fno-hosted}.
1886 @xref{Standards,,Language Standards Supported by GCC}, for details of
1887 freestanding and hosted environments.
1891 @cindex OpenACC accelerator programming
1892 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1893 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1894 compiler generates accelerated code according to the OpenACC Application
1895 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1896 implies @option{-pthread}, and thus is only supported on targets that
1897 have support for @option{-pthread}.
1899 Note that this is an experimental feature, incomplete, and subject to
1900 change in future versions of GCC. See
1901 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1905 @cindex OpenMP parallel
1906 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1907 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1908 compiler generates parallel code according to the OpenMP Application
1909 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1910 implies @option{-pthread}, and thus is only supported on targets that
1911 have support for @option{-pthread}. @option{-fopenmp} implies
1912 @option{-fopenmp-simd}.
1915 @opindex fopenmp-simd
1918 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1919 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1924 @cindex Enable Cilk Plus
1925 Enable the usage of Cilk Plus language extension features for C/C++.
1926 When the option @option{-fcilkplus} is specified, enable the usage of
1927 the Cilk Plus Language extension features for C/C++. The present
1928 implementation follows ABI version 1.2. This is an experimental
1929 feature that is only partially complete, and whose interface may
1930 change in future versions of GCC as the official specification
1931 changes. Currently, all features but @code{_Cilk_for} have been
1936 When the option @option{-fgnu-tm} is specified, the compiler
1937 generates code for the Linux variant of Intel's current Transactional
1938 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1939 an experimental feature whose interface may change in future versions
1940 of GCC, as the official specification changes. Please note that not
1941 all architectures are supported for this feature.
1943 For more information on GCC's support for transactional memory,
1944 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1945 Transactional Memory Library}.
1947 Note that the transactional memory feature is not supported with
1948 non-call exceptions (@option{-fnon-call-exceptions}).
1950 @item -fms-extensions
1951 @opindex fms-extensions
1952 Accept some non-standard constructs used in Microsoft header files.
1954 In C++ code, this allows member names in structures to be similar
1955 to previous types declarations.
1964 Some cases of unnamed fields in structures and unions are only
1965 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1966 fields within structs/unions}, for details.
1968 Note that this option is off for all targets but x86
1969 targets using ms-abi.
1971 @item -fplan9-extensions
1972 @opindex fplan9-extensions
1973 Accept some non-standard constructs used in Plan 9 code.
1975 This enables @option{-fms-extensions}, permits passing pointers to
1976 structures with anonymous fields to functions that expect pointers to
1977 elements of the type of the field, and permits referring to anonymous
1978 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1979 struct/union fields within structs/unions}, for details. This is only
1980 supported for C, not C++.
1984 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1985 options for strict ISO C conformance) implies @option{-trigraphs}.
1987 @cindex traditional C language
1988 @cindex C language, traditional
1990 @itemx -traditional-cpp
1991 @opindex traditional-cpp
1992 @opindex traditional
1993 Formerly, these options caused GCC to attempt to emulate a pre-standard
1994 C compiler. They are now only supported with the @option{-E} switch.
1995 The preprocessor continues to support a pre-standard mode. See the GNU
1996 CPP manual for details.
1998 @item -fcond-mismatch
1999 @opindex fcond-mismatch
2000 Allow conditional expressions with mismatched types in the second and
2001 third arguments. The value of such an expression is void. This option
2002 is not supported for C++.
2004 @item -flax-vector-conversions
2005 @opindex flax-vector-conversions
2006 Allow implicit conversions between vectors with differing numbers of
2007 elements and/or incompatible element types. This option should not be
2010 @item -funsigned-char
2011 @opindex funsigned-char
2012 Let the type @code{char} be unsigned, like @code{unsigned char}.
2014 Each kind of machine has a default for what @code{char} should
2015 be. It is either like @code{unsigned char} by default or like
2016 @code{signed char} by default.
2018 Ideally, a portable program should always use @code{signed char} or
2019 @code{unsigned char} when it depends on the signedness of an object.
2020 But many programs have been written to use plain @code{char} and
2021 expect it to be signed, or expect it to be unsigned, depending on the
2022 machines they were written for. This option, and its inverse, let you
2023 make such a program work with the opposite default.
2025 The type @code{char} is always a distinct type from each of
2026 @code{signed char} or @code{unsigned char}, even though its behavior
2027 is always just like one of those two.
2030 @opindex fsigned-char
2031 Let the type @code{char} be signed, like @code{signed char}.
2033 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2034 the negative form of @option{-funsigned-char}. Likewise, the option
2035 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2037 @item -fsigned-bitfields
2038 @itemx -funsigned-bitfields
2039 @itemx -fno-signed-bitfields
2040 @itemx -fno-unsigned-bitfields
2041 @opindex fsigned-bitfields
2042 @opindex funsigned-bitfields
2043 @opindex fno-signed-bitfields
2044 @opindex fno-unsigned-bitfields
2045 These options control whether a bit-field is signed or unsigned, when the
2046 declaration does not use either @code{signed} or @code{unsigned}. By
2047 default, such a bit-field is signed, because this is consistent: the
2048 basic integer types such as @code{int} are signed types.
2051 @node C++ Dialect Options
2052 @section Options Controlling C++ Dialect
2054 @cindex compiler options, C++
2055 @cindex C++ options, command-line
2056 @cindex options, C++
2057 This section describes the command-line options that are only meaningful
2058 for C++ programs. You can also use most of the GNU compiler options
2059 regardless of what language your program is in. For example, you
2060 might compile a file @file{firstClass.C} like this:
2063 g++ -g -frepo -O -c firstClass.C
2067 In this example, only @option{-frepo} is an option meant
2068 only for C++ programs; you can use the other options with any
2069 language supported by GCC@.
2071 Here is a list of options that are @emph{only} for compiling C++ programs:
2075 @item -fabi-version=@var{n}
2076 @opindex fabi-version
2077 Use version @var{n} of the C++ ABI@. The default is version 0.
2079 Version 0 refers to the version conforming most closely to
2080 the C++ ABI specification. Therefore, the ABI obtained using version 0
2081 will change in different versions of G++ as ABI bugs are fixed.
2083 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2085 Version 2 is the version of the C++ ABI that first appeared in G++
2086 3.4, and was the default through G++ 4.9.
2088 Version 3 corrects an error in mangling a constant address as a
2091 Version 4, which first appeared in G++ 4.5, implements a standard
2092 mangling for vector types.
2094 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2095 attribute const/volatile on function pointer types, decltype of a
2096 plain decl, and use of a function parameter in the declaration of
2099 Version 6, which first appeared in G++ 4.7, corrects the promotion
2100 behavior of C++11 scoped enums and the mangling of template argument
2101 packs, const/static_cast, prefix ++ and --, and a class scope function
2102 used as a template argument.
2104 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2105 builtin type and corrects the mangling of lambdas in default argument
2108 Version 8, which first appeared in G++ 4.9, corrects the substitution
2109 behavior of function types with function-cv-qualifiers.
2111 See also @option{-Wabi}.
2113 @item -fabi-compat-version=@var{n}
2114 @opindex fabi-compat-version
2115 Starting with GCC 4.5, on targets that support strong aliases, G++
2116 works around mangling changes by creating an alias with the correct
2117 mangled name when defining a symbol with an incorrect mangled name.
2118 This switch specifies which ABI version to use for the alias.
2120 With @option{-fabi-version=0} (the default), this defaults to 2. If
2121 another ABI version is explicitly selected, this defaults to 0.
2123 The compatibility version is also set by @option{-Wabi=@var{n}}.
2125 @item -fno-access-control
2126 @opindex fno-access-control
2127 Turn off all access checking. This switch is mainly useful for working
2128 around bugs in the access control code.
2132 Check that the pointer returned by @code{operator new} is non-null
2133 before attempting to modify the storage allocated. This check is
2134 normally unnecessary because the C++ standard specifies that
2135 @code{operator new} only returns @code{0} if it is declared
2136 @code{throw()}, in which case the compiler always checks the
2137 return value even without this option. In all other cases, when
2138 @code{operator new} has a non-empty exception specification, memory
2139 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2140 @samp{new (nothrow)}.
2142 @item -fconstexpr-depth=@var{n}
2143 @opindex fconstexpr-depth
2144 Set the maximum nested evaluation depth for C++11 constexpr functions
2145 to @var{n}. A limit is needed to detect endless recursion during
2146 constant expression evaluation. The minimum specified by the standard
2149 @item -fdeduce-init-list
2150 @opindex fdeduce-init-list
2151 Enable deduction of a template type parameter as
2152 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2155 template <class T> auto forward(T t) -> decltype (realfn (t))
2162 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2166 This deduction was implemented as a possible extension to the
2167 originally proposed semantics for the C++11 standard, but was not part
2168 of the final standard, so it is disabled by default. This option is
2169 deprecated, and may be removed in a future version of G++.
2171 @item -ffriend-injection
2172 @opindex ffriend-injection
2173 Inject friend functions into the enclosing namespace, so that they are
2174 visible outside the scope of the class in which they are declared.
2175 Friend functions were documented to work this way in the old Annotated
2176 C++ Reference Manual, and versions of G++ before 4.1 always worked
2177 that way. However, in ISO C++ a friend function that is not declared
2178 in an enclosing scope can only be found using argument dependent
2179 lookup. This option causes friends to be injected as they were in
2182 This option is for compatibility, and may be removed in a future
2185 @item -fno-elide-constructors
2186 @opindex fno-elide-constructors
2187 The C++ standard allows an implementation to omit creating a temporary
2188 that is only used to initialize another object of the same type.
2189 Specifying this option disables that optimization, and forces G++ to
2190 call the copy constructor in all cases.
2192 @item -fno-enforce-eh-specs
2193 @opindex fno-enforce-eh-specs
2194 Don't generate code to check for violation of exception specifications
2195 at run time. This option violates the C++ standard, but may be useful
2196 for reducing code size in production builds, much like defining
2197 @code{NDEBUG}. This does not give user code permission to throw
2198 exceptions in violation of the exception specifications; the compiler
2199 still optimizes based on the specifications, so throwing an
2200 unexpected exception results in undefined behavior at run time.
2202 @item -fextern-tls-init
2203 @itemx -fno-extern-tls-init
2204 @opindex fextern-tls-init
2205 @opindex fno-extern-tls-init
2206 The C++11 and OpenMP standards allow @code{thread_local} and
2207 @code{threadprivate} variables to have dynamic (runtime)
2208 initialization. To support this, any use of such a variable goes
2209 through a wrapper function that performs any necessary initialization.
2210 When the use and definition of the variable are in the same
2211 translation unit, this overhead can be optimized away, but when the
2212 use is in a different translation unit there is significant overhead
2213 even if the variable doesn't actually need dynamic initialization. If
2214 the programmer can be sure that no use of the variable in a
2215 non-defining TU needs to trigger dynamic initialization (either
2216 because the variable is statically initialized, or a use of the
2217 variable in the defining TU will be executed before any uses in
2218 another TU), they can avoid this overhead with the
2219 @option{-fno-extern-tls-init} option.
2221 On targets that support symbol aliases, the default is
2222 @option{-fextern-tls-init}. On targets that do not support symbol
2223 aliases, the default is @option{-fno-extern-tls-init}.
2226 @itemx -fno-for-scope
2228 @opindex fno-for-scope
2229 If @option{-ffor-scope} is specified, the scope of variables declared in
2230 a @i{for-init-statement} is limited to the @code{for} loop itself,
2231 as specified by the C++ standard.
2232 If @option{-fno-for-scope} is specified, the scope of variables declared in
2233 a @i{for-init-statement} extends to the end of the enclosing scope,
2234 as was the case in old versions of G++, and other (traditional)
2235 implementations of C++.
2237 If neither flag is given, the default is to follow the standard,
2238 but to allow and give a warning for old-style code that would
2239 otherwise be invalid, or have different behavior.
2241 @item -fno-gnu-keywords
2242 @opindex fno-gnu-keywords
2243 Do not recognize @code{typeof} as a keyword, so that code can use this
2244 word as an identifier. You can use the keyword @code{__typeof__} instead.
2245 @option{-ansi} implies @option{-fno-gnu-keywords}.
2247 @item -fno-implicit-templates
2248 @opindex fno-implicit-templates
2249 Never emit code for non-inline templates that are instantiated
2250 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2251 @xref{Template Instantiation}, for more information.
2253 @item -fno-implicit-inline-templates
2254 @opindex fno-implicit-inline-templates
2255 Don't emit code for implicit instantiations of inline templates, either.
2256 The default is to handle inlines differently so that compiles with and
2257 without optimization need the same set of explicit instantiations.
2259 @item -fno-implement-inlines
2260 @opindex fno-implement-inlines
2261 To save space, do not emit out-of-line copies of inline functions
2262 controlled by @code{#pragma implementation}. This causes linker
2263 errors if these functions are not inlined everywhere they are called.
2265 @item -fms-extensions
2266 @opindex fms-extensions
2267 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2268 int and getting a pointer to member function via non-standard syntax.
2270 @item -fno-nonansi-builtins
2271 @opindex fno-nonansi-builtins
2272 Disable built-in declarations of functions that are not mandated by
2273 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2274 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2277 @opindex fnothrow-opt
2278 Treat a @code{throw()} exception specification as if it were a
2279 @code{noexcept} specification to reduce or eliminate the text size
2280 overhead relative to a function with no exception specification. If
2281 the function has local variables of types with non-trivial
2282 destructors, the exception specification actually makes the
2283 function smaller because the EH cleanups for those variables can be
2284 optimized away. The semantic effect is that an exception thrown out of
2285 a function with such an exception specification results in a call
2286 to @code{terminate} rather than @code{unexpected}.
2288 @item -fno-operator-names
2289 @opindex fno-operator-names
2290 Do not treat the operator name keywords @code{and}, @code{bitand},
2291 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2292 synonyms as keywords.
2294 @item -fno-optional-diags
2295 @opindex fno-optional-diags
2296 Disable diagnostics that the standard says a compiler does not need to
2297 issue. Currently, the only such diagnostic issued by G++ is the one for
2298 a name having multiple meanings within a class.
2301 @opindex fpermissive
2302 Downgrade some diagnostics about nonconformant code from errors to
2303 warnings. Thus, using @option{-fpermissive} allows some
2304 nonconforming code to compile.
2306 @item -fno-pretty-templates
2307 @opindex fno-pretty-templates
2308 When an error message refers to a specialization of a function
2309 template, the compiler normally prints the signature of the
2310 template followed by the template arguments and any typedefs or
2311 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2312 rather than @code{void f(int)}) so that it's clear which template is
2313 involved. When an error message refers to a specialization of a class
2314 template, the compiler omits any template arguments that match
2315 the default template arguments for that template. If either of these
2316 behaviors make it harder to understand the error message rather than
2317 easier, you can use @option{-fno-pretty-templates} to disable them.
2321 Enable automatic template instantiation at link time. This option also
2322 implies @option{-fno-implicit-templates}. @xref{Template
2323 Instantiation}, for more information.
2327 Disable generation of information about every class with virtual
2328 functions for use by the C++ run-time type identification features
2329 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2330 of the language, you can save some space by using this flag. Note that
2331 exception handling uses the same information, but G++ generates it as
2332 needed. The @code{dynamic_cast} operator can still be used for casts that
2333 do not require run-time type information, i.e.@: casts to @code{void *} or to
2334 unambiguous base classes.
2336 @item -fsized-deallocation
2337 @opindex fsized-deallocation
2338 Enable the built-in global declarations
2340 void operator delete (void *, std::size_t) noexcept;
2341 void operator delete[] (void *, std::size_t) noexcept;
2343 as introduced in C++14. This is useful for user-defined replacement
2344 deallocation functions that, for example, use the size of the object
2345 to make deallocation faster. Enabled by default under
2346 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2347 warns about places that might want to add a definition.
2351 Emit statistics about front-end processing at the end of the compilation.
2352 This information is generally only useful to the G++ development team.
2354 @item -fstrict-enums
2355 @opindex fstrict-enums
2356 Allow the compiler to optimize using the assumption that a value of
2357 enumerated type can only be one of the values of the enumeration (as
2358 defined in the C++ standard; basically, a value that can be
2359 represented in the minimum number of bits needed to represent all the
2360 enumerators). This assumption may not be valid if the program uses a
2361 cast to convert an arbitrary integer value to the enumerated type.
2363 @item -ftemplate-backtrace-limit=@var{n}
2364 @opindex ftemplate-backtrace-limit
2365 Set the maximum number of template instantiation notes for a single
2366 warning or error to @var{n}. The default value is 10.
2368 @item -ftemplate-depth=@var{n}
2369 @opindex ftemplate-depth
2370 Set the maximum instantiation depth for template classes to @var{n}.
2371 A limit on the template instantiation depth is needed to detect
2372 endless recursions during template class instantiation. ANSI/ISO C++
2373 conforming programs must not rely on a maximum depth greater than 17
2374 (changed to 1024 in C++11). The default value is 900, as the compiler
2375 can run out of stack space before hitting 1024 in some situations.
2377 @item -fno-threadsafe-statics
2378 @opindex fno-threadsafe-statics
2379 Do not emit the extra code to use the routines specified in the C++
2380 ABI for thread-safe initialization of local statics. You can use this
2381 option to reduce code size slightly in code that doesn't need to be
2384 @item -fuse-cxa-atexit
2385 @opindex fuse-cxa-atexit
2386 Register destructors for objects with static storage duration with the
2387 @code{__cxa_atexit} function rather than the @code{atexit} function.
2388 This option is required for fully standards-compliant handling of static
2389 destructors, but only works if your C library supports
2390 @code{__cxa_atexit}.
2392 @item -fno-use-cxa-get-exception-ptr
2393 @opindex fno-use-cxa-get-exception-ptr
2394 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2395 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2396 if the runtime routine is not available.
2398 @item -fvisibility-inlines-hidden
2399 @opindex fvisibility-inlines-hidden
2400 This switch declares that the user does not attempt to compare
2401 pointers to inline functions or methods where the addresses of the two functions
2402 are taken in different shared objects.
2404 The effect of this is that GCC may, effectively, mark inline methods with
2405 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2406 appear in the export table of a DSO and do not require a PLT indirection
2407 when used within the DSO@. Enabling this option can have a dramatic effect
2408 on load and link times of a DSO as it massively reduces the size of the
2409 dynamic export table when the library makes heavy use of templates.
2411 The behavior of this switch is not quite the same as marking the
2412 methods as hidden directly, because it does not affect static variables
2413 local to the function or cause the compiler to deduce that
2414 the function is defined in only one shared object.
2416 You may mark a method as having a visibility explicitly to negate the
2417 effect of the switch for that method. For example, if you do want to
2418 compare pointers to a particular inline method, you might mark it as
2419 having default visibility. Marking the enclosing class with explicit
2420 visibility has no effect.
2422 Explicitly instantiated inline methods are unaffected by this option
2423 as their linkage might otherwise cross a shared library boundary.
2424 @xref{Template Instantiation}.
2426 @item -fvisibility-ms-compat
2427 @opindex fvisibility-ms-compat
2428 This flag attempts to use visibility settings to make GCC's C++
2429 linkage model compatible with that of Microsoft Visual Studio.
2431 The flag makes these changes to GCC's linkage model:
2435 It sets the default visibility to @code{hidden}, like
2436 @option{-fvisibility=hidden}.
2439 Types, but not their members, are not hidden by default.
2442 The One Definition Rule is relaxed for types without explicit
2443 visibility specifications that are defined in more than one
2444 shared object: those declarations are permitted if they are
2445 permitted when this option is not used.
2448 In new code it is better to use @option{-fvisibility=hidden} and
2449 export those classes that are intended to be externally visible.
2450 Unfortunately it is possible for code to rely, perhaps accidentally,
2451 on the Visual Studio behavior.
2453 Among the consequences of these changes are that static data members
2454 of the same type with the same name but defined in different shared
2455 objects are different, so changing one does not change the other;
2456 and that pointers to function members defined in different shared
2457 objects may not compare equal. When this flag is given, it is a
2458 violation of the ODR to define types with the same name differently.
2460 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2461 @opindex fvtable-verify
2462 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2463 feature that verifies at run time, for every virtual call, that
2464 the vtable pointer through which the call is made is valid for the type of
2465 the object, and has not been corrupted or overwritten. If an invalid vtable
2466 pointer is detected at run time, an error is reported and execution of the
2467 program is immediately halted.
2469 This option causes run-time data structures to be built at program startup,
2470 which are used for verifying the vtable pointers.
2471 The options @samp{std} and @samp{preinit}
2472 control the timing of when these data structures are built. In both cases the
2473 data structures are built before execution reaches @code{main}. Using
2474 @option{-fvtable-verify=std} causes the data structures to be built after
2475 shared libraries have been loaded and initialized.
2476 @option{-fvtable-verify=preinit} causes them to be built before shared
2477 libraries have been loaded and initialized.
2479 If this option appears multiple times in the command line with different
2480 values specified, @samp{none} takes highest priority over both @samp{std} and
2481 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2485 When used in conjunction with @option{-fvtable-verify=std} or
2486 @option{-fvtable-verify=preinit}, causes debug versions of the
2487 runtime functions for the vtable verification feature to be called.
2488 This flag also causes the compiler to log information about which
2489 vtable pointers it finds for each class.
2490 This information is written to a file named @file{vtv_set_ptr_data.log}
2491 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2492 if that is defined or the current working directory otherwise.
2494 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2495 file, be sure to delete any existing one.
2498 @opindex fvtv-counts
2499 This is a debugging flag. When used in conjunction with
2500 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2501 causes the compiler to keep track of the total number of virtual calls
2502 it encounters and the number of verifications it inserts. It also
2503 counts the number of calls to certain run-time library functions
2504 that it inserts and logs this information for each compilation unit.
2505 The compiler writes this information to a file named
2506 @file{vtv_count_data.log} in the directory named by the environment
2507 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2508 directory otherwise. It also counts the size of the vtable pointer sets
2509 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2510 in the same directory.
2512 Note: This feature @emph{appends} data to the log files. To get fresh log
2513 files, be sure to delete any existing ones.
2517 Do not use weak symbol support, even if it is provided by the linker.
2518 By default, G++ uses weak symbols if they are available. This
2519 option exists only for testing, and should not be used by end-users;
2520 it results in inferior code and has no benefits. This option may
2521 be removed in a future release of G++.
2525 Do not search for header files in the standard directories specific to
2526 C++, but do still search the other standard directories. (This option
2527 is used when building the C++ library.)
2530 In addition, these optimization, warning, and code generation options
2531 have meanings only for C++ programs:
2534 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2537 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2538 G++ to warn when it generates code that is probably not compatible with the
2539 vendor-neutral C++ ABI@. Since G++ now defaults to
2540 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2541 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2542 or an older compatibility version is selected (with
2543 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2545 Although an effort has been made to warn about
2546 all such cases, there are probably some cases that are not warned about,
2547 even though G++ is generating incompatible code. There may also be
2548 cases where warnings are emitted even though the code that is generated
2551 You should rewrite your code to avoid these warnings if you are
2552 concerned about the fact that code generated by G++ may not be binary
2553 compatible with code generated by other compilers.
2555 @option{-Wabi} can also be used with an explicit version number to
2556 warn about compatibility with a particular @option{-fabi-version}
2557 level, e.g. @option{-Wabi=2} to warn about changes relative to
2558 @option{-fabi-version=2}. Specifying a version number also sets
2559 @option{-fabi-compat-version=@var{n}}.
2561 The known incompatibilities in @option{-fabi-version=2} (which was the
2562 default from GCC 3.4 to 4.9) include:
2567 A template with a non-type template parameter of reference type was
2568 mangled incorrectly:
2571 template <int &> struct S @{@};
2575 This was fixed in @option{-fabi-version=3}.
2578 SIMD vector types declared using @code{__attribute ((vector_size))} were
2579 mangled in a non-standard way that does not allow for overloading of
2580 functions taking vectors of different sizes.
2582 The mangling was changed in @option{-fabi-version=4}.
2585 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2586 qualifiers, and @code{decltype} of a plain declaration was folded away.
2588 These mangling issues were fixed in @option{-fabi-version=5}.
2591 Scoped enumerators passed as arguments to a variadic function are
2592 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2593 On most targets this does not actually affect the parameter passing
2594 ABI, as there is no way to pass an argument smaller than @code{int}.
2596 Also, the ABI changed the mangling of template argument packs,
2597 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2598 a class scope function used as a template argument.
2600 These issues were corrected in @option{-fabi-version=6}.
2603 Lambdas in default argument scope were mangled incorrectly, and the
2604 ABI changed the mangling of @code{nullptr_t}.
2606 These issues were corrected in @option{-fabi-version=7}.
2609 When mangling a function type with function-cv-qualifiers, the
2610 un-qualified function type was incorrectly treated as a substitution
2613 This was fixed in @option{-fabi-version=8}.
2616 It also warns about psABI-related changes. The known psABI changes at this
2622 For SysV/x86-64, unions with @code{long double} members are
2623 passed in memory as specified in psABI. For example:
2633 @code{union U} is always passed in memory.
2637 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2638 @opindex Wctor-dtor-privacy
2639 @opindex Wno-ctor-dtor-privacy
2640 Warn when a class seems unusable because all the constructors or
2641 destructors in that class are private, and it has neither friends nor
2642 public static member functions. Also warn if there are no non-private
2643 methods, and there's at least one private member function that isn't
2644 a constructor or destructor.
2646 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2647 @opindex Wdelete-non-virtual-dtor
2648 @opindex Wno-delete-non-virtual-dtor
2649 Warn when @code{delete} is used to destroy an instance of a class that
2650 has virtual functions and non-virtual destructor. It is unsafe to delete
2651 an instance of a derived class through a pointer to a base class if the
2652 base class does not have a virtual destructor. This warning is enabled
2655 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2656 @opindex Wliteral-suffix
2657 @opindex Wno-literal-suffix
2658 Warn when a string or character literal is followed by a ud-suffix which does
2659 not begin with an underscore. As a conforming extension, GCC treats such
2660 suffixes as separate preprocessing tokens in order to maintain backwards
2661 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2665 #define __STDC_FORMAT_MACROS
2666 #include <inttypes.h>
2671 printf("My int64: %"PRId64"\n", i64);
2675 In this case, @code{PRId64} is treated as a separate preprocessing token.
2677 This warning is enabled by default.
2679 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2681 @opindex Wno-narrowing
2682 Warn when a narrowing conversion prohibited by C++11 occurs within
2686 int i = @{ 2.2 @}; // error: narrowing from double to int
2689 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2691 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses for
2692 non-constants the diagnostic required by the standard. Note that this
2693 does not affect the meaning of well-formed code; narrowing conversions
2694 are still considered ill-formed in SFINAE context.
2696 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2698 @opindex Wno-noexcept
2699 Warn when a noexcept-expression evaluates to false because of a call
2700 to a function that does not have a non-throwing exception
2701 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2702 the compiler to never throw an exception.
2704 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2705 @opindex Wnon-virtual-dtor
2706 @opindex Wno-non-virtual-dtor
2707 Warn when a class has virtual functions and an accessible non-virtual
2708 destructor itself or in an accessible polymorphic base class, in which
2709 case it is possible but unsafe to delete an instance of a derived
2710 class through a pointer to the class itself or base class. This
2711 warning is automatically enabled if @option{-Weffc++} is specified.
2713 @item -Wreorder @r{(C++ and Objective-C++ only)}
2715 @opindex Wno-reorder
2716 @cindex reordering, warning
2717 @cindex warning for reordering of member initializers
2718 Warn when the order of member initializers given in the code does not
2719 match the order in which they must be executed. For instance:
2725 A(): j (0), i (1) @{ @}
2730 The compiler rearranges the member initializers for @code{i}
2731 and @code{j} to match the declaration order of the members, emitting
2732 a warning to that effect. This warning is enabled by @option{-Wall}.
2734 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2735 @opindex fext-numeric-literals
2736 @opindex fno-ext-numeric-literals
2737 Accept imaginary, fixed-point, or machine-defined
2738 literal number suffixes as GNU extensions.
2739 When this option is turned off these suffixes are treated
2740 as C++11 user-defined literal numeric suffixes.
2741 This is on by default for all pre-C++11 dialects and all GNU dialects:
2742 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2743 @option{-std=gnu++14}.
2744 This option is off by default
2745 for ISO C++11 onwards (@option{-std=c++11}, ...).
2748 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2751 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2754 Warn about violations of the following style guidelines from Scott Meyers'
2755 @cite{Effective C++} series of books:
2759 Define a copy constructor and an assignment operator for classes
2760 with dynamically-allocated memory.
2763 Prefer initialization to assignment in constructors.
2766 Have @code{operator=} return a reference to @code{*this}.
2769 Don't try to return a reference when you must return an object.
2772 Distinguish between prefix and postfix forms of increment and
2773 decrement operators.
2776 Never overload @code{&&}, @code{||}, or @code{,}.
2780 This option also enables @option{-Wnon-virtual-dtor}, which is also
2781 one of the effective C++ recommendations. However, the check is
2782 extended to warn about the lack of virtual destructor in accessible
2783 non-polymorphic bases classes too.
2785 When selecting this option, be aware that the standard library
2786 headers do not obey all of these guidelines; use @samp{grep -v}
2787 to filter out those warnings.
2789 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2790 @opindex Wstrict-null-sentinel
2791 @opindex Wno-strict-null-sentinel
2792 Warn about the use of an uncasted @code{NULL} as sentinel. When
2793 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2794 to @code{__null}. Although it is a null pointer constant rather than a
2795 null pointer, it is guaranteed to be of the same size as a pointer.
2796 But this use is not portable across different compilers.
2798 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2799 @opindex Wno-non-template-friend
2800 @opindex Wnon-template-friend
2801 Disable warnings when non-templatized friend functions are declared
2802 within a template. Since the advent of explicit template specification
2803 support in G++, if the name of the friend is an unqualified-id (i.e.,
2804 @samp{friend foo(int)}), the C++ language specification demands that the
2805 friend declare or define an ordinary, nontemplate function. (Section
2806 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2807 could be interpreted as a particular specialization of a templatized
2808 function. Because this non-conforming behavior is no longer the default
2809 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2810 check existing code for potential trouble spots and is on by default.
2811 This new compiler behavior can be turned off with
2812 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2813 but disables the helpful warning.
2815 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2816 @opindex Wold-style-cast
2817 @opindex Wno-old-style-cast
2818 Warn if an old-style (C-style) cast to a non-void type is used within
2819 a C++ program. The new-style casts (@code{dynamic_cast},
2820 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2821 less vulnerable to unintended effects and much easier to search for.
2823 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2824 @opindex Woverloaded-virtual
2825 @opindex Wno-overloaded-virtual
2826 @cindex overloaded virtual function, warning
2827 @cindex warning for overloaded virtual function
2828 Warn when a function declaration hides virtual functions from a
2829 base class. For example, in:
2836 struct B: public A @{
2841 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2852 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2853 @opindex Wno-pmf-conversions
2854 @opindex Wpmf-conversions
2855 Disable the diagnostic for converting a bound pointer to member function
2858 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2859 @opindex Wsign-promo
2860 @opindex Wno-sign-promo
2861 Warn when overload resolution chooses a promotion from unsigned or
2862 enumerated type to a signed type, over a conversion to an unsigned type of
2863 the same size. Previous versions of G++ tried to preserve
2864 unsignedness, but the standard mandates the current behavior.
2867 @node Objective-C and Objective-C++ Dialect Options
2868 @section Options Controlling Objective-C and Objective-C++ Dialects
2870 @cindex compiler options, Objective-C and Objective-C++
2871 @cindex Objective-C and Objective-C++ options, command-line
2872 @cindex options, Objective-C and Objective-C++
2873 (NOTE: This manual does not describe the Objective-C and Objective-C++
2874 languages themselves. @xref{Standards,,Language Standards
2875 Supported by GCC}, for references.)
2877 This section describes the command-line options that are only meaningful
2878 for Objective-C and Objective-C++ programs. You can also use most of
2879 the language-independent GNU compiler options.
2880 For example, you might compile a file @file{some_class.m} like this:
2883 gcc -g -fgnu-runtime -O -c some_class.m
2887 In this example, @option{-fgnu-runtime} is an option meant only for
2888 Objective-C and Objective-C++ programs; you can use the other options with
2889 any language supported by GCC@.
2891 Note that since Objective-C is an extension of the C language, Objective-C
2892 compilations may also use options specific to the C front-end (e.g.,
2893 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2894 C++-specific options (e.g., @option{-Wabi}).
2896 Here is a list of options that are @emph{only} for compiling Objective-C
2897 and Objective-C++ programs:
2900 @item -fconstant-string-class=@var{class-name}
2901 @opindex fconstant-string-class
2902 Use @var{class-name} as the name of the class to instantiate for each
2903 literal string specified with the syntax @code{@@"@dots{}"}. The default
2904 class name is @code{NXConstantString} if the GNU runtime is being used, and
2905 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2906 @option{-fconstant-cfstrings} option, if also present, overrides the
2907 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2908 to be laid out as constant CoreFoundation strings.
2911 @opindex fgnu-runtime
2912 Generate object code compatible with the standard GNU Objective-C
2913 runtime. This is the default for most types of systems.
2915 @item -fnext-runtime
2916 @opindex fnext-runtime
2917 Generate output compatible with the NeXT runtime. This is the default
2918 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2919 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2922 @item -fno-nil-receivers
2923 @opindex fno-nil-receivers
2924 Assume that all Objective-C message dispatches (@code{[receiver
2925 message:arg]}) in this translation unit ensure that the receiver is
2926 not @code{nil}. This allows for more efficient entry points in the
2927 runtime to be used. This option is only available in conjunction with
2928 the NeXT runtime and ABI version 0 or 1.
2930 @item -fobjc-abi-version=@var{n}
2931 @opindex fobjc-abi-version
2932 Use version @var{n} of the Objective-C ABI for the selected runtime.
2933 This option is currently supported only for the NeXT runtime. In that
2934 case, Version 0 is the traditional (32-bit) ABI without support for
2935 properties and other Objective-C 2.0 additions. Version 1 is the
2936 traditional (32-bit) ABI with support for properties and other
2937 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2938 nothing is specified, the default is Version 0 on 32-bit target
2939 machines, and Version 2 on 64-bit target machines.
2941 @item -fobjc-call-cxx-cdtors
2942 @opindex fobjc-call-cxx-cdtors
2943 For each Objective-C class, check if any of its instance variables is a
2944 C++ object with a non-trivial default constructor. If so, synthesize a
2945 special @code{- (id) .cxx_construct} instance method which runs
2946 non-trivial default constructors on any such instance variables, in order,
2947 and then return @code{self}. Similarly, check if any instance variable
2948 is a C++ object with a non-trivial destructor, and if so, synthesize a
2949 special @code{- (void) .cxx_destruct} method which runs
2950 all such default destructors, in reverse order.
2952 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2953 methods thusly generated only operate on instance variables
2954 declared in the current Objective-C class, and not those inherited
2955 from superclasses. It is the responsibility of the Objective-C
2956 runtime to invoke all such methods in an object's inheritance
2957 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2958 by the runtime immediately after a new object instance is allocated;
2959 the @code{- (void) .cxx_destruct} methods are invoked immediately
2960 before the runtime deallocates an object instance.
2962 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2963 support for invoking the @code{- (id) .cxx_construct} and
2964 @code{- (void) .cxx_destruct} methods.
2966 @item -fobjc-direct-dispatch
2967 @opindex fobjc-direct-dispatch
2968 Allow fast jumps to the message dispatcher. On Darwin this is
2969 accomplished via the comm page.
2971 @item -fobjc-exceptions
2972 @opindex fobjc-exceptions
2973 Enable syntactic support for structured exception handling in
2974 Objective-C, similar to what is offered by C++ and Java. This option
2975 is required to use the Objective-C keywords @code{@@try},
2976 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2977 @code{@@synchronized}. This option is available with both the GNU
2978 runtime and the NeXT runtime (but not available in conjunction with
2979 the NeXT runtime on Mac OS X 10.2 and earlier).
2983 Enable garbage collection (GC) in Objective-C and Objective-C++
2984 programs. This option is only available with the NeXT runtime; the
2985 GNU runtime has a different garbage collection implementation that
2986 does not require special compiler flags.
2988 @item -fobjc-nilcheck
2989 @opindex fobjc-nilcheck
2990 For the NeXT runtime with version 2 of the ABI, check for a nil
2991 receiver in method invocations before doing the actual method call.
2992 This is the default and can be disabled using
2993 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2994 checked for nil in this way no matter what this flag is set to.
2995 Currently this flag does nothing when the GNU runtime, or an older
2996 version of the NeXT runtime ABI, is used.
2998 @item -fobjc-std=objc1
3000 Conform to the language syntax of Objective-C 1.0, the language
3001 recognized by GCC 4.0. This only affects the Objective-C additions to
3002 the C/C++ language; it does not affect conformance to C/C++ standards,
3003 which is controlled by the separate C/C++ dialect option flags. When
3004 this option is used with the Objective-C or Objective-C++ compiler,
3005 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3006 This is useful if you need to make sure that your Objective-C code can
3007 be compiled with older versions of GCC@.
3009 @item -freplace-objc-classes
3010 @opindex freplace-objc-classes
3011 Emit a special marker instructing @command{ld(1)} not to statically link in
3012 the resulting object file, and allow @command{dyld(1)} to load it in at
3013 run time instead. This is used in conjunction with the Fix-and-Continue
3014 debugging mode, where the object file in question may be recompiled and
3015 dynamically reloaded in the course of program execution, without the need
3016 to restart the program itself. Currently, Fix-and-Continue functionality
3017 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3022 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3023 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3024 compile time) with static class references that get initialized at load time,
3025 which improves run-time performance. Specifying the @option{-fzero-link} flag
3026 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3027 to be retained. This is useful in Zero-Link debugging mode, since it allows
3028 for individual class implementations to be modified during program execution.
3029 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3030 regardless of command-line options.
3032 @item -fno-local-ivars
3033 @opindex fno-local-ivars
3034 @opindex flocal-ivars
3035 By default instance variables in Objective-C can be accessed as if
3036 they were local variables from within the methods of the class they're
3037 declared in. This can lead to shadowing between instance variables
3038 and other variables declared either locally inside a class method or
3039 globally with the same name. Specifying the @option{-fno-local-ivars}
3040 flag disables this behavior thus avoiding variable shadowing issues.
3042 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3043 @opindex fivar-visibility
3044 Set the default instance variable visibility to the specified option
3045 so that instance variables declared outside the scope of any access
3046 modifier directives default to the specified visibility.
3050 Dump interface declarations for all classes seen in the source file to a
3051 file named @file{@var{sourcename}.decl}.
3053 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3054 @opindex Wassign-intercept
3055 @opindex Wno-assign-intercept
3056 Warn whenever an Objective-C assignment is being intercepted by the
3059 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3060 @opindex Wno-protocol
3062 If a class is declared to implement a protocol, a warning is issued for
3063 every method in the protocol that is not implemented by the class. The
3064 default behavior is to issue a warning for every method not explicitly
3065 implemented in the class, even if a method implementation is inherited
3066 from the superclass. If you use the @option{-Wno-protocol} option, then
3067 methods inherited from the superclass are considered to be implemented,
3068 and no warning is issued for them.
3070 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3072 @opindex Wno-selector
3073 Warn if multiple methods of different types for the same selector are
3074 found during compilation. The check is performed on the list of methods
3075 in the final stage of compilation. Additionally, a check is performed
3076 for each selector appearing in a @code{@@selector(@dots{})}
3077 expression, and a corresponding method for that selector has been found
3078 during compilation. Because these checks scan the method table only at
3079 the end of compilation, these warnings are not produced if the final
3080 stage of compilation is not reached, for example because an error is
3081 found during compilation, or because the @option{-fsyntax-only} option is
3084 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3085 @opindex Wstrict-selector-match
3086 @opindex Wno-strict-selector-match
3087 Warn if multiple methods with differing argument and/or return types are
3088 found for a given selector when attempting to send a message using this
3089 selector to a receiver of type @code{id} or @code{Class}. When this flag
3090 is off (which is the default behavior), the compiler omits such warnings
3091 if any differences found are confined to types that share the same size
3094 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3095 @opindex Wundeclared-selector
3096 @opindex Wno-undeclared-selector
3097 Warn if a @code{@@selector(@dots{})} expression referring to an
3098 undeclared selector is found. A selector is considered undeclared if no
3099 method with that name has been declared before the
3100 @code{@@selector(@dots{})} expression, either explicitly in an
3101 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3102 an @code{@@implementation} section. This option always performs its
3103 checks as soon as a @code{@@selector(@dots{})} expression is found,
3104 while @option{-Wselector} only performs its checks in the final stage of
3105 compilation. This also enforces the coding style convention
3106 that methods and selectors must be declared before being used.
3108 @item -print-objc-runtime-info
3109 @opindex print-objc-runtime-info
3110 Generate C header describing the largest structure that is passed by
3115 @node Language Independent Options
3116 @section Options to Control Diagnostic Messages Formatting
3117 @cindex options to control diagnostics formatting
3118 @cindex diagnostic messages
3119 @cindex message formatting
3121 Traditionally, diagnostic messages have been formatted irrespective of
3122 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3123 options described below
3124 to control the formatting algorithm for diagnostic messages,
3125 e.g.@: how many characters per line, how often source location
3126 information should be reported. Note that some language front ends may not
3127 honor these options.
3130 @item -fmessage-length=@var{n}
3131 @opindex fmessage-length
3132 Try to format error messages so that they fit on lines of about
3133 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3134 done; each error message appears on a single line. This is the
3135 default for all front ends.
3137 @item -fdiagnostics-show-location=once
3138 @opindex fdiagnostics-show-location
3139 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3140 reporter to emit source location information @emph{once}; that is, in
3141 case the message is too long to fit on a single physical line and has to
3142 be wrapped, the source location won't be emitted (as prefix) again,
3143 over and over, in subsequent continuation lines. This is the default
3146 @item -fdiagnostics-show-location=every-line
3147 Only meaningful in line-wrapping mode. Instructs the diagnostic
3148 messages reporter to emit the same source location information (as
3149 prefix) for physical lines that result from the process of breaking
3150 a message which is too long to fit on a single line.
3152 @item -fdiagnostics-color[=@var{WHEN}]
3153 @itemx -fno-diagnostics-color
3154 @opindex fdiagnostics-color
3155 @cindex highlight, color, colour
3156 @vindex GCC_COLORS @r{environment variable}
3157 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3158 or @samp{auto}. The default depends on how the compiler has been configured,
3159 it can be any of the above @var{WHEN} options or also @samp{never}
3160 if @env{GCC_COLORS} environment variable isn't present in the environment,
3161 and @samp{auto} otherwise.
3162 @samp{auto} means to use color only when the standard error is a terminal.
3163 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3164 aliases for @option{-fdiagnostics-color=always} and
3165 @option{-fdiagnostics-color=never}, respectively.
3167 The colors are defined by the environment variable @env{GCC_COLORS}.
3168 Its value is a colon-separated list of capabilities and Select Graphic
3169 Rendition (SGR) substrings. SGR commands are interpreted by the
3170 terminal or terminal emulator. (See the section in the documentation
3171 of your text terminal for permitted values and their meanings as
3172 character attributes.) These substring values are integers in decimal
3173 representation and can be concatenated with semicolons.
3174 Common values to concatenate include
3176 @samp{4} for underline,
3178 @samp{7} for inverse,
3179 @samp{39} for default foreground color,
3180 @samp{30} to @samp{37} for foreground colors,
3181 @samp{90} to @samp{97} for 16-color mode foreground colors,
3182 @samp{38;5;0} to @samp{38;5;255}
3183 for 88-color and 256-color modes foreground colors,
3184 @samp{49} for default background color,
3185 @samp{40} to @samp{47} for background colors,
3186 @samp{100} to @samp{107} for 16-color mode background colors,
3187 and @samp{48;5;0} to @samp{48;5;255}
3188 for 88-color and 256-color modes background colors.
3190 The default @env{GCC_COLORS} is
3192 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3195 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3196 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3197 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3198 string disables colors.
3199 Supported capabilities are as follows.
3203 @vindex error GCC_COLORS @r{capability}
3204 SGR substring for error: markers.
3207 @vindex warning GCC_COLORS @r{capability}
3208 SGR substring for warning: markers.
3211 @vindex note GCC_COLORS @r{capability}
3212 SGR substring for note: markers.
3215 @vindex caret GCC_COLORS @r{capability}
3216 SGR substring for caret line.
3219 @vindex locus GCC_COLORS @r{capability}
3220 SGR substring for location information, @samp{file:line} or
3221 @samp{file:line:column} etc.
3224 @vindex quote GCC_COLORS @r{capability}
3225 SGR substring for information printed within quotes.
3228 @item -fno-diagnostics-show-option
3229 @opindex fno-diagnostics-show-option
3230 @opindex fdiagnostics-show-option
3231 By default, each diagnostic emitted includes text indicating the
3232 command-line option that directly controls the diagnostic (if such an
3233 option is known to the diagnostic machinery). Specifying the
3234 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3236 @item -fno-diagnostics-show-caret
3237 @opindex fno-diagnostics-show-caret
3238 @opindex fdiagnostics-show-caret
3239 By default, each diagnostic emitted includes the original source line
3240 and a caret '^' indicating the column. This option suppresses this
3241 information. The source line is truncated to @var{n} characters, if
3242 the @option{-fmessage-length=n} is given. When the output is done
3243 to the terminal, the width is limited to the width given by the
3244 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3248 @node Warning Options
3249 @section Options to Request or Suppress Warnings
3250 @cindex options to control warnings
3251 @cindex warning messages
3252 @cindex messages, warning
3253 @cindex suppressing warnings
3255 Warnings are diagnostic messages that report constructions that
3256 are not inherently erroneous but that are risky or suggest there
3257 may have been an error.
3259 The following language-independent options do not enable specific
3260 warnings but control the kinds of diagnostics produced by GCC@.
3263 @cindex syntax checking
3265 @opindex fsyntax-only
3266 Check the code for syntax errors, but don't do anything beyond that.
3268 @item -fmax-errors=@var{n}
3269 @opindex fmax-errors
3270 Limits the maximum number of error messages to @var{n}, at which point
3271 GCC bails out rather than attempting to continue processing the source
3272 code. If @var{n} is 0 (the default), there is no limit on the number
3273 of error messages produced. If @option{-Wfatal-errors} is also
3274 specified, then @option{-Wfatal-errors} takes precedence over this
3279 Inhibit all warning messages.
3284 Make all warnings into errors.
3289 Make the specified warning into an error. The specifier for a warning
3290 is appended; for example @option{-Werror=switch} turns the warnings
3291 controlled by @option{-Wswitch} into errors. This switch takes a
3292 negative form, to be used to negate @option{-Werror} for specific
3293 warnings; for example @option{-Wno-error=switch} makes
3294 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3297 The warning message for each controllable warning includes the
3298 option that controls the warning. That option can then be used with
3299 @option{-Werror=} and @option{-Wno-error=} as described above.
3300 (Printing of the option in the warning message can be disabled using the
3301 @option{-fno-diagnostics-show-option} flag.)
3303 Note that specifying @option{-Werror=}@var{foo} automatically implies
3304 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3307 @item -Wfatal-errors
3308 @opindex Wfatal-errors
3309 @opindex Wno-fatal-errors
3310 This option causes the compiler to abort compilation on the first error
3311 occurred rather than trying to keep going and printing further error
3316 You can request many specific warnings with options beginning with
3317 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3318 implicit declarations. Each of these specific warning options also
3319 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3320 example, @option{-Wno-implicit}. This manual lists only one of the
3321 two forms, whichever is not the default. For further
3322 language-specific options also refer to @ref{C++ Dialect Options} and
3323 @ref{Objective-C and Objective-C++ Dialect Options}.
3325 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3326 options, such as @option{-Wunused}, which may turn on further options,
3327 such as @option{-Wunused-value}. The combined effect of positive and
3328 negative forms is that more specific options have priority over less
3329 specific ones, independently of their position in the command-line. For
3330 options of the same specificity, the last one takes effect. Options
3331 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3332 as if they appeared at the end of the command-line.
3334 When an unrecognized warning option is requested (e.g.,
3335 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3336 that the option is not recognized. However, if the @option{-Wno-} form
3337 is used, the behavior is slightly different: no diagnostic is
3338 produced for @option{-Wno-unknown-warning} unless other diagnostics
3339 are being produced. This allows the use of new @option{-Wno-} options
3340 with old compilers, but if something goes wrong, the compiler
3341 warns that an unrecognized option is present.
3348 Issue all the warnings demanded by strict ISO C and ISO C++;
3349 reject all programs that use forbidden extensions, and some other
3350 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3351 version of the ISO C standard specified by any @option{-std} option used.
3353 Valid ISO C and ISO C++ programs should compile properly with or without
3354 this option (though a rare few require @option{-ansi} or a
3355 @option{-std} option specifying the required version of ISO C)@. However,
3356 without this option, certain GNU extensions and traditional C and C++
3357 features are supported as well. With this option, they are rejected.
3359 @option{-Wpedantic} does not cause warning messages for use of the
3360 alternate keywords whose names begin and end with @samp{__}. Pedantic
3361 warnings are also disabled in the expression that follows
3362 @code{__extension__}. However, only system header files should use
3363 these escape routes; application programs should avoid them.
3364 @xref{Alternate Keywords}.
3366 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3367 C conformance. They soon find that it does not do quite what they want:
3368 it finds some non-ISO practices, but not all---only those for which
3369 ISO C @emph{requires} a diagnostic, and some others for which
3370 diagnostics have been added.
3372 A feature to report any failure to conform to ISO C might be useful in
3373 some instances, but would require considerable additional work and would
3374 be quite different from @option{-Wpedantic}. We don't have plans to
3375 support such a feature in the near future.
3377 Where the standard specified with @option{-std} represents a GNU
3378 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3379 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3380 extended dialect is based. Warnings from @option{-Wpedantic} are given
3381 where they are required by the base standard. (It does not make sense
3382 for such warnings to be given only for features not in the specified GNU
3383 C dialect, since by definition the GNU dialects of C include all
3384 features the compiler supports with the given option, and there would be
3385 nothing to warn about.)
3387 @item -pedantic-errors
3388 @opindex pedantic-errors
3389 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3390 requires a diagnostic, in some cases where there is undefined behavior
3391 at compile-time and in some other cases that do not prevent compilation
3392 of programs that are valid according to the standard. This is not
3393 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3394 by this option and not enabled by the latter and vice versa.
3399 This enables all the warnings about constructions that some users
3400 consider questionable, and that are easy to avoid (or modify to
3401 prevent the warning), even in conjunction with macros. This also
3402 enables some language-specific warnings described in @ref{C++ Dialect
3403 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3405 @option{-Wall} turns on the following warning flags:
3407 @gccoptlist{-Waddress @gol
3408 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3409 -Wc++11-compat -Wc++14-compat@gol
3410 -Wchar-subscripts @gol
3411 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3412 -Wimplicit-int @r{(C and Objective-C only)} @gol
3413 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3416 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3417 -Wmaybe-uninitialized @gol
3418 -Wmissing-braces @r{(only for C/ObjC)} @gol
3425 -Wsequence-point @gol
3426 -Wsign-compare @r{(only in C++)} @gol
3427 -Wstrict-aliasing @gol
3428 -Wstrict-overflow=1 @gol
3431 -Wuninitialized @gol
3432 -Wunknown-pragmas @gol
3433 -Wunused-function @gol
3436 -Wunused-variable @gol
3437 -Wvolatile-register-var @gol
3440 Note that some warning flags are not implied by @option{-Wall}. Some of
3441 them warn about constructions that users generally do not consider
3442 questionable, but which occasionally you might wish to check for;
3443 others warn about constructions that are necessary or hard to avoid in
3444 some cases, and there is no simple way to modify the code to suppress
3445 the warning. Some of them are enabled by @option{-Wextra} but many of
3446 them must be enabled individually.
3452 This enables some extra warning flags that are not enabled by
3453 @option{-Wall}. (This option used to be called @option{-W}. The older
3454 name is still supported, but the newer name is more descriptive.)
3456 @gccoptlist{-Wclobbered @gol
3458 -Wignored-qualifiers @gol
3459 -Wmissing-field-initializers @gol
3460 -Wmissing-parameter-type @r{(C only)} @gol
3461 -Wold-style-declaration @r{(C only)} @gol
3462 -Woverride-init @gol
3465 -Wuninitialized @gol
3466 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3467 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3470 The option @option{-Wextra} also prints warning messages for the
3476 A pointer is compared against integer zero with @code{<}, @code{<=},
3477 @code{>}, or @code{>=}.
3480 (C++ only) An enumerator and a non-enumerator both appear in a
3481 conditional expression.
3484 (C++ only) Ambiguous virtual bases.
3487 (C++ only) Subscripting an array that has been declared @code{register}.
3490 (C++ only) Taking the address of a variable that has been declared
3494 (C++ only) A base class is not initialized in a derived class's copy
3499 @item -Wchar-subscripts
3500 @opindex Wchar-subscripts
3501 @opindex Wno-char-subscripts
3502 Warn if an array subscript has type @code{char}. This is a common cause
3503 of error, as programmers often forget that this type is signed on some
3505 This warning is enabled by @option{-Wall}.
3509 @opindex Wno-comment
3510 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3511 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3512 This warning is enabled by @option{-Wall}.
3514 @item -Wno-coverage-mismatch
3515 @opindex Wno-coverage-mismatch
3516 Warn if feedback profiles do not match when using the
3517 @option{-fprofile-use} option.
3518 If a source file is changed between compiling with @option{-fprofile-gen} and
3519 with @option{-fprofile-use}, the files with the profile feedback can fail
3520 to match the source file and GCC cannot use the profile feedback
3521 information. By default, this warning is enabled and is treated as an
3522 error. @option{-Wno-coverage-mismatch} can be used to disable the
3523 warning or @option{-Wno-error=coverage-mismatch} can be used to
3524 disable the error. Disabling the error for this warning can result in
3525 poorly optimized code and is useful only in the
3526 case of very minor changes such as bug fixes to an existing code-base.
3527 Completely disabling the warning is not recommended.
3530 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3532 Suppress warning messages emitted by @code{#warning} directives.
3534 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3535 @opindex Wdouble-promotion
3536 @opindex Wno-double-promotion
3537 Give a warning when a value of type @code{float} is implicitly
3538 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3539 floating-point unit implement @code{float} in hardware, but emulate
3540 @code{double} in software. On such a machine, doing computations
3541 using @code{double} values is much more expensive because of the
3542 overhead required for software emulation.
3544 It is easy to accidentally do computations with @code{double} because
3545 floating-point literals are implicitly of type @code{double}. For
3549 float area(float radius)
3551 return 3.14159 * radius * radius;
3555 the compiler performs the entire computation with @code{double}
3556 because the floating-point literal is a @code{double}.
3559 @itemx -Wformat=@var{n}
3562 @opindex ffreestanding
3563 @opindex fno-builtin
3565 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3566 the arguments supplied have types appropriate to the format string
3567 specified, and that the conversions specified in the format string make
3568 sense. This includes standard functions, and others specified by format
3569 attributes (@pxref{Function Attributes}), in the @code{printf},
3570 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3571 not in the C standard) families (or other target-specific families).
3572 Which functions are checked without format attributes having been
3573 specified depends on the standard version selected, and such checks of
3574 functions without the attribute specified are disabled by
3575 @option{-ffreestanding} or @option{-fno-builtin}.
3577 The formats are checked against the format features supported by GNU
3578 libc version 2.2. These include all ISO C90 and C99 features, as well
3579 as features from the Single Unix Specification and some BSD and GNU
3580 extensions. Other library implementations may not support all these
3581 features; GCC does not support warning about features that go beyond a
3582 particular library's limitations. However, if @option{-Wpedantic} is used
3583 with @option{-Wformat}, warnings are given about format features not
3584 in the selected standard version (but not for @code{strfmon} formats,
3585 since those are not in any version of the C standard). @xref{C Dialect
3586 Options,,Options Controlling C Dialect}.
3593 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3594 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3595 @option{-Wformat} also checks for null format arguments for several
3596 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3597 aspects of this level of format checking can be disabled by the
3598 options: @option{-Wno-format-contains-nul},
3599 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3600 @option{-Wformat} is enabled by @option{-Wall}.
3602 @item -Wno-format-contains-nul
3603 @opindex Wno-format-contains-nul
3604 @opindex Wformat-contains-nul
3605 If @option{-Wformat} is specified, do not warn about format strings that
3608 @item -Wno-format-extra-args
3609 @opindex Wno-format-extra-args
3610 @opindex Wformat-extra-args
3611 If @option{-Wformat} is specified, do not warn about excess arguments to a
3612 @code{printf} or @code{scanf} format function. The C standard specifies
3613 that such arguments are ignored.
3615 Where the unused arguments lie between used arguments that are
3616 specified with @samp{$} operand number specifications, normally
3617 warnings are still given, since the implementation could not know what
3618 type to pass to @code{va_arg} to skip the unused arguments. However,
3619 in the case of @code{scanf} formats, this option suppresses the
3620 warning if the unused arguments are all pointers, since the Single
3621 Unix Specification says that such unused arguments are allowed.
3623 @item -Wno-format-zero-length
3624 @opindex Wno-format-zero-length
3625 @opindex Wformat-zero-length
3626 If @option{-Wformat} is specified, do not warn about zero-length formats.
3627 The C standard specifies that zero-length formats are allowed.
3632 Enable @option{-Wformat} plus additional format checks. Currently
3633 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3636 @item -Wformat-nonliteral
3637 @opindex Wformat-nonliteral
3638 @opindex Wno-format-nonliteral
3639 If @option{-Wformat} is specified, also warn if the format string is not a
3640 string literal and so cannot be checked, unless the format function
3641 takes its format arguments as a @code{va_list}.
3643 @item -Wformat-security
3644 @opindex Wformat-security
3645 @opindex Wno-format-security
3646 If @option{-Wformat} is specified, also warn about uses of format
3647 functions that represent possible security problems. At present, this
3648 warns about calls to @code{printf} and @code{scanf} functions where the
3649 format string is not a string literal and there are no format arguments,
3650 as in @code{printf (foo);}. This may be a security hole if the format
3651 string came from untrusted input and contains @samp{%n}. (This is
3652 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3653 in future warnings may be added to @option{-Wformat-security} that are not
3654 included in @option{-Wformat-nonliteral}.)
3656 @item -Wformat-signedness
3657 @opindex Wformat-signedness
3658 @opindex Wno-format-signedness
3659 If @option{-Wformat} is specified, also warn if the format string
3660 requires an unsigned argument and the argument is signed and vice versa.
3663 @opindex Wformat-y2k
3664 @opindex Wno-format-y2k
3665 If @option{-Wformat} is specified, also warn about @code{strftime}
3666 formats that may yield only a two-digit year.
3671 @opindex Wno-nonnull
3672 Warn about passing a null pointer for arguments marked as
3673 requiring a non-null value by the @code{nonnull} function attribute.
3675 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3676 can be disabled with the @option{-Wno-nonnull} option.
3678 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3680 @opindex Wno-init-self
3681 Warn about uninitialized variables that are initialized with themselves.
3682 Note this option can only be used with the @option{-Wuninitialized} option.
3684 For example, GCC warns about @code{i} being uninitialized in the
3685 following snippet only when @option{-Winit-self} has been specified:
3696 This warning is enabled by @option{-Wall} in C++.
3698 @item -Wimplicit-int @r{(C and Objective-C only)}
3699 @opindex Wimplicit-int
3700 @opindex Wno-implicit-int
3701 Warn when a declaration does not specify a type.
3702 This warning is enabled by @option{-Wall}.
3704 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3705 @opindex Wimplicit-function-declaration
3706 @opindex Wno-implicit-function-declaration
3707 Give a warning whenever a function is used before being declared. In
3708 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3709 enabled by default and it is made into an error by
3710 @option{-pedantic-errors}. This warning is also enabled by
3713 @item -Wimplicit @r{(C and Objective-C only)}
3715 @opindex Wno-implicit
3716 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3717 This warning is enabled by @option{-Wall}.
3719 @item -Wignored-qualifiers @r{(C and C++ only)}
3720 @opindex Wignored-qualifiers
3721 @opindex Wno-ignored-qualifiers
3722 Warn if the return type of a function has a type qualifier
3723 such as @code{const}. For ISO C such a type qualifier has no effect,
3724 since the value returned by a function is not an lvalue.
3725 For C++, the warning is only emitted for scalar types or @code{void}.
3726 ISO C prohibits qualified @code{void} return types on function
3727 definitions, so such return types always receive a warning
3728 even without this option.
3730 This warning is also enabled by @option{-Wextra}.
3735 Warn if the type of @code{main} is suspicious. @code{main} should be
3736 a function with external linkage, returning int, taking either zero
3737 arguments, two, or three arguments of appropriate types. This warning
3738 is enabled by default in C++ and is enabled by either @option{-Wall}
3739 or @option{-Wpedantic}.
3741 @item -Wmissing-braces
3742 @opindex Wmissing-braces
3743 @opindex Wno-missing-braces
3744 Warn if an aggregate or union initializer is not fully bracketed. In
3745 the following example, the initializer for @code{a} is not fully
3746 bracketed, but that for @code{b} is fully bracketed. This warning is
3747 enabled by @option{-Wall} in C.
3750 int a[2][2] = @{ 0, 1, 2, 3 @};
3751 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3754 This warning is enabled by @option{-Wall}.
3756 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3757 @opindex Wmissing-include-dirs
3758 @opindex Wno-missing-include-dirs
3759 Warn if a user-supplied include directory does not exist.
3762 @opindex Wparentheses
3763 @opindex Wno-parentheses
3764 Warn if parentheses are omitted in certain contexts, such
3765 as when there is an assignment in a context where a truth value
3766 is expected, or when operators are nested whose precedence people
3767 often get confused about.
3769 Also warn if a comparison like @code{x<=y<=z} appears; this is
3770 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3771 interpretation from that of ordinary mathematical notation.
3773 Also warn about constructions where there may be confusion to which
3774 @code{if} statement an @code{else} branch belongs. Here is an example of
3789 In C/C++, every @code{else} branch belongs to the innermost possible
3790 @code{if} statement, which in this example is @code{if (b)}. This is
3791 often not what the programmer expected, as illustrated in the above
3792 example by indentation the programmer chose. When there is the
3793 potential for this confusion, GCC issues a warning when this flag
3794 is specified. To eliminate the warning, add explicit braces around
3795 the innermost @code{if} statement so there is no way the @code{else}
3796 can belong to the enclosing @code{if}. The resulting code
3813 Also warn for dangerous uses of the GNU extension to
3814 @code{?:} with omitted middle operand. When the condition
3815 in the @code{?}: operator is a boolean expression, the omitted value is
3816 always 1. Often programmers expect it to be a value computed
3817 inside the conditional expression instead.
3819 This warning is enabled by @option{-Wall}.
3821 @item -Wsequence-point
3822 @opindex Wsequence-point
3823 @opindex Wno-sequence-point
3824 Warn about code that may have undefined semantics because of violations
3825 of sequence point rules in the C and C++ standards.
3827 The C and C++ standards define the order in which expressions in a C/C++
3828 program are evaluated in terms of @dfn{sequence points}, which represent
3829 a partial ordering between the execution of parts of the program: those
3830 executed before the sequence point, and those executed after it. These
3831 occur after the evaluation of a full expression (one which is not part
3832 of a larger expression), after the evaluation of the first operand of a
3833 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3834 function is called (but after the evaluation of its arguments and the
3835 expression denoting the called function), and in certain other places.
3836 Other than as expressed by the sequence point rules, the order of
3837 evaluation of subexpressions of an expression is not specified. All
3838 these rules describe only a partial order rather than a total order,
3839 since, for example, if two functions are called within one expression
3840 with no sequence point between them, the order in which the functions
3841 are called is not specified. However, the standards committee have
3842 ruled that function calls do not overlap.
3844 It is not specified when between sequence points modifications to the
3845 values of objects take effect. Programs whose behavior depends on this
3846 have undefined behavior; the C and C++ standards specify that ``Between
3847 the previous and next sequence point an object shall have its stored
3848 value modified at most once by the evaluation of an expression.
3849 Furthermore, the prior value shall be read only to determine the value
3850 to be stored.''. If a program breaks these rules, the results on any
3851 particular implementation are entirely unpredictable.
3853 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3854 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3855 diagnosed by this option, and it may give an occasional false positive
3856 result, but in general it has been found fairly effective at detecting
3857 this sort of problem in programs.
3859 The standard is worded confusingly, therefore there is some debate
3860 over the precise meaning of the sequence point rules in subtle cases.
3861 Links to discussions of the problem, including proposed formal
3862 definitions, may be found on the GCC readings page, at
3863 @uref{http://gcc.gnu.org/@/readings.html}.
3865 This warning is enabled by @option{-Wall} for C and C++.
3867 @item -Wno-return-local-addr
3868 @opindex Wno-return-local-addr
3869 @opindex Wreturn-local-addr
3870 Do not warn about returning a pointer (or in C++, a reference) to a
3871 variable that goes out of scope after the function returns.
3874 @opindex Wreturn-type
3875 @opindex Wno-return-type
3876 Warn whenever a function is defined with a return type that defaults
3877 to @code{int}. Also warn about any @code{return} statement with no
3878 return value in a function whose return type is not @code{void}
3879 (falling off the end of the function body is considered returning
3880 without a value), and about a @code{return} statement with an
3881 expression in a function whose return type is @code{void}.
3883 For C++, a function without return type always produces a diagnostic
3884 message, even when @option{-Wno-return-type} is specified. The only
3885 exceptions are @code{main} and functions defined in system headers.
3887 This warning is enabled by @option{-Wall}.
3889 @item -Wshift-count-negative
3890 @opindex Wshift-count-negative
3891 @opindex Wno-shift-count-negative
3892 Warn if shift count is negative. This warning is enabled by default.
3894 @item -Wshift-count-overflow
3895 @opindex Wshift-count-overflow
3896 @opindex Wno-shift-count-overflow
3897 Warn if shift count >= width of type. This warning is enabled by default.
3902 Warn whenever a @code{switch} statement has an index of enumerated type
3903 and lacks a @code{case} for one or more of the named codes of that
3904 enumeration. (The presence of a @code{default} label prevents this
3905 warning.) @code{case} labels outside the enumeration range also
3906 provoke warnings when this option is used (even if there is a
3907 @code{default} label).
3908 This warning is enabled by @option{-Wall}.
3910 @item -Wswitch-default
3911 @opindex Wswitch-default
3912 @opindex Wno-switch-default
3913 Warn whenever a @code{switch} statement does not have a @code{default}
3917 @opindex Wswitch-enum
3918 @opindex Wno-switch-enum
3919 Warn whenever a @code{switch} statement has an index of enumerated type
3920 and lacks a @code{case} for one or more of the named codes of that
3921 enumeration. @code{case} labels outside the enumeration range also
3922 provoke warnings when this option is used. The only difference
3923 between @option{-Wswitch} and this option is that this option gives a
3924 warning about an omitted enumeration code even if there is a
3925 @code{default} label.
3928 @opindex Wswitch-bool
3929 @opindex Wno-switch-bool
3930 Warn whenever a @code{switch} statement has an index of boolean type.
3931 It is possible to suppress this warning by casting the controlling
3932 expression to a type other than @code{bool}. For example:
3935 switch ((int) (a == 4))
3941 This warning is enabled by default for C and C++ programs.
3943 @item -Wsync-nand @r{(C and C++ only)}
3945 @opindex Wno-sync-nand
3946 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3947 built-in functions are used. These functions changed semantics in GCC 4.4.
3951 @opindex Wno-trigraphs
3952 Warn if any trigraphs are encountered that might change the meaning of
3953 the program (trigraphs within comments are not warned about).
3954 This warning is enabled by @option{-Wall}.
3956 @item -Wunused-but-set-parameter
3957 @opindex Wunused-but-set-parameter
3958 @opindex Wno-unused-but-set-parameter
3959 Warn whenever a function parameter is assigned to, but otherwise unused
3960 (aside from its declaration).
3962 To suppress this warning use the @code{unused} attribute
3963 (@pxref{Variable Attributes}).
3965 This warning is also enabled by @option{-Wunused} together with
3968 @item -Wunused-but-set-variable
3969 @opindex Wunused-but-set-variable
3970 @opindex Wno-unused-but-set-variable
3971 Warn whenever a local variable is assigned to, but otherwise unused
3972 (aside from its declaration).
3973 This warning is enabled by @option{-Wall}.
3975 To suppress this warning use the @code{unused} attribute
3976 (@pxref{Variable Attributes}).
3978 This warning is also enabled by @option{-Wunused}, which is enabled
3981 @item -Wunused-function
3982 @opindex Wunused-function
3983 @opindex Wno-unused-function
3984 Warn whenever a static function is declared but not defined or a
3985 non-inline static function is unused.
3986 This warning is enabled by @option{-Wall}.
3988 @item -Wunused-label
3989 @opindex Wunused-label
3990 @opindex Wno-unused-label
3991 Warn whenever a label is declared but not used.
3992 This warning is enabled by @option{-Wall}.
3994 To suppress this warning use the @code{unused} attribute
3995 (@pxref{Variable Attributes}).
3997 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3998 @opindex Wunused-local-typedefs
3999 Warn when a typedef locally defined in a function is not used.
4000 This warning is enabled by @option{-Wall}.
4002 @item -Wunused-parameter
4003 @opindex Wunused-parameter
4004 @opindex Wno-unused-parameter
4005 Warn whenever a function parameter is unused aside from its declaration.
4007 To suppress this warning use the @code{unused} attribute
4008 (@pxref{Variable Attributes}).
4010 @item -Wno-unused-result
4011 @opindex Wunused-result
4012 @opindex Wno-unused-result
4013 Do not warn if a caller of a function marked with attribute
4014 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4015 its return value. The default is @option{-Wunused-result}.
4017 @item -Wunused-variable
4018 @opindex Wunused-variable
4019 @opindex Wno-unused-variable
4020 Warn whenever a local variable or non-constant static variable is unused
4021 aside from its declaration.
4022 This warning is enabled by @option{-Wall}.
4024 To suppress this warning use the @code{unused} attribute
4025 (@pxref{Variable Attributes}).
4027 @item -Wunused-value
4028 @opindex Wunused-value
4029 @opindex Wno-unused-value
4030 Warn whenever a statement computes a result that is explicitly not
4031 used. To suppress this warning cast the unused expression to
4032 @code{void}. This includes an expression-statement or the left-hand
4033 side of a comma expression that contains no side effects. For example,
4034 an expression such as @code{x[i,j]} causes a warning, while
4035 @code{x[(void)i,j]} does not.
4037 This warning is enabled by @option{-Wall}.
4042 All the above @option{-Wunused} options combined.
4044 In order to get a warning about an unused function parameter, you must
4045 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4046 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4048 @item -Wuninitialized
4049 @opindex Wuninitialized
4050 @opindex Wno-uninitialized
4051 Warn if an automatic variable is used without first being initialized
4052 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4053 warn if a non-static reference or non-static @code{const} member
4054 appears in a class without constructors.
4056 If you want to warn about code that uses the uninitialized value of the
4057 variable in its own initializer, use the @option{-Winit-self} option.
4059 These warnings occur for individual uninitialized or clobbered
4060 elements of structure, union or array variables as well as for
4061 variables that are uninitialized or clobbered as a whole. They do
4062 not occur for variables or elements declared @code{volatile}. Because
4063 these warnings depend on optimization, the exact variables or elements
4064 for which there are warnings depends on the precise optimization
4065 options and version of GCC used.
4067 Note that there may be no warning about a variable that is used only
4068 to compute a value that itself is never used, because such
4069 computations may be deleted by data flow analysis before the warnings
4072 @item -Wmaybe-uninitialized
4073 @opindex Wmaybe-uninitialized
4074 @opindex Wno-maybe-uninitialized
4075 For an automatic variable, if there exists a path from the function
4076 entry to a use of the variable that is initialized, but there exist
4077 some other paths for which the variable is not initialized, the compiler
4078 emits a warning if it cannot prove the uninitialized paths are not
4079 executed at run time. These warnings are made optional because GCC is
4080 not smart enough to see all the reasons why the code might be correct
4081 in spite of appearing to have an error. Here is one example of how
4102 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4103 always initialized, but GCC doesn't know this. To suppress the
4104 warning, you need to provide a default case with assert(0) or
4107 @cindex @code{longjmp} warnings
4108 This option also warns when a non-volatile automatic variable might be
4109 changed by a call to @code{longjmp}. These warnings as well are possible
4110 only in optimizing compilation.
4112 The compiler sees only the calls to @code{setjmp}. It cannot know
4113 where @code{longjmp} will be called; in fact, a signal handler could
4114 call it at any point in the code. As a result, you may get a warning
4115 even when there is in fact no problem because @code{longjmp} cannot
4116 in fact be called at the place that would cause a problem.
4118 Some spurious warnings can be avoided if you declare all the functions
4119 you use that never return as @code{noreturn}. @xref{Function
4122 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4124 @item -Wunknown-pragmas
4125 @opindex Wunknown-pragmas
4126 @opindex Wno-unknown-pragmas
4127 @cindex warning for unknown pragmas
4128 @cindex unknown pragmas, warning
4129 @cindex pragmas, warning of unknown
4130 Warn when a @code{#pragma} directive is encountered that is not understood by
4131 GCC@. If this command-line option is used, warnings are even issued
4132 for unknown pragmas in system header files. This is not the case if
4133 the warnings are only enabled by the @option{-Wall} command-line option.
4136 @opindex Wno-pragmas
4138 Do not warn about misuses of pragmas, such as incorrect parameters,
4139 invalid syntax, or conflicts between pragmas. See also
4140 @option{-Wunknown-pragmas}.
4142 @item -Wstrict-aliasing
4143 @opindex Wstrict-aliasing
4144 @opindex Wno-strict-aliasing
4145 This option is only active when @option{-fstrict-aliasing} is active.
4146 It warns about code that might break the strict aliasing rules that the
4147 compiler is using for optimization. The warning does not catch all
4148 cases, but does attempt to catch the more common pitfalls. It is
4149 included in @option{-Wall}.
4150 It is equivalent to @option{-Wstrict-aliasing=3}
4152 @item -Wstrict-aliasing=n
4153 @opindex Wstrict-aliasing=n
4154 This option is only active when @option{-fstrict-aliasing} is active.
4155 It warns about code that might break the strict aliasing rules that the
4156 compiler is using for optimization.
4157 Higher levels correspond to higher accuracy (fewer false positives).
4158 Higher levels also correspond to more effort, similar to the way @option{-O}
4160 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4162 Level 1: Most aggressive, quick, least accurate.
4163 Possibly useful when higher levels
4164 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4165 false negatives. However, it has many false positives.
4166 Warns for all pointer conversions between possibly incompatible types,
4167 even if never dereferenced. Runs in the front end only.
4169 Level 2: Aggressive, quick, not too precise.
4170 May still have many false positives (not as many as level 1 though),
4171 and few false negatives (but possibly more than level 1).
4172 Unlike level 1, it only warns when an address is taken. Warns about
4173 incomplete types. Runs in the front end only.
4175 Level 3 (default for @option{-Wstrict-aliasing}):
4176 Should have very few false positives and few false
4177 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4178 Takes care of the common pun+dereference pattern in the front end:
4179 @code{*(int*)&some_float}.
4180 If optimization is enabled, it also runs in the back end, where it deals
4181 with multiple statement cases using flow-sensitive points-to information.
4182 Only warns when the converted pointer is dereferenced.
4183 Does not warn about incomplete types.
4185 @item -Wstrict-overflow
4186 @itemx -Wstrict-overflow=@var{n}
4187 @opindex Wstrict-overflow
4188 @opindex Wno-strict-overflow
4189 This option is only active when @option{-fstrict-overflow} is active.
4190 It warns about cases where the compiler optimizes based on the
4191 assumption that signed overflow does not occur. Note that it does not
4192 warn about all cases where the code might overflow: it only warns
4193 about cases where the compiler implements some optimization. Thus
4194 this warning depends on the optimization level.
4196 An optimization that assumes that signed overflow does not occur is
4197 perfectly safe if the values of the variables involved are such that
4198 overflow never does, in fact, occur. Therefore this warning can
4199 easily give a false positive: a warning about code that is not
4200 actually a problem. To help focus on important issues, several
4201 warning levels are defined. No warnings are issued for the use of
4202 undefined signed overflow when estimating how many iterations a loop
4203 requires, in particular when determining whether a loop will be
4207 @item -Wstrict-overflow=1
4208 Warn about cases that are both questionable and easy to avoid. For
4209 example, with @option{-fstrict-overflow}, the compiler simplifies
4210 @code{x + 1 > x} to @code{1}. This level of
4211 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4212 are not, and must be explicitly requested.
4214 @item -Wstrict-overflow=2
4215 Also warn about other cases where a comparison is simplified to a
4216 constant. For example: @code{abs (x) >= 0}. This can only be
4217 simplified when @option{-fstrict-overflow} is in effect, because
4218 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4219 zero. @option{-Wstrict-overflow} (with no level) is the same as
4220 @option{-Wstrict-overflow=2}.
4222 @item -Wstrict-overflow=3
4223 Also warn about other cases where a comparison is simplified. For
4224 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4226 @item -Wstrict-overflow=4
4227 Also warn about other simplifications not covered by the above cases.
4228 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4230 @item -Wstrict-overflow=5
4231 Also warn about cases where the compiler reduces the magnitude of a
4232 constant involved in a comparison. For example: @code{x + 2 > y} is
4233 simplified to @code{x + 1 >= y}. This is reported only at the
4234 highest warning level because this simplification applies to many
4235 comparisons, so this warning level gives a very large number of
4239 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4240 @opindex Wsuggest-attribute=
4241 @opindex Wno-suggest-attribute=
4242 Warn for cases where adding an attribute may be beneficial. The
4243 attributes currently supported are listed below.
4246 @item -Wsuggest-attribute=pure
4247 @itemx -Wsuggest-attribute=const
4248 @itemx -Wsuggest-attribute=noreturn
4249 @opindex Wsuggest-attribute=pure
4250 @opindex Wno-suggest-attribute=pure
4251 @opindex Wsuggest-attribute=const
4252 @opindex Wno-suggest-attribute=const
4253 @opindex Wsuggest-attribute=noreturn
4254 @opindex Wno-suggest-attribute=noreturn
4256 Warn about functions that might be candidates for attributes
4257 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4258 functions visible in other compilation units or (in the case of @code{pure} and
4259 @code{const}) if it cannot prove that the function returns normally. A function
4260 returns normally if it doesn't contain an infinite loop or return abnormally
4261 by throwing, calling @code{abort} or trapping. This analysis requires option
4262 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4263 higher. Higher optimization levels improve the accuracy of the analysis.
4265 @item -Wsuggest-attribute=format
4266 @itemx -Wmissing-format-attribute
4267 @opindex Wsuggest-attribute=format
4268 @opindex Wmissing-format-attribute
4269 @opindex Wno-suggest-attribute=format
4270 @opindex Wno-missing-format-attribute
4274 Warn about function pointers that might be candidates for @code{format}
4275 attributes. Note these are only possible candidates, not absolute ones.
4276 GCC guesses that function pointers with @code{format} attributes that
4277 are used in assignment, initialization, parameter passing or return
4278 statements should have a corresponding @code{format} attribute in the
4279 resulting type. I.e.@: the left-hand side of the assignment or
4280 initialization, the type of the parameter variable, or the return type
4281 of the containing function respectively should also have a @code{format}
4282 attribute to avoid the warning.
4284 GCC also warns about function definitions that might be
4285 candidates for @code{format} attributes. Again, these are only
4286 possible candidates. GCC guesses that @code{format} attributes
4287 might be appropriate for any function that calls a function like
4288 @code{vprintf} or @code{vscanf}, but this might not always be the
4289 case, and some functions for which @code{format} attributes are
4290 appropriate may not be detected.
4293 @item -Wsuggest-final-types
4294 @opindex Wno-suggest-final-types
4295 @opindex Wsuggest-final-types
4296 Warn about types with virtual methods where code quality would be improved
4297 if the type were declared with the C++11 @code{final} specifier,
4299 declared in an anonymous namespace. This allows GCC to more aggressively
4300 devirtualize the polymorphic calls. This warning is more effective with link
4301 time optimization, where the information about the class hierarchy graph is
4304 @item -Wsuggest-final-methods
4305 @opindex Wno-suggest-final-methods
4306 @opindex Wsuggest-final-methods
4307 Warn about virtual methods where code quality would be improved if the method
4308 were declared with the C++11 @code{final} specifier,
4309 or, if possible, its type were
4310 declared in an anonymous namespace or with the @code{final} specifier.
4312 more effective with link time optimization, where the information about the
4313 class hierarchy graph is more complete. It is recommended to first consider
4314 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4317 @item -Wsuggest-override
4318 Warn about overriding virtual functions that are not marked with the override
4321 @item -Warray-bounds
4322 @itemx -Warray-bounds=@var{n}
4323 @opindex Wno-array-bounds
4324 @opindex Warray-bounds
4325 This option is only active when @option{-ftree-vrp} is active
4326 (default for @option{-O2} and above). It warns about subscripts to arrays
4327 that are always out of bounds. This warning is enabled by @option{-Wall}.
4330 @item -Warray-bounds=1
4331 This is the warning level of @option{-Warray-bounds} and is enabled
4332 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4334 @item -Warray-bounds=2
4335 This warning level also warns about out of bounds access for
4336 arrays at the end of a struct and for arrays accessed through
4337 pointers. This warning level may give a larger number of
4338 false positives and is deactivated by default.
4342 @item -Wbool-compare
4343 @opindex Wno-bool-compare
4344 @opindex Wbool-compare
4345 Warn about boolean expression compared with an integer value different from
4346 @code{true}/@code{false}. For instance, the following comparison is
4351 if ((n > 1) == 2) @{ @dots{} @}
4353 This warning is enabled by @option{-Wall}.
4355 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4356 @opindex Wno-discarded-qualifiers
4357 @opindex Wdiscarded-qualifiers
4358 Do not warn if type qualifiers on pointers are being discarded.
4359 Typically, the compiler warns if a @code{const char *} variable is
4360 passed to a function that takes a @code{char *} parameter. This option
4361 can be used to suppress such a warning.
4363 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4364 @opindex Wno-discarded-array-qualifiers
4365 @opindex Wdiscarded-array-qualifiers
4366 Do not warn if type qualifiers on arrays which are pointer targets
4367 are being discarded. Typically, the compiler warns if a
4368 @code{const int (*)[]} variable is passed to a function that
4369 takes a @code{int (*)[]} parameter. This option can be used to
4370 suppress such a warning.
4372 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4373 @opindex Wno-incompatible-pointer-types
4374 @opindex Wincompatible-pointer-types
4375 Do not warn when there is a conversion between pointers that have incompatible
4376 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4377 which warns for pointer argument passing or assignment with different
4380 @item -Wno-int-conversion @r{(C and Objective-C only)}
4381 @opindex Wno-int-conversion
4382 @opindex Wint-conversion
4383 Do not warn about incompatible integer to pointer and pointer to integer
4384 conversions. This warning is about implicit conversions; for explicit
4385 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4386 @option{-Wno-pointer-to-int-cast} may be used.
4388 @item -Wno-div-by-zero
4389 @opindex Wno-div-by-zero
4390 @opindex Wdiv-by-zero
4391 Do not warn about compile-time integer division by zero. Floating-point
4392 division by zero is not warned about, as it can be a legitimate way of
4393 obtaining infinities and NaNs.
4395 @item -Wsystem-headers
4396 @opindex Wsystem-headers
4397 @opindex Wno-system-headers
4398 @cindex warnings from system headers
4399 @cindex system headers, warnings from
4400 Print warning messages for constructs found in system header files.
4401 Warnings from system headers are normally suppressed, on the assumption
4402 that they usually do not indicate real problems and would only make the
4403 compiler output harder to read. Using this command-line option tells
4404 GCC to emit warnings from system headers as if they occurred in user
4405 code. However, note that using @option{-Wall} in conjunction with this
4406 option does @emph{not} warn about unknown pragmas in system
4407 headers---for that, @option{-Wunknown-pragmas} must also be used.
4410 @opindex Wtrampolines
4411 @opindex Wno-trampolines
4412 Warn about trampolines generated for pointers to nested functions.
4413 A trampoline is a small piece of data or code that is created at run
4414 time on the stack when the address of a nested function is taken, and is
4415 used to call the nested function indirectly. For some targets, it is
4416 made up of data only and thus requires no special treatment. But, for
4417 most targets, it is made up of code and thus requires the stack to be
4418 made executable in order for the program to work properly.
4421 @opindex Wfloat-equal
4422 @opindex Wno-float-equal
4423 Warn if floating-point values are used in equality comparisons.
4425 The idea behind this is that sometimes it is convenient (for the
4426 programmer) to consider floating-point values as approximations to
4427 infinitely precise real numbers. If you are doing this, then you need
4428 to compute (by analyzing the code, or in some other way) the maximum or
4429 likely maximum error that the computation introduces, and allow for it
4430 when performing comparisons (and when producing output, but that's a
4431 different problem). In particular, instead of testing for equality, you
4432 should check to see whether the two values have ranges that overlap; and
4433 this is done with the relational operators, so equality comparisons are
4436 @item -Wtraditional @r{(C and Objective-C only)}
4437 @opindex Wtraditional
4438 @opindex Wno-traditional
4439 Warn about certain constructs that behave differently in traditional and
4440 ISO C@. Also warn about ISO C constructs that have no traditional C
4441 equivalent, and/or problematic constructs that should be avoided.
4445 Macro parameters that appear within string literals in the macro body.
4446 In traditional C macro replacement takes place within string literals,
4447 but in ISO C it does not.
4450 In traditional C, some preprocessor directives did not exist.
4451 Traditional preprocessors only considered a line to be a directive
4452 if the @samp{#} appeared in column 1 on the line. Therefore
4453 @option{-Wtraditional} warns about directives that traditional C
4454 understands but ignores because the @samp{#} does not appear as the
4455 first character on the line. It also suggests you hide directives like
4456 @code{#pragma} not understood by traditional C by indenting them. Some
4457 traditional implementations do not recognize @code{#elif}, so this option
4458 suggests avoiding it altogether.
4461 A function-like macro that appears without arguments.
4464 The unary plus operator.
4467 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4468 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4469 constants.) Note, these suffixes appear in macros defined in the system
4470 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4471 Use of these macros in user code might normally lead to spurious
4472 warnings, however GCC's integrated preprocessor has enough context to
4473 avoid warning in these cases.
4476 A function declared external in one block and then used after the end of
4480 A @code{switch} statement has an operand of type @code{long}.
4483 A non-@code{static} function declaration follows a @code{static} one.
4484 This construct is not accepted by some traditional C compilers.
4487 The ISO type of an integer constant has a different width or
4488 signedness from its traditional type. This warning is only issued if
4489 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4490 typically represent bit patterns, are not warned about.
4493 Usage of ISO string concatenation is detected.
4496 Initialization of automatic aggregates.
4499 Identifier conflicts with labels. Traditional C lacks a separate
4500 namespace for labels.
4503 Initialization of unions. If the initializer is zero, the warning is
4504 omitted. This is done under the assumption that the zero initializer in
4505 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4506 initializer warnings and relies on default initialization to zero in the
4510 Conversions by prototypes between fixed/floating-point values and vice
4511 versa. The absence of these prototypes when compiling with traditional
4512 C causes serious problems. This is a subset of the possible
4513 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4516 Use of ISO C style function definitions. This warning intentionally is
4517 @emph{not} issued for prototype declarations or variadic functions
4518 because these ISO C features appear in your code when using
4519 libiberty's traditional C compatibility macros, @code{PARAMS} and
4520 @code{VPARAMS}. This warning is also bypassed for nested functions
4521 because that feature is already a GCC extension and thus not relevant to
4522 traditional C compatibility.
4525 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4526 @opindex Wtraditional-conversion
4527 @opindex Wno-traditional-conversion
4528 Warn if a prototype causes a type conversion that is different from what
4529 would happen to the same argument in the absence of a prototype. This
4530 includes conversions of fixed point to floating and vice versa, and
4531 conversions changing the width or signedness of a fixed-point argument
4532 except when the same as the default promotion.
4534 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4535 @opindex Wdeclaration-after-statement
4536 @opindex Wno-declaration-after-statement
4537 Warn when a declaration is found after a statement in a block. This
4538 construct, known from C++, was introduced with ISO C99 and is by default
4539 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4540 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4545 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4547 @item -Wno-endif-labels
4548 @opindex Wno-endif-labels
4549 @opindex Wendif-labels
4550 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4555 Warn whenever a local variable or type declaration shadows another
4556 variable, parameter, type, class member (in C++), or instance variable
4557 (in Objective-C) or whenever a built-in function is shadowed. Note
4558 that in C++, the compiler warns if a local variable shadows an
4559 explicit typedef, but not if it shadows a struct/class/enum.
4561 @item -Wno-shadow-ivar @r{(Objective-C only)}
4562 @opindex Wno-shadow-ivar
4563 @opindex Wshadow-ivar
4564 Do not warn whenever a local variable shadows an instance variable in an
4567 @item -Wlarger-than=@var{len}
4568 @opindex Wlarger-than=@var{len}
4569 @opindex Wlarger-than-@var{len}
4570 Warn whenever an object of larger than @var{len} bytes is defined.
4572 @item -Wframe-larger-than=@var{len}
4573 @opindex Wframe-larger-than
4574 Warn if the size of a function frame is larger than @var{len} bytes.
4575 The computation done to determine the stack frame size is approximate
4576 and not conservative.
4577 The actual requirements may be somewhat greater than @var{len}
4578 even if you do not get a warning. In addition, any space allocated
4579 via @code{alloca}, variable-length arrays, or related constructs
4580 is not included by the compiler when determining
4581 whether or not to issue a warning.
4583 @item -Wno-free-nonheap-object
4584 @opindex Wno-free-nonheap-object
4585 @opindex Wfree-nonheap-object
4586 Do not warn when attempting to free an object that was not allocated
4589 @item -Wstack-usage=@var{len}
4590 @opindex Wstack-usage
4591 Warn if the stack usage of a function might be larger than @var{len} bytes.
4592 The computation done to determine the stack usage is conservative.
4593 Any space allocated via @code{alloca}, variable-length arrays, or related
4594 constructs is included by the compiler when determining whether or not to
4597 The message is in keeping with the output of @option{-fstack-usage}.
4601 If the stack usage is fully static but exceeds the specified amount, it's:
4604 warning: stack usage is 1120 bytes
4607 If the stack usage is (partly) dynamic but bounded, it's:
4610 warning: stack usage might be 1648 bytes
4613 If the stack usage is (partly) dynamic and not bounded, it's:
4616 warning: stack usage might be unbounded
4620 @item -Wunsafe-loop-optimizations
4621 @opindex Wunsafe-loop-optimizations
4622 @opindex Wno-unsafe-loop-optimizations
4623 Warn if the loop cannot be optimized because the compiler cannot
4624 assume anything on the bounds of the loop indices. With
4625 @option{-funsafe-loop-optimizations} warn if the compiler makes
4628 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4629 @opindex Wno-pedantic-ms-format
4630 @opindex Wpedantic-ms-format
4631 When used in combination with @option{-Wformat}
4632 and @option{-pedantic} without GNU extensions, this option
4633 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4634 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4635 which depend on the MS runtime.
4637 @item -Wpointer-arith
4638 @opindex Wpointer-arith
4639 @opindex Wno-pointer-arith
4640 Warn about anything that depends on the ``size of'' a function type or
4641 of @code{void}. GNU C assigns these types a size of 1, for
4642 convenience in calculations with @code{void *} pointers and pointers
4643 to functions. In C++, warn also when an arithmetic operation involves
4644 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4647 @opindex Wtype-limits
4648 @opindex Wno-type-limits
4649 Warn if a comparison is always true or always false due to the limited
4650 range of the data type, but do not warn for constant expressions. For
4651 example, warn if an unsigned variable is compared against zero with
4652 @code{<} or @code{>=}. This warning is also enabled by
4655 @item -Wbad-function-cast @r{(C and Objective-C only)}
4656 @opindex Wbad-function-cast
4657 @opindex Wno-bad-function-cast
4658 Warn when a function call is cast to a non-matching type.
4659 For example, warn if a call to a function returning an integer type
4660 is cast to a pointer type.
4662 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4663 @opindex Wc90-c99-compat
4664 @opindex Wno-c90-c99-compat
4665 Warn about features not present in ISO C90, but present in ISO C99.
4666 For instance, warn about use of variable length arrays, @code{long long}
4667 type, @code{bool} type, compound literals, designated initializers, and so
4668 on. This option is independent of the standards mode. Warnings are disabled
4669 in the expression that follows @code{__extension__}.
4671 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4672 @opindex Wc99-c11-compat
4673 @opindex Wno-c99-c11-compat
4674 Warn about features not present in ISO C99, but present in ISO C11.
4675 For instance, warn about use of anonymous structures and unions,
4676 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4677 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4678 and so on. This option is independent of the standards mode. Warnings are
4679 disabled in the expression that follows @code{__extension__}.
4681 @item -Wc++-compat @r{(C and Objective-C only)}
4682 @opindex Wc++-compat
4683 Warn about ISO C constructs that are outside of the common subset of
4684 ISO C and ISO C++, e.g.@: request for implicit conversion from
4685 @code{void *} to a pointer to non-@code{void} type.
4687 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4688 @opindex Wc++11-compat
4689 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4690 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4691 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4692 enabled by @option{-Wall}.
4694 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4695 @opindex Wc++14-compat
4696 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4697 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4701 @opindex Wno-cast-qual
4702 Warn whenever a pointer is cast so as to remove a type qualifier from
4703 the target type. For example, warn if a @code{const char *} is cast
4704 to an ordinary @code{char *}.
4706 Also warn when making a cast that introduces a type qualifier in an
4707 unsafe way. For example, casting @code{char **} to @code{const char **}
4708 is unsafe, as in this example:
4711 /* p is char ** value. */
4712 const char **q = (const char **) p;
4713 /* Assignment of readonly string to const char * is OK. */
4715 /* Now char** pointer points to read-only memory. */
4720 @opindex Wcast-align
4721 @opindex Wno-cast-align
4722 Warn whenever a pointer is cast such that the required alignment of the
4723 target is increased. For example, warn if a @code{char *} is cast to
4724 an @code{int *} on machines where integers can only be accessed at
4725 two- or four-byte boundaries.
4727 @item -Wwrite-strings
4728 @opindex Wwrite-strings
4729 @opindex Wno-write-strings
4730 When compiling C, give string constants the type @code{const
4731 char[@var{length}]} so that copying the address of one into a
4732 non-@code{const} @code{char *} pointer produces a warning. These
4733 warnings help you find at compile time code that can try to write
4734 into a string constant, but only if you have been very careful about
4735 using @code{const} in declarations and prototypes. Otherwise, it is
4736 just a nuisance. This is why we did not make @option{-Wall} request
4739 When compiling C++, warn about the deprecated conversion from string
4740 literals to @code{char *}. This warning is enabled by default for C++
4745 @opindex Wno-clobbered
4746 Warn for variables that might be changed by @code{longjmp} or
4747 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4749 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4750 @opindex Wconditionally-supported
4751 @opindex Wno-conditionally-supported
4752 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4755 @opindex Wconversion
4756 @opindex Wno-conversion
4757 Warn for implicit conversions that may alter a value. This includes
4758 conversions between real and integer, like @code{abs (x)} when
4759 @code{x} is @code{double}; conversions between signed and unsigned,
4760 like @code{unsigned ui = -1}; and conversions to smaller types, like
4761 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4762 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4763 changed by the conversion like in @code{abs (2.0)}. Warnings about
4764 conversions between signed and unsigned integers can be disabled by
4765 using @option{-Wno-sign-conversion}.
4767 For C++, also warn for confusing overload resolution for user-defined
4768 conversions; and conversions that never use a type conversion
4769 operator: conversions to @code{void}, the same type, a base class or a
4770 reference to them. Warnings about conversions between signed and
4771 unsigned integers are disabled by default in C++ unless
4772 @option{-Wsign-conversion} is explicitly enabled.
4774 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4775 @opindex Wconversion-null
4776 @opindex Wno-conversion-null
4777 Do not warn for conversions between @code{NULL} and non-pointer
4778 types. @option{-Wconversion-null} is enabled by default.
4780 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4781 @opindex Wzero-as-null-pointer-constant
4782 @opindex Wno-zero-as-null-pointer-constant
4783 Warn when a literal '0' is used as null pointer constant. This can
4784 be useful to facilitate the conversion to @code{nullptr} in C++11.
4788 @opindex Wno-date-time
4789 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4790 are encountered as they might prevent bit-wise-identical reproducible
4793 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4794 @opindex Wdelete-incomplete
4795 @opindex Wno-delete-incomplete
4796 Warn when deleting a pointer to incomplete type, which may cause
4797 undefined behavior at runtime. This warning is enabled by default.
4799 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4800 @opindex Wuseless-cast
4801 @opindex Wno-useless-cast
4802 Warn when an expression is casted to its own type.
4805 @opindex Wempty-body
4806 @opindex Wno-empty-body
4807 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
4808 while} statement. This warning is also enabled by @option{-Wextra}.
4810 @item -Wenum-compare
4811 @opindex Wenum-compare
4812 @opindex Wno-enum-compare
4813 Warn about a comparison between values of different enumerated types.
4814 In C++ enumeral mismatches in conditional expressions are also
4815 diagnosed and the warning is enabled by default. In C this warning is
4816 enabled by @option{-Wall}.
4818 @item -Wjump-misses-init @r{(C, Objective-C only)}
4819 @opindex Wjump-misses-init
4820 @opindex Wno-jump-misses-init
4821 Warn if a @code{goto} statement or a @code{switch} statement jumps
4822 forward across the initialization of a variable, or jumps backward to a
4823 label after the variable has been initialized. This only warns about
4824 variables that are initialized when they are declared. This warning is
4825 only supported for C and Objective-C; in C++ this sort of branch is an
4828 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4829 can be disabled with the @option{-Wno-jump-misses-init} option.
4831 @item -Wsign-compare
4832 @opindex Wsign-compare
4833 @opindex Wno-sign-compare
4834 @cindex warning for comparison of signed and unsigned values
4835 @cindex comparison of signed and unsigned values, warning
4836 @cindex signed and unsigned values, comparison warning
4837 Warn when a comparison between signed and unsigned values could produce
4838 an incorrect result when the signed value is converted to unsigned.
4839 This warning is also enabled by @option{-Wextra}; to get the other warnings
4840 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4842 @item -Wsign-conversion
4843 @opindex Wsign-conversion
4844 @opindex Wno-sign-conversion
4845 Warn for implicit conversions that may change the sign of an integer
4846 value, like assigning a signed integer expression to an unsigned
4847 integer variable. An explicit cast silences the warning. In C, this
4848 option is enabled also by @option{-Wconversion}.
4850 @item -Wfloat-conversion
4851 @opindex Wfloat-conversion
4852 @opindex Wno-float-conversion
4853 Warn for implicit conversions that reduce the precision of a real value.
4854 This includes conversions from real to integer, and from higher precision
4855 real to lower precision real values. This option is also enabled by
4856 @option{-Wconversion}.
4858 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
4859 @opindex Wsized-deallocation
4860 @opindex Wno-sized-deallocation
4861 Warn about a definition of an unsized deallocation function
4863 void operator delete (void *) noexcept;
4864 void operator delete[] (void *) noexcept;
4866 without a definition of the corresponding sized deallocation function
4868 void operator delete (void *, std::size_t) noexcept;
4869 void operator delete[] (void *, std::size_t) noexcept;
4871 or vice versa. Enabled by @option{-Wextra} along with
4872 @option{-fsized-deallocation}.
4874 @item -Wsizeof-pointer-memaccess
4875 @opindex Wsizeof-pointer-memaccess
4876 @opindex Wno-sizeof-pointer-memaccess
4877 Warn for suspicious length parameters to certain string and memory built-in
4878 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4879 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4880 but a pointer, and suggests a possible fix, or about
4881 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4884 @item -Wsizeof-array-argument
4885 @opindex Wsizeof-array-argument
4886 @opindex Wno-sizeof-array-argument
4887 Warn when the @code{sizeof} operator is applied to a parameter that is
4888 declared as an array in a function definition. This warning is enabled by
4889 default for C and C++ programs.
4891 @item -Wmemset-transposed-args
4892 @opindex Wmemset-transposed-args
4893 @opindex Wno-memset-transposed-args
4894 Warn for suspicious calls to the @code{memset} built-in function, if the
4895 second argument is not zero and the third argument is zero. This warns e.g.@
4896 about @code{memset (buf, sizeof buf, 0)} where most probably
4897 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4898 is only emitted if the third argument is literal zero, if it is some expression
4899 that is folded to zero, or e.g. a cast of zero to some type etc., it
4900 is far less likely that user has mistakenly exchanged the arguments and
4901 no warning is emitted. This warning is enabled by @option{-Wall}.
4905 @opindex Wno-address
4906 Warn about suspicious uses of memory addresses. These include using
4907 the address of a function in a conditional expression, such as
4908 @code{void func(void); if (func)}, and comparisons against the memory
4909 address of a string literal, such as @code{if (x == "abc")}. Such
4910 uses typically indicate a programmer error: the address of a function
4911 always evaluates to true, so their use in a conditional usually
4912 indicate that the programmer forgot the parentheses in a function
4913 call; and comparisons against string literals result in unspecified
4914 behavior and are not portable in C, so they usually indicate that the
4915 programmer intended to use @code{strcmp}. This warning is enabled by
4919 @opindex Wlogical-op
4920 @opindex Wno-logical-op
4921 Warn about suspicious uses of logical operators in expressions.
4922 This includes using logical operators in contexts where a
4923 bit-wise operator is likely to be expected.
4925 @item -Wlogical-not-parentheses
4926 @opindex Wlogical-not-parentheses
4927 @opindex Wno-logical-not-parentheses
4928 Warn about logical not used on the left hand side operand of a comparison.
4929 This option does not warn if the RHS operand is of a boolean type. Its
4930 purpose is to detect suspicious code like the following:
4934 if (!a > 1) @{ @dots{} @}
4937 It is possible to suppress the warning by wrapping the LHS into
4940 if ((!a) > 1) @{ @dots{} @}
4943 This warning is enabled by @option{-Wall}.
4945 @item -Waggregate-return
4946 @opindex Waggregate-return
4947 @opindex Wno-aggregate-return
4948 Warn if any functions that return structures or unions are defined or
4949 called. (In languages where you can return an array, this also elicits
4952 @item -Wno-aggressive-loop-optimizations
4953 @opindex Wno-aggressive-loop-optimizations
4954 @opindex Waggressive-loop-optimizations
4955 Warn if in a loop with constant number of iterations the compiler detects
4956 undefined behavior in some statement during one or more of the iterations.
4958 @item -Wno-attributes
4959 @opindex Wno-attributes
4960 @opindex Wattributes
4961 Do not warn if an unexpected @code{__attribute__} is used, such as
4962 unrecognized attributes, function attributes applied to variables,
4963 etc. This does not stop errors for incorrect use of supported
4966 @item -Wno-builtin-macro-redefined
4967 @opindex Wno-builtin-macro-redefined
4968 @opindex Wbuiltin-macro-redefined
4969 Do not warn if certain built-in macros are redefined. This suppresses
4970 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4971 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4973 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4974 @opindex Wstrict-prototypes
4975 @opindex Wno-strict-prototypes
4976 Warn if a function is declared or defined without specifying the
4977 argument types. (An old-style function definition is permitted without
4978 a warning if preceded by a declaration that specifies the argument
4981 @item -Wold-style-declaration @r{(C and Objective-C only)}
4982 @opindex Wold-style-declaration
4983 @opindex Wno-old-style-declaration
4984 Warn for obsolescent usages, according to the C Standard, in a
4985 declaration. For example, warn if storage-class specifiers like
4986 @code{static} are not the first things in a declaration. This warning
4987 is also enabled by @option{-Wextra}.
4989 @item -Wold-style-definition @r{(C and Objective-C only)}
4990 @opindex Wold-style-definition
4991 @opindex Wno-old-style-definition
4992 Warn if an old-style function definition is used. A warning is given
4993 even if there is a previous prototype.
4995 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4996 @opindex Wmissing-parameter-type
4997 @opindex Wno-missing-parameter-type
4998 A function parameter is declared without a type specifier in K&R-style
5005 This warning is also enabled by @option{-Wextra}.
5007 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5008 @opindex Wmissing-prototypes
5009 @opindex Wno-missing-prototypes
5010 Warn if a global function is defined without a previous prototype
5011 declaration. This warning is issued even if the definition itself
5012 provides a prototype. Use this option to detect global functions
5013 that do not have a matching prototype declaration in a header file.
5014 This option is not valid for C++ because all function declarations
5015 provide prototypes and a non-matching declaration declares an
5016 overload rather than conflict with an earlier declaration.
5017 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5019 @item -Wmissing-declarations
5020 @opindex Wmissing-declarations
5021 @opindex Wno-missing-declarations
5022 Warn if a global function is defined without a previous declaration.
5023 Do so even if the definition itself provides a prototype.
5024 Use this option to detect global functions that are not declared in
5025 header files. In C, no warnings are issued for functions with previous
5026 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5027 missing prototypes. In C++, no warnings are issued for function templates,
5028 or for inline functions, or for functions in anonymous namespaces.
5030 @item -Wmissing-field-initializers
5031 @opindex Wmissing-field-initializers
5032 @opindex Wno-missing-field-initializers
5036 Warn if a structure's initializer has some fields missing. For
5037 example, the following code causes such a warning, because
5038 @code{x.h} is implicitly zero:
5041 struct s @{ int f, g, h; @};
5042 struct s x = @{ 3, 4 @};
5045 This option does not warn about designated initializers, so the following
5046 modification does not trigger a warning:
5049 struct s @{ int f, g, h; @};
5050 struct s x = @{ .f = 3, .g = 4 @};
5053 In C++ this option does not warn either about the empty @{ @}
5054 initializer, for example:
5057 struct s @{ int f, g, h; @};
5061 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5062 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5064 @item -Wno-multichar
5065 @opindex Wno-multichar
5067 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5068 Usually they indicate a typo in the user's code, as they have
5069 implementation-defined values, and should not be used in portable code.
5071 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5072 @opindex Wnormalized=
5073 @opindex Wnormalized
5074 @opindex Wno-normalized
5077 @cindex character set, input normalization
5078 In ISO C and ISO C++, two identifiers are different if they are
5079 different sequences of characters. However, sometimes when characters
5080 outside the basic ASCII character set are used, you can have two
5081 different character sequences that look the same. To avoid confusion,
5082 the ISO 10646 standard sets out some @dfn{normalization rules} which
5083 when applied ensure that two sequences that look the same are turned into
5084 the same sequence. GCC can warn you if you are using identifiers that
5085 have not been normalized; this option controls that warning.
5087 There are four levels of warning supported by GCC@. The default is
5088 @option{-Wnormalized=nfc}, which warns about any identifier that is
5089 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5090 recommended form for most uses. It is equivalent to
5091 @option{-Wnormalized}.
5093 Unfortunately, there are some characters allowed in identifiers by
5094 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5095 identifiers. That is, there's no way to use these symbols in portable
5096 ISO C or C++ and have all your identifiers in NFC@.
5097 @option{-Wnormalized=id} suppresses the warning for these characters.
5098 It is hoped that future versions of the standards involved will correct
5099 this, which is why this option is not the default.
5101 You can switch the warning off for all characters by writing
5102 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5103 only do this if you are using some other normalization scheme (like
5104 ``D''), because otherwise you can easily create bugs that are
5105 literally impossible to see.
5107 Some characters in ISO 10646 have distinct meanings but look identical
5108 in some fonts or display methodologies, especially once formatting has
5109 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5110 LETTER N'', displays just like a regular @code{n} that has been
5111 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5112 normalization scheme to convert all these into a standard form as
5113 well, and GCC warns if your code is not in NFKC if you use
5114 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5115 about every identifier that contains the letter O because it might be
5116 confused with the digit 0, and so is not the default, but may be
5117 useful as a local coding convention if the programming environment
5118 cannot be fixed to display these characters distinctly.
5120 @item -Wno-deprecated
5121 @opindex Wno-deprecated
5122 @opindex Wdeprecated
5123 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5125 @item -Wno-deprecated-declarations
5126 @opindex Wno-deprecated-declarations
5127 @opindex Wdeprecated-declarations
5128 Do not warn about uses of functions (@pxref{Function Attributes}),
5129 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5130 Attributes}) marked as deprecated by using the @code{deprecated}
5134 @opindex Wno-overflow
5136 Do not warn about compile-time overflow in constant expressions.
5141 Warn about One Definition Rule violations during link-time optimization.
5142 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5145 @opindex Wopenm-simd
5146 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5147 simd directive set by user. The @option{-fsimd-cost-model=unlimited} can
5148 be used to relax the cost model.
5150 @item -Woverride-init @r{(C and Objective-C only)}
5151 @opindex Woverride-init
5152 @opindex Wno-override-init
5156 Warn if an initialized field without side effects is overridden when
5157 using designated initializers (@pxref{Designated Inits, , Designated
5160 This warning is included in @option{-Wextra}. To get other
5161 @option{-Wextra} warnings without this one, use @option{-Wextra
5162 -Wno-override-init}.
5167 Warn if a structure is given the packed attribute, but the packed
5168 attribute has no effect on the layout or size of the structure.
5169 Such structures may be mis-aligned for little benefit. For
5170 instance, in this code, the variable @code{f.x} in @code{struct bar}
5171 is misaligned even though @code{struct bar} does not itself
5172 have the packed attribute:
5179 @} __attribute__((packed));
5187 @item -Wpacked-bitfield-compat
5188 @opindex Wpacked-bitfield-compat
5189 @opindex Wno-packed-bitfield-compat
5190 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5191 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5192 the change can lead to differences in the structure layout. GCC
5193 informs you when the offset of such a field has changed in GCC 4.4.
5194 For example there is no longer a 4-bit padding between field @code{a}
5195 and @code{b} in this structure:
5202 @} __attribute__ ((packed));
5205 This warning is enabled by default. Use
5206 @option{-Wno-packed-bitfield-compat} to disable this warning.
5211 Warn if padding is included in a structure, either to align an element
5212 of the structure or to align the whole structure. Sometimes when this
5213 happens it is possible to rearrange the fields of the structure to
5214 reduce the padding and so make the structure smaller.
5216 @item -Wredundant-decls
5217 @opindex Wredundant-decls
5218 @opindex Wno-redundant-decls
5219 Warn if anything is declared more than once in the same scope, even in
5220 cases where multiple declaration is valid and changes nothing.
5222 @item -Wnested-externs @r{(C and Objective-C only)}
5223 @opindex Wnested-externs
5224 @opindex Wno-nested-externs
5225 Warn if an @code{extern} declaration is encountered within a function.
5227 @item -Wno-inherited-variadic-ctor
5228 @opindex Winherited-variadic-ctor
5229 @opindex Wno-inherited-variadic-ctor
5230 Suppress warnings about use of C++11 inheriting constructors when the
5231 base class inherited from has a C variadic constructor; the warning is
5232 on by default because the ellipsis is not inherited.
5237 Warn if a function that is declared as inline cannot be inlined.
5238 Even with this option, the compiler does not warn about failures to
5239 inline functions declared in system headers.
5241 The compiler uses a variety of heuristics to determine whether or not
5242 to inline a function. For example, the compiler takes into account
5243 the size of the function being inlined and the amount of inlining
5244 that has already been done in the current function. Therefore,
5245 seemingly insignificant changes in the source program can cause the
5246 warnings produced by @option{-Winline} to appear or disappear.
5248 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5249 @opindex Wno-invalid-offsetof
5250 @opindex Winvalid-offsetof
5251 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5252 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5253 to a non-standard-layout type is undefined. In existing C++ implementations,
5254 however, @code{offsetof} typically gives meaningful results.
5255 This flag is for users who are aware that they are
5256 writing nonportable code and who have deliberately chosen to ignore the
5259 The restrictions on @code{offsetof} may be relaxed in a future version
5260 of the C++ standard.
5262 @item -Wno-int-to-pointer-cast
5263 @opindex Wno-int-to-pointer-cast
5264 @opindex Wint-to-pointer-cast
5265 Suppress warnings from casts to pointer type of an integer of a
5266 different size. In C++, casting to a pointer type of smaller size is
5267 an error. @option{Wint-to-pointer-cast} is enabled by default.
5270 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5271 @opindex Wno-pointer-to-int-cast
5272 @opindex Wpointer-to-int-cast
5273 Suppress warnings from casts from a pointer to an integer type of a
5277 @opindex Winvalid-pch
5278 @opindex Wno-invalid-pch
5279 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5280 the search path but can't be used.
5284 @opindex Wno-long-long
5285 Warn if @code{long long} type is used. This is enabled by either
5286 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5287 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5289 @item -Wvariadic-macros
5290 @opindex Wvariadic-macros
5291 @opindex Wno-variadic-macros
5292 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5293 alternate syntax is used in ISO C99 mode. This is enabled by either
5294 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5295 messages, use @option{-Wno-variadic-macros}.
5299 @opindex Wno-varargs
5300 Warn upon questionable usage of the macros used to handle variable
5301 arguments like @code{va_start}. This is default. To inhibit the
5302 warning messages, use @option{-Wno-varargs}.
5304 @item -Wvector-operation-performance
5305 @opindex Wvector-operation-performance
5306 @opindex Wno-vector-operation-performance
5307 Warn if vector operation is not implemented via SIMD capabilities of the
5308 architecture. Mainly useful for the performance tuning.
5309 Vector operation can be implemented @code{piecewise}, which means that the
5310 scalar operation is performed on every vector element;
5311 @code{in parallel}, which means that the vector operation is implemented
5312 using scalars of wider type, which normally is more performance efficient;
5313 and @code{as a single scalar}, which means that vector fits into a
5316 @item -Wno-virtual-move-assign
5317 @opindex Wvirtual-move-assign
5318 @opindex Wno-virtual-move-assign
5319 Suppress warnings about inheriting from a virtual base with a
5320 non-trivial C++11 move assignment operator. This is dangerous because
5321 if the virtual base is reachable along more than one path, it is
5322 moved multiple times, which can mean both objects end up in the
5323 moved-from state. If the move assignment operator is written to avoid
5324 moving from a moved-from object, this warning can be disabled.
5329 Warn if variable length array is used in the code.
5330 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5331 the variable length array.
5333 @item -Wvolatile-register-var
5334 @opindex Wvolatile-register-var
5335 @opindex Wno-volatile-register-var
5336 Warn if a register variable is declared volatile. The volatile
5337 modifier does not inhibit all optimizations that may eliminate reads
5338 and/or writes to register variables. This warning is enabled by
5341 @item -Wdisabled-optimization
5342 @opindex Wdisabled-optimization
5343 @opindex Wno-disabled-optimization
5344 Warn if a requested optimization pass is disabled. This warning does
5345 not generally indicate that there is anything wrong with your code; it
5346 merely indicates that GCC's optimizers are unable to handle the code
5347 effectively. Often, the problem is that your code is too big or too
5348 complex; GCC refuses to optimize programs when the optimization
5349 itself is likely to take inordinate amounts of time.
5351 @item -Wpointer-sign @r{(C and Objective-C only)}
5352 @opindex Wpointer-sign
5353 @opindex Wno-pointer-sign
5354 Warn for pointer argument passing or assignment with different signedness.
5355 This option is only supported for C and Objective-C@. It is implied by
5356 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5357 @option{-Wno-pointer-sign}.
5359 @item -Wstack-protector
5360 @opindex Wstack-protector
5361 @opindex Wno-stack-protector
5362 This option is only active when @option{-fstack-protector} is active. It
5363 warns about functions that are not protected against stack smashing.
5365 @item -Woverlength-strings
5366 @opindex Woverlength-strings
5367 @opindex Wno-overlength-strings
5368 Warn about string constants that are longer than the ``minimum
5369 maximum'' length specified in the C standard. Modern compilers
5370 generally allow string constants that are much longer than the
5371 standard's minimum limit, but very portable programs should avoid
5372 using longer strings.
5374 The limit applies @emph{after} string constant concatenation, and does
5375 not count the trailing NUL@. In C90, the limit was 509 characters; in
5376 C99, it was raised to 4095. C++98 does not specify a normative
5377 minimum maximum, so we do not diagnose overlength strings in C++@.
5379 This option is implied by @option{-Wpedantic}, and can be disabled with
5380 @option{-Wno-overlength-strings}.
5382 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5383 @opindex Wunsuffixed-float-constants
5385 Issue a warning for any floating constant that does not have
5386 a suffix. When used together with @option{-Wsystem-headers} it
5387 warns about such constants in system header files. This can be useful
5388 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5389 from the decimal floating-point extension to C99.
5391 @item -Wno-designated-init @r{(C and Objective-C only)}
5392 Suppress warnings when a positional initializer is used to initialize
5393 a structure that has been marked with the @code{designated_init}
5398 @node Debugging Options
5399 @section Options for Debugging Your Program or GCC
5400 @cindex options, debugging
5401 @cindex debugging information options
5403 GCC has various special options that are used for debugging
5404 either your program or GCC:
5409 Produce debugging information in the operating system's native format
5410 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5413 On most systems that use stabs format, @option{-g} enables use of extra
5414 debugging information that only GDB can use; this extra information
5415 makes debugging work better in GDB but probably makes other debuggers
5417 refuse to read the program. If you want to control for certain whether
5418 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5419 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5421 GCC allows you to use @option{-g} with
5422 @option{-O}. The shortcuts taken by optimized code may occasionally
5423 produce surprising results: some variables you declared may not exist
5424 at all; flow of control may briefly move where you did not expect it;
5425 some statements may not be executed because they compute constant
5426 results or their values are already at hand; some statements may
5427 execute in different places because they have been moved out of loops.
5429 Nevertheless it proves possible to debug optimized output. This makes
5430 it reasonable to use the optimizer for programs that might have bugs.
5432 The following options are useful when GCC is generated with the
5433 capability for more than one debugging format.
5436 @opindex gsplit-dwarf
5437 Separate as much dwarf debugging information as possible into a
5438 separate output file with the extension .dwo. This option allows
5439 the build system to avoid linking files with debug information. To
5440 be useful, this option requires a debugger capable of reading .dwo
5445 Produce debugging information for use by GDB@. This means to use the
5446 most expressive format available (DWARF 2, stabs, or the native format
5447 if neither of those are supported), including GDB extensions if at all
5452 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5454 @item -ggnu-pubnames
5455 @opindex ggnu-pubnames
5456 Generate .debug_pubnames and .debug_pubtypes sections in a format
5457 suitable for conversion into a GDB@ index. This option is only useful
5458 with a linker that can produce GDB@ index version 7.
5462 Produce debugging information in stabs format (if that is supported),
5463 without GDB extensions. This is the format used by DBX on most BSD
5464 systems. On MIPS, Alpha and System V Release 4 systems this option
5465 produces stabs debugging output that is not understood by DBX or SDB@.
5466 On System V Release 4 systems this option requires the GNU assembler.
5468 @item -feliminate-unused-debug-symbols
5469 @opindex feliminate-unused-debug-symbols
5470 Produce debugging information in stabs format (if that is supported),
5471 for only symbols that are actually used.
5473 @item -femit-class-debug-always
5474 @opindex femit-class-debug-always
5475 Instead of emitting debugging information for a C++ class in only one
5476 object file, emit it in all object files using the class. This option
5477 should be used only with debuggers that are unable to handle the way GCC
5478 normally emits debugging information for classes because using this
5479 option increases the size of debugging information by as much as a
5482 @item -fdebug-types-section
5483 @opindex fdebug-types-section
5484 @opindex fno-debug-types-section
5485 When using DWARF Version 4 or higher, type DIEs can be put into
5486 their own @code{.debug_types} section instead of making them part of the
5487 @code{.debug_info} section. It is more efficient to put them in a separate
5488 comdat sections since the linker can then remove duplicates.
5489 But not all DWARF consumers support @code{.debug_types} sections yet
5490 and on some objects @code{.debug_types} produces larger instead of smaller
5491 debugging information.
5495 Produce debugging information in stabs format (if that is supported),
5496 using GNU extensions understood only by the GNU debugger (GDB)@. The
5497 use of these extensions is likely to make other debuggers crash or
5498 refuse to read the program.
5502 Produce debugging information in COFF format (if that is supported).
5503 This is the format used by SDB on most System V systems prior to
5508 Produce debugging information in XCOFF format (if that is supported).
5509 This is the format used by the DBX debugger on IBM RS/6000 systems.
5513 Produce debugging information in XCOFF format (if that is supported),
5514 using GNU extensions understood only by the GNU debugger (GDB)@. The
5515 use of these extensions is likely to make other debuggers crash or
5516 refuse to read the program, and may cause assemblers other than the GNU
5517 assembler (GAS) to fail with an error.
5519 @item -gdwarf-@var{version}
5520 @opindex gdwarf-@var{version}
5521 Produce debugging information in DWARF format (if that is supported).
5522 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5523 for most targets is 4. DWARF Version 5 is only experimental.
5525 Note that with DWARF Version 2, some ports require and always
5526 use some non-conflicting DWARF 3 extensions in the unwind tables.
5528 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5529 for maximum benefit.
5531 @item -grecord-gcc-switches
5532 @opindex grecord-gcc-switches
5533 This switch causes the command-line options used to invoke the
5534 compiler that may affect code generation to be appended to the
5535 DW_AT_producer attribute in DWARF debugging information. The options
5536 are concatenated with spaces separating them from each other and from
5537 the compiler version. See also @option{-frecord-gcc-switches} for another
5538 way of storing compiler options into the object file. This is the default.
5540 @item -gno-record-gcc-switches
5541 @opindex gno-record-gcc-switches
5542 Disallow appending command-line options to the DW_AT_producer attribute
5543 in DWARF debugging information.
5545 @item -gstrict-dwarf
5546 @opindex gstrict-dwarf
5547 Disallow using extensions of later DWARF standard version than selected
5548 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5549 DWARF extensions from later standard versions is allowed.
5551 @item -gno-strict-dwarf
5552 @opindex gno-strict-dwarf
5553 Allow using extensions of later DWARF standard version than selected with
5554 @option{-gdwarf-@var{version}}.
5556 @item -gz@r{[}=@var{type}@r{]}
5558 Produce compressed debug sections in DWARF format, if that is supported.
5559 If @var{type} is not given, the default type depends on the capabilities
5560 of the assembler and linker used. @var{type} may be one of
5561 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5562 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5563 compression in traditional GNU format). If the linker doesn't support
5564 writing compressed debug sections, the option is rejected. Otherwise,
5565 if the assembler does not support them, @option{-gz} is silently ignored
5566 when producing object files.
5570 Produce debugging information in Alpha/VMS debug format (if that is
5571 supported). This is the format used by DEBUG on Alpha/VMS systems.
5574 @itemx -ggdb@var{level}
5575 @itemx -gstabs@var{level}
5576 @itemx -gcoff@var{level}
5577 @itemx -gxcoff@var{level}
5578 @itemx -gvms@var{level}
5579 Request debugging information and also use @var{level} to specify how
5580 much information. The default level is 2.
5582 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5585 Level 1 produces minimal information, enough for making backtraces in
5586 parts of the program that you don't plan to debug. This includes
5587 descriptions of functions and external variables, and line number
5588 tables, but no information about local variables.
5590 Level 3 includes extra information, such as all the macro definitions
5591 present in the program. Some debuggers support macro expansion when
5592 you use @option{-g3}.
5594 @option{-gdwarf-2} does not accept a concatenated debug level, because
5595 GCC used to support an option @option{-gdwarf} that meant to generate
5596 debug information in version 1 of the DWARF format (which is very
5597 different from version 2), and it would have been too confusing. That
5598 debug format is long obsolete, but the option cannot be changed now.
5599 Instead use an additional @option{-g@var{level}} option to change the
5600 debug level for DWARF.
5604 Turn off generation of debug info, if leaving out this option
5605 generates it, or turn it on at level 2 otherwise. The position of this
5606 argument in the command line does not matter; it takes effect after all
5607 other options are processed, and it does so only once, no matter how
5608 many times it is given. This is mainly intended to be used with
5609 @option{-fcompare-debug}.
5611 @item -fsanitize=address
5612 @opindex fsanitize=address
5613 Enable AddressSanitizer, a fast memory error detector.
5614 Memory access instructions are instrumented to detect
5615 out-of-bounds and use-after-free bugs.
5616 See @uref{http://code.google.com/p/address-sanitizer/} for
5617 more details. The run-time behavior can be influenced using the
5618 @env{ASAN_OPTIONS} environment variable; see
5619 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5620 a list of supported options.
5622 @item -fsanitize=kernel-address
5623 @opindex fsanitize=kernel-address
5624 Enable AddressSanitizer for Linux kernel.
5625 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5627 @item -fsanitize=thread
5628 @opindex fsanitize=thread
5629 Enable ThreadSanitizer, a fast data race detector.
5630 Memory access instructions are instrumented to detect
5631 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5632 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5633 environment variable; see
5634 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5637 @item -fsanitize=leak
5638 @opindex fsanitize=leak
5639 Enable LeakSanitizer, a memory leak detector.
5640 This option only matters for linking of executables and if neither
5641 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5642 case the executable is linked against a library that overrides @code{malloc}
5643 and other allocator functions. See
5644 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5645 details. The run-time behavior can be influenced using the
5646 @env{LSAN_OPTIONS} environment variable.
5648 @item -fsanitize=undefined
5649 @opindex fsanitize=undefined
5650 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5651 Various computations are instrumented to detect undefined behavior
5652 at runtime. Current suboptions are:
5656 @item -fsanitize=shift
5657 @opindex fsanitize=shift
5658 This option enables checking that the result of a shift operation is
5659 not undefined. Note that what exactly is considered undefined differs
5660 slightly between C and C++, as well as between ISO C90 and C99, etc.
5662 @item -fsanitize=integer-divide-by-zero
5663 @opindex fsanitize=integer-divide-by-zero
5664 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5666 @item -fsanitize=unreachable
5667 @opindex fsanitize=unreachable
5668 With this option, the compiler turns the @code{__builtin_unreachable}
5669 call into a diagnostics message call instead. When reaching the
5670 @code{__builtin_unreachable} call, the behavior is undefined.
5672 @item -fsanitize=vla-bound
5673 @opindex fsanitize=vla-bound
5674 This option instructs the compiler to check that the size of a variable
5675 length array is positive.
5677 @item -fsanitize=null
5678 @opindex fsanitize=null
5679 This option enables pointer checking. Particularly, the application
5680 built with this option turned on will issue an error message when it
5681 tries to dereference a NULL pointer, or if a reference (possibly an
5682 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5683 on an object pointed by a NULL pointer.
5685 @item -fsanitize=return
5686 @opindex fsanitize=return
5687 This option enables return statement checking. Programs
5688 built with this option turned on will issue an error message
5689 when the end of a non-void function is reached without actually
5690 returning a value. This option works in C++ only.
5692 @item -fsanitize=signed-integer-overflow
5693 @opindex fsanitize=signed-integer-overflow
5694 This option enables signed integer overflow checking. We check that
5695 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5696 does not overflow in the signed arithmetics. Note, integer promotion
5697 rules must be taken into account. That is, the following is not an
5700 signed char a = SCHAR_MAX;
5704 @item -fsanitize=bounds
5705 @opindex fsanitize=bounds
5706 This option enables instrumentation of array bounds. Various out of bounds
5707 accesses are detected. Flexible array members and initializers of variables
5708 with static storage are not instrumented.
5710 @item -fsanitize=alignment
5711 @opindex fsanitize=alignment
5713 This option enables checking of alignment of pointers when they are
5714 dereferenced, or when a reference is bound to insufficiently aligned target,
5715 or when a method or constructor is invoked on insufficiently aligned object.
5717 @item -fsanitize=object-size
5718 @opindex fsanitize=object-size
5719 This option enables instrumentation of memory references using the
5720 @code{__builtin_object_size} function. Various out of bounds pointer
5721 accesses are detected.
5723 @item -fsanitize=float-divide-by-zero
5724 @opindex fsanitize=float-divide-by-zero
5725 Detect floating-point division by zero. Unlike other similar options,
5726 @option{-fsanitize=float-divide-by-zero} is not enabled by
5727 @option{-fsanitize=undefined}, since floating-point division by zero can
5728 be a legitimate way of obtaining infinities and NaNs.
5730 @item -fsanitize=float-cast-overflow
5731 @opindex fsanitize=float-cast-overflow
5732 This option enables floating-point type to integer conversion checking.
5733 We check that the result of the conversion does not overflow.
5734 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
5735 not enabled by @option{-fsanitize=undefined}.
5736 This option does not work well with @code{FE_INVALID} exceptions enabled.
5738 @item -fsanitize=nonnull-attribute
5739 @opindex fsanitize=nonnull-attribute
5741 This option enables instrumentation of calls, checking whether null values
5742 are not passed to arguments marked as requiring a non-null value by the
5743 @code{nonnull} function attribute.
5745 @item -fsanitize=returns-nonnull-attribute
5746 @opindex fsanitize=returns-nonnull-attribute
5748 This option enables instrumentation of return statements in functions
5749 marked with @code{returns_nonnull} function attribute, to detect returning
5750 of null values from such functions.
5752 @item -fsanitize=bool
5753 @opindex fsanitize=bool
5755 This option enables instrumentation of loads from bool. If a value other
5756 than 0/1 is loaded, a run-time error is issued.
5758 @item -fsanitize=enum
5759 @opindex fsanitize=enum
5761 This option enables instrumentation of loads from an enum type. If
5762 a value outside the range of values for the enum type is loaded,
5763 a run-time error is issued.
5765 @item -fsanitize=vptr
5766 @opindex fsanitize=vptr
5768 This option enables instrumentation of C++ member function calls, member
5769 accesses and some conversions between pointers to base and derived classes,
5770 to verify the referenced object has the correct dynamic type.
5774 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5775 @option{-fsanitize=undefined} gives a diagnostic message.
5776 This currently works only for the C family of languages.
5778 @item -fno-sanitize=all
5779 @opindex fno-sanitize=all
5781 This option disables all previously enabled sanitizers.
5782 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
5785 @item -fasan-shadow-offset=@var{number}
5786 @opindex fasan-shadow-offset
5787 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5788 It is useful for experimenting with different shadow memory layouts in
5789 Kernel AddressSanitizer.
5791 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5792 @opindex fsanitize-recover
5793 @opindex fno-sanitize-recover
5794 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5795 mentioned in comma-separated list of @var{opts}. Enabling this option
5796 for a sanitizer component causes it to attempt to continue
5797 running the program as if no error happened. This means multiple
5798 runtime errors can be reported in a single program run, and the exit
5799 code of the program may indicate success even when errors
5800 have been reported. The @option{-fno-sanitize-recover=} can be used to alter
5801 this behavior: only the first detected error is reported
5802 and program then exits with a non-zero exit code.
5804 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5805 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5806 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5807 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5808 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
5809 accepted, the former enables recovery for all sanitizers that support it,
5810 the latter disables recovery for all sanitizers that support it.
5812 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5814 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5817 Similarly @option{-fno-sanitize-recover} is equivalent to
5819 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5822 @item -fsanitize-undefined-trap-on-error
5823 @opindex fsanitize-undefined-trap-on-error
5824 The @option{-fsanitize-undefined-trap-on-error} instructs the compiler to
5825 report undefined behavior using @code{__builtin_trap} rather than
5826 a @code{libubsan} library routine. The advantage of this is that the
5827 @code{libubsan} library is not needed and is not linked in, so this
5828 is usable even in freestanding environments.
5830 @item -fdump-final-insns@r{[}=@var{file}@r{]}
5831 @opindex fdump-final-insns
5832 Dump the final internal representation (RTL) to @var{file}. If the
5833 optional argument is omitted (or if @var{file} is @code{.}), the name
5834 of the dump file is determined by appending @code{.gkd} to the
5835 compilation output file name.
5837 @item -fcompare-debug@r{[}=@var{opts}@r{]}
5838 @opindex fcompare-debug
5839 @opindex fno-compare-debug
5840 If no error occurs during compilation, run the compiler a second time,
5841 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
5842 passed to the second compilation. Dump the final internal
5843 representation in both compilations, and print an error if they differ.
5845 If the equal sign is omitted, the default @option{-gtoggle} is used.
5847 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
5848 and nonzero, implicitly enables @option{-fcompare-debug}. If
5849 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
5850 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
5853 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
5854 is equivalent to @option{-fno-compare-debug}, which disables the dumping
5855 of the final representation and the second compilation, preventing even
5856 @env{GCC_COMPARE_DEBUG} from taking effect.
5858 To verify full coverage during @option{-fcompare-debug} testing, set
5859 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
5860 which GCC rejects as an invalid option in any actual compilation
5861 (rather than preprocessing, assembly or linking). To get just a
5862 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
5863 not overridden} will do.
5865 @item -fcompare-debug-second
5866 @opindex fcompare-debug-second
5867 This option is implicitly passed to the compiler for the second
5868 compilation requested by @option{-fcompare-debug}, along with options to
5869 silence warnings, and omitting other options that would cause
5870 side-effect compiler outputs to files or to the standard output. Dump
5871 files and preserved temporary files are renamed so as to contain the
5872 @code{.gk} additional extension during the second compilation, to avoid
5873 overwriting those generated by the first.
5875 When this option is passed to the compiler driver, it causes the
5876 @emph{first} compilation to be skipped, which makes it useful for little
5877 other than debugging the compiler proper.
5879 @item -feliminate-dwarf2-dups
5880 @opindex feliminate-dwarf2-dups
5881 Compress DWARF 2 debugging information by eliminating duplicated
5882 information about each symbol. This option only makes sense when
5883 generating DWARF 2 debugging information with @option{-gdwarf-2}.
5885 @item -femit-struct-debug-baseonly
5886 @opindex femit-struct-debug-baseonly
5887 Emit debug information for struct-like types
5888 only when the base name of the compilation source file
5889 matches the base name of file in which the struct is defined.
5891 This option substantially reduces the size of debugging information,
5892 but at significant potential loss in type information to the debugger.
5893 See @option{-femit-struct-debug-reduced} for a less aggressive option.
5894 See @option{-femit-struct-debug-detailed} for more detailed control.
5896 This option works only with DWARF 2.
5898 @item -femit-struct-debug-reduced
5899 @opindex femit-struct-debug-reduced
5900 Emit debug information for struct-like types
5901 only when the base name of the compilation source file
5902 matches the base name of file in which the type is defined,
5903 unless the struct is a template or defined in a system header.
5905 This option significantly reduces the size of debugging information,
5906 with some potential loss in type information to the debugger.
5907 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
5908 See @option{-femit-struct-debug-detailed} for more detailed control.
5910 This option works only with DWARF 2.
5912 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
5913 @opindex femit-struct-debug-detailed
5914 Specify the struct-like types
5915 for which the compiler generates debug information.
5916 The intent is to reduce duplicate struct debug information
5917 between different object files within the same program.
5919 This option is a detailed version of
5920 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
5921 which serves for most needs.
5923 A specification has the syntax@*
5924 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
5926 The optional first word limits the specification to
5927 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
5928 A struct type is used directly when it is the type of a variable, member.
5929 Indirect uses arise through pointers to structs.
5930 That is, when use of an incomplete struct is valid, the use is indirect.
5932 @samp{struct one direct; struct two * indirect;}.
5934 The optional second word limits the specification to
5935 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
5936 Generic structs are a bit complicated to explain.
5937 For C++, these are non-explicit specializations of template classes,
5938 or non-template classes within the above.
5939 Other programming languages have generics,
5940 but @option{-femit-struct-debug-detailed} does not yet implement them.
5942 The third word specifies the source files for those
5943 structs for which the compiler should emit debug information.
5944 The values @samp{none} and @samp{any} have the normal meaning.
5945 The value @samp{base} means that
5946 the base of name of the file in which the type declaration appears
5947 must match the base of the name of the main compilation file.
5948 In practice, this means that when compiling @file{foo.c}, debug information
5949 is generated for types declared in that file and @file{foo.h},
5950 but not other header files.
5951 The value @samp{sys} means those types satisfying @samp{base}
5952 or declared in system or compiler headers.
5954 You may need to experiment to determine the best settings for your application.
5956 The default is @option{-femit-struct-debug-detailed=all}.
5958 This option works only with DWARF 2.
5960 @item -fno-merge-debug-strings
5961 @opindex fmerge-debug-strings
5962 @opindex fno-merge-debug-strings
5963 Direct the linker to not merge together strings in the debugging
5964 information that are identical in different object files. Merging is
5965 not supported by all assemblers or linkers. Merging decreases the size
5966 of the debug information in the output file at the cost of increasing
5967 link processing time. Merging is enabled by default.
5969 @item -fdebug-prefix-map=@var{old}=@var{new}
5970 @opindex fdebug-prefix-map
5971 When compiling files in directory @file{@var{old}}, record debugging
5972 information describing them as in @file{@var{new}} instead.
5974 @item -fno-dwarf2-cfi-asm
5975 @opindex fdwarf2-cfi-asm
5976 @opindex fno-dwarf2-cfi-asm
5977 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5978 instead of using GAS @code{.cfi_*} directives.
5980 @cindex @command{prof}
5983 Generate extra code to write profile information suitable for the
5984 analysis program @command{prof}. You must use this option when compiling
5985 the source files you want data about, and you must also use it when
5988 @cindex @command{gprof}
5991 Generate extra code to write profile information suitable for the
5992 analysis program @command{gprof}. You must use this option when compiling
5993 the source files you want data about, and you must also use it when
5998 Makes the compiler print out each function name as it is compiled, and
5999 print some statistics about each pass when it finishes.
6002 @opindex ftime-report
6003 Makes the compiler print some statistics about the time consumed by each
6004 pass when it finishes.
6007 @opindex fmem-report
6008 Makes the compiler print some statistics about permanent memory
6009 allocation when it finishes.
6011 @item -fmem-report-wpa
6012 @opindex fmem-report-wpa
6013 Makes the compiler print some statistics about permanent memory
6014 allocation for the WPA phase only.
6016 @item -fpre-ipa-mem-report
6017 @opindex fpre-ipa-mem-report
6018 @item -fpost-ipa-mem-report
6019 @opindex fpost-ipa-mem-report
6020 Makes the compiler print some statistics about permanent memory
6021 allocation before or after interprocedural optimization.
6023 @item -fprofile-report
6024 @opindex fprofile-report
6025 Makes the compiler print some statistics about consistency of the
6026 (estimated) profile and effect of individual passes.
6029 @opindex fstack-usage
6030 Makes the compiler output stack usage information for the program, on a
6031 per-function basis. The filename for the dump is made by appending
6032 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6033 the output file, if explicitly specified and it is not an executable,
6034 otherwise it is the basename of the source file. An entry is made up
6039 The name of the function.
6043 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6046 The qualifier @code{static} means that the function manipulates the stack
6047 statically: a fixed number of bytes are allocated for the frame on function
6048 entry and released on function exit; no stack adjustments are otherwise made
6049 in the function. The second field is this fixed number of bytes.
6051 The qualifier @code{dynamic} means that the function manipulates the stack
6052 dynamically: in addition to the static allocation described above, stack
6053 adjustments are made in the body of the function, for example to push/pop
6054 arguments around function calls. If the qualifier @code{bounded} is also
6055 present, the amount of these adjustments is bounded at compile time and
6056 the second field is an upper bound of the total amount of stack used by
6057 the function. If it is not present, the amount of these adjustments is
6058 not bounded at compile time and the second field only represents the
6061 @item -fprofile-arcs
6062 @opindex fprofile-arcs
6063 Add code so that program flow @dfn{arcs} are instrumented. During
6064 execution the program records how many times each branch and call is
6065 executed and how many times it is taken or returns. When the compiled
6066 program exits it saves this data to a file called
6067 @file{@var{auxname}.gcda} for each source file. The data may be used for
6068 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6069 test coverage analysis (@option{-ftest-coverage}). Each object file's
6070 @var{auxname} is generated from the name of the output file, if
6071 explicitly specified and it is not the final executable, otherwise it is
6072 the basename of the source file. In both cases any suffix is removed
6073 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6074 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6075 @xref{Cross-profiling}.
6077 @cindex @command{gcov}
6081 This option is used to compile and link code instrumented for coverage
6082 analysis. The option is a synonym for @option{-fprofile-arcs}
6083 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6084 linking). See the documentation for those options for more details.
6089 Compile the source files with @option{-fprofile-arcs} plus optimization
6090 and code generation options. For test coverage analysis, use the
6091 additional @option{-ftest-coverage} option. You do not need to profile
6092 every source file in a program.
6095 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6096 (the latter implies the former).
6099 Run the program on a representative workload to generate the arc profile
6100 information. This may be repeated any number of times. You can run
6101 concurrent instances of your program, and provided that the file system
6102 supports locking, the data files will be correctly updated. Also
6103 @code{fork} calls are detected and correctly handled (double counting
6107 For profile-directed optimizations, compile the source files again with
6108 the same optimization and code generation options plus
6109 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6110 Control Optimization}).
6113 For test coverage analysis, use @command{gcov} to produce human readable
6114 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6115 @command{gcov} documentation for further information.
6119 With @option{-fprofile-arcs}, for each function of your program GCC
6120 creates a program flow graph, then finds a spanning tree for the graph.
6121 Only arcs that are not on the spanning tree have to be instrumented: the
6122 compiler adds code to count the number of times that these arcs are
6123 executed. When an arc is the only exit or only entrance to a block, the
6124 instrumentation code can be added to the block; otherwise, a new basic
6125 block must be created to hold the instrumentation code.
6128 @item -ftest-coverage
6129 @opindex ftest-coverage
6130 Produce a notes file that the @command{gcov} code-coverage utility
6131 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6132 show program coverage. Each source file's note file is called
6133 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6134 above for a description of @var{auxname} and instructions on how to
6135 generate test coverage data. Coverage data matches the source files
6136 more closely if you do not optimize.
6138 @item -fdbg-cnt-list
6139 @opindex fdbg-cnt-list
6140 Print the name and the counter upper bound for all debug counters.
6143 @item -fdbg-cnt=@var{counter-value-list}
6145 Set the internal debug counter upper bound. @var{counter-value-list}
6146 is a comma-separated list of @var{name}:@var{value} pairs
6147 which sets the upper bound of each debug counter @var{name} to @var{value}.
6148 All debug counters have the initial upper bound of @code{UINT_MAX};
6149 thus @code{dbg_cnt} returns true always unless the upper bound
6150 is set by this option.
6151 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6152 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6154 @item -fenable-@var{kind}-@var{pass}
6155 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6159 This is a set of options that are used to explicitly disable/enable
6160 optimization passes. These options are intended for use for debugging GCC.
6161 Compiler users should use regular options for enabling/disabling
6166 @item -fdisable-ipa-@var{pass}
6167 Disable 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 -fdisable-rtl-@var{pass}
6172 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6173 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6174 statically invoked in the compiler multiple times, the pass name should be
6175 appended with a sequential number starting from 1. @var{range-list} is a
6176 comma-separated list of function ranges or assembler names. Each range is a number
6177 pair separated by a colon. The range is inclusive in both ends. If the range
6178 is trivial, the number pair can be simplified as a single number. If the
6179 function's call graph node's @var{uid} falls within one of the specified ranges,
6180 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6181 function header of a dump file, and the pass names can be dumped by using
6182 option @option{-fdump-passes}.
6184 @item -fdisable-tree-@var{pass}
6185 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6186 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6189 @item -fenable-ipa-@var{pass}
6190 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6191 statically invoked in the compiler multiple times, the pass name should be
6192 appended with a sequential number starting from 1.
6194 @item -fenable-rtl-@var{pass}
6195 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6196 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6197 description and examples.
6199 @item -fenable-tree-@var{pass}
6200 @itemx -fenable-tree-@var{pass}=@var{range-list}
6201 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6202 of option arguments.
6206 Here are some examples showing uses of these options.
6210 # disable ccp1 for all functions
6212 # disable complete unroll for function whose cgraph node uid is 1
6213 -fenable-tree-cunroll=1
6214 # disable gcse2 for functions at the following ranges [1,1],
6215 # [300,400], and [400,1000]
6216 # disable gcse2 for functions foo and foo2
6217 -fdisable-rtl-gcse2=foo,foo2
6218 # disable early inlining
6219 -fdisable-tree-einline
6220 # disable ipa inlining
6221 -fdisable-ipa-inline
6222 # enable tree full unroll
6223 -fenable-tree-unroll
6227 @item -d@var{letters}
6228 @itemx -fdump-rtl-@var{pass}
6229 @itemx -fdump-rtl-@var{pass}=@var{filename}
6231 @opindex fdump-rtl-@var{pass}
6232 Says to make debugging dumps during compilation at times specified by
6233 @var{letters}. This is used for debugging the RTL-based passes of the
6234 compiler. The file names for most of the dumps are made by appending
6235 a pass number and a word to the @var{dumpname}, and the files are
6236 created in the directory of the output file. In case of
6237 @option{=@var{filename}} option, the dump is output on the given file
6238 instead of the pass numbered dump files. Note that the pass number is
6239 computed statically as passes get registered into the pass manager.
6240 Thus the numbering is not related to the dynamic order of execution of
6241 passes. In particular, a pass installed by a plugin could have a
6242 number over 200 even if it executed quite early. @var{dumpname} is
6243 generated from the name of the output file, if explicitly specified
6244 and it is not an executable, otherwise it is the basename of the
6245 source file. These switches may have different effects when
6246 @option{-E} is used for preprocessing.
6248 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6249 @option{-d} option @var{letters}. Here are the possible
6250 letters for use in @var{pass} and @var{letters}, and their meanings:
6254 @item -fdump-rtl-alignments
6255 @opindex fdump-rtl-alignments
6256 Dump after branch alignments have been computed.
6258 @item -fdump-rtl-asmcons
6259 @opindex fdump-rtl-asmcons
6260 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6262 @item -fdump-rtl-auto_inc_dec
6263 @opindex fdump-rtl-auto_inc_dec
6264 Dump after auto-inc-dec discovery. This pass is only run on
6265 architectures that have auto inc or auto dec instructions.
6267 @item -fdump-rtl-barriers
6268 @opindex fdump-rtl-barriers
6269 Dump after cleaning up the barrier instructions.
6271 @item -fdump-rtl-bbpart
6272 @opindex fdump-rtl-bbpart
6273 Dump after partitioning hot and cold basic blocks.
6275 @item -fdump-rtl-bbro
6276 @opindex fdump-rtl-bbro
6277 Dump after block reordering.
6279 @item -fdump-rtl-btl1
6280 @itemx -fdump-rtl-btl2
6281 @opindex fdump-rtl-btl2
6282 @opindex fdump-rtl-btl2
6283 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6284 after the two branch
6285 target load optimization passes.
6287 @item -fdump-rtl-bypass
6288 @opindex fdump-rtl-bypass
6289 Dump after jump bypassing and control flow optimizations.
6291 @item -fdump-rtl-combine
6292 @opindex fdump-rtl-combine
6293 Dump after the RTL instruction combination pass.
6295 @item -fdump-rtl-compgotos
6296 @opindex fdump-rtl-compgotos
6297 Dump after duplicating the computed gotos.
6299 @item -fdump-rtl-ce1
6300 @itemx -fdump-rtl-ce2
6301 @itemx -fdump-rtl-ce3
6302 @opindex fdump-rtl-ce1
6303 @opindex fdump-rtl-ce2
6304 @opindex fdump-rtl-ce3
6305 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6306 @option{-fdump-rtl-ce3} enable dumping after the three
6307 if conversion passes.
6309 @item -fdump-rtl-cprop_hardreg
6310 @opindex fdump-rtl-cprop_hardreg
6311 Dump after hard register copy propagation.
6313 @item -fdump-rtl-csa
6314 @opindex fdump-rtl-csa
6315 Dump after combining stack adjustments.
6317 @item -fdump-rtl-cse1
6318 @itemx -fdump-rtl-cse2
6319 @opindex fdump-rtl-cse1
6320 @opindex fdump-rtl-cse2
6321 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6322 the two common subexpression elimination passes.
6324 @item -fdump-rtl-dce
6325 @opindex fdump-rtl-dce
6326 Dump after the standalone dead code elimination passes.
6328 @item -fdump-rtl-dbr
6329 @opindex fdump-rtl-dbr
6330 Dump after delayed branch scheduling.
6332 @item -fdump-rtl-dce1
6333 @itemx -fdump-rtl-dce2
6334 @opindex fdump-rtl-dce1
6335 @opindex fdump-rtl-dce2
6336 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6337 the two dead store elimination passes.
6340 @opindex fdump-rtl-eh
6341 Dump after finalization of EH handling code.
6343 @item -fdump-rtl-eh_ranges
6344 @opindex fdump-rtl-eh_ranges
6345 Dump after conversion of EH handling range regions.
6347 @item -fdump-rtl-expand
6348 @opindex fdump-rtl-expand
6349 Dump after RTL generation.
6351 @item -fdump-rtl-fwprop1
6352 @itemx -fdump-rtl-fwprop2
6353 @opindex fdump-rtl-fwprop1
6354 @opindex fdump-rtl-fwprop2
6355 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6356 dumping after the two forward propagation passes.
6358 @item -fdump-rtl-gcse1
6359 @itemx -fdump-rtl-gcse2
6360 @opindex fdump-rtl-gcse1
6361 @opindex fdump-rtl-gcse2
6362 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6363 after global common subexpression elimination.
6365 @item -fdump-rtl-init-regs
6366 @opindex fdump-rtl-init-regs
6367 Dump after the initialization of the registers.
6369 @item -fdump-rtl-initvals
6370 @opindex fdump-rtl-initvals
6371 Dump after the computation of the initial value sets.
6373 @item -fdump-rtl-into_cfglayout
6374 @opindex fdump-rtl-into_cfglayout
6375 Dump after converting to cfglayout mode.
6377 @item -fdump-rtl-ira
6378 @opindex fdump-rtl-ira
6379 Dump after iterated register allocation.
6381 @item -fdump-rtl-jump
6382 @opindex fdump-rtl-jump
6383 Dump after the second jump optimization.
6385 @item -fdump-rtl-loop2
6386 @opindex fdump-rtl-loop2
6387 @option{-fdump-rtl-loop2} enables dumping after the rtl
6388 loop optimization passes.
6390 @item -fdump-rtl-mach
6391 @opindex fdump-rtl-mach
6392 Dump after performing the machine dependent reorganization pass, if that
6395 @item -fdump-rtl-mode_sw
6396 @opindex fdump-rtl-mode_sw
6397 Dump after removing redundant mode switches.
6399 @item -fdump-rtl-rnreg
6400 @opindex fdump-rtl-rnreg
6401 Dump after register renumbering.
6403 @item -fdump-rtl-outof_cfglayout
6404 @opindex fdump-rtl-outof_cfglayout
6405 Dump after converting from cfglayout mode.
6407 @item -fdump-rtl-peephole2
6408 @opindex fdump-rtl-peephole2
6409 Dump after the peephole pass.
6411 @item -fdump-rtl-postreload
6412 @opindex fdump-rtl-postreload
6413 Dump after post-reload optimizations.
6415 @item -fdump-rtl-pro_and_epilogue
6416 @opindex fdump-rtl-pro_and_epilogue
6417 Dump after generating the function prologues and epilogues.
6419 @item -fdump-rtl-sched1
6420 @itemx -fdump-rtl-sched2
6421 @opindex fdump-rtl-sched1
6422 @opindex fdump-rtl-sched2
6423 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6424 after the basic block scheduling passes.
6426 @item -fdump-rtl-ree
6427 @opindex fdump-rtl-ree
6428 Dump after sign/zero extension elimination.
6430 @item -fdump-rtl-seqabstr
6431 @opindex fdump-rtl-seqabstr
6432 Dump after common sequence discovery.
6434 @item -fdump-rtl-shorten
6435 @opindex fdump-rtl-shorten
6436 Dump after shortening branches.
6438 @item -fdump-rtl-sibling
6439 @opindex fdump-rtl-sibling
6440 Dump after sibling call optimizations.
6442 @item -fdump-rtl-split1
6443 @itemx -fdump-rtl-split2
6444 @itemx -fdump-rtl-split3
6445 @itemx -fdump-rtl-split4
6446 @itemx -fdump-rtl-split5
6447 @opindex fdump-rtl-split1
6448 @opindex fdump-rtl-split2
6449 @opindex fdump-rtl-split3
6450 @opindex fdump-rtl-split4
6451 @opindex fdump-rtl-split5
6452 These options enable dumping after five rounds of
6453 instruction splitting.
6455 @item -fdump-rtl-sms
6456 @opindex fdump-rtl-sms
6457 Dump after modulo scheduling. This pass is only run on some
6460 @item -fdump-rtl-stack
6461 @opindex fdump-rtl-stack
6462 Dump after conversion from GCC's ``flat register file'' registers to the
6463 x87's stack-like registers. This pass is only run on x86 variants.
6465 @item -fdump-rtl-subreg1
6466 @itemx -fdump-rtl-subreg2
6467 @opindex fdump-rtl-subreg1
6468 @opindex fdump-rtl-subreg2
6469 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6470 the two subreg expansion passes.
6472 @item -fdump-rtl-unshare
6473 @opindex fdump-rtl-unshare
6474 Dump after all rtl has been unshared.
6476 @item -fdump-rtl-vartrack
6477 @opindex fdump-rtl-vartrack
6478 Dump after variable tracking.
6480 @item -fdump-rtl-vregs
6481 @opindex fdump-rtl-vregs
6482 Dump after converting virtual registers to hard registers.
6484 @item -fdump-rtl-web
6485 @opindex fdump-rtl-web
6486 Dump after live range splitting.
6488 @item -fdump-rtl-regclass
6489 @itemx -fdump-rtl-subregs_of_mode_init
6490 @itemx -fdump-rtl-subregs_of_mode_finish
6491 @itemx -fdump-rtl-dfinit
6492 @itemx -fdump-rtl-dfinish
6493 @opindex fdump-rtl-regclass
6494 @opindex fdump-rtl-subregs_of_mode_init
6495 @opindex fdump-rtl-subregs_of_mode_finish
6496 @opindex fdump-rtl-dfinit
6497 @opindex fdump-rtl-dfinish
6498 These dumps are defined but always produce empty files.
6501 @itemx -fdump-rtl-all
6503 @opindex fdump-rtl-all
6504 Produce all the dumps listed above.
6508 Annotate the assembler output with miscellaneous debugging information.
6512 Dump all macro definitions, at the end of preprocessing, in addition to
6517 Produce a core dump whenever an error occurs.
6521 Annotate the assembler output with a comment indicating which
6522 pattern and alternative is used. The length of each instruction is
6527 Dump the RTL in the assembler output as a comment before each instruction.
6528 Also turns on @option{-dp} annotation.
6532 Just generate RTL for a function instead of compiling it. Usually used
6533 with @option{-fdump-rtl-expand}.
6537 @opindex fdump-noaddr
6538 When doing debugging dumps, suppress address output. This makes it more
6539 feasible to use diff on debugging dumps for compiler invocations with
6540 different compiler binaries and/or different
6541 text / bss / data / heap / stack / dso start locations.
6544 @opindex freport-bug
6545 Collect and dump debug information into temporary file if ICE in C/C++
6548 @item -fdump-unnumbered
6549 @opindex fdump-unnumbered
6550 When doing debugging dumps, suppress instruction numbers and address output.
6551 This makes it more feasible to use diff on debugging dumps for compiler
6552 invocations with different options, in particular with and without
6555 @item -fdump-unnumbered-links
6556 @opindex fdump-unnumbered-links
6557 When doing debugging dumps (see @option{-d} option above), suppress
6558 instruction numbers for the links to the previous and next instructions
6561 @item -fdump-translation-unit @r{(C++ only)}
6562 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6563 @opindex fdump-translation-unit
6564 Dump a representation of the tree structure for the entire translation
6565 unit to a file. The file name is made by appending @file{.tu} to the
6566 source file name, and the file is created in the same directory as the
6567 output file. If the @samp{-@var{options}} form is used, @var{options}
6568 controls the details of the dump as described for the
6569 @option{-fdump-tree} options.
6571 @item -fdump-class-hierarchy @r{(C++ only)}
6572 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6573 @opindex fdump-class-hierarchy
6574 Dump a representation of each class's hierarchy and virtual function
6575 table layout to a file. The file name is made by appending
6576 @file{.class} to the source file name, and the file is created in the
6577 same directory as the output file. If the @samp{-@var{options}} form
6578 is used, @var{options} controls the details of the dump as described
6579 for the @option{-fdump-tree} options.
6581 @item -fdump-ipa-@var{switch}
6583 Control the dumping at various stages of inter-procedural analysis
6584 language tree to a file. The file name is generated by appending a
6585 switch specific suffix to the source file name, and the file is created
6586 in the same directory as the output file. The following dumps are
6591 Enables all inter-procedural analysis dumps.
6594 Dumps information about call-graph optimization, unused function removal,
6595 and inlining decisions.
6598 Dump after function inlining.
6603 @opindex fdump-passes
6604 Dump the list of optimization passes that are turned on and off by
6605 the current command-line options.
6607 @item -fdump-statistics-@var{option}
6608 @opindex fdump-statistics
6609 Enable and control dumping of pass statistics in a separate file. The
6610 file name is generated by appending a suffix ending in
6611 @samp{.statistics} to the source file name, and the file is created in
6612 the same directory as the output file. If the @samp{-@var{option}}
6613 form is used, @samp{-stats} causes counters to be summed over the
6614 whole compilation unit while @samp{-details} dumps every event as
6615 the passes generate them. The default with no option is to sum
6616 counters for each function compiled.
6618 @item -fdump-tree-@var{switch}
6619 @itemx -fdump-tree-@var{switch}-@var{options}
6620 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6622 Control the dumping at various stages of processing the intermediate
6623 language tree to a file. The file name is generated by appending a
6624 switch-specific suffix to the source file name, and the file is
6625 created in the same directory as the output file. In case of
6626 @option{=@var{filename}} option, the dump is output on the given file
6627 instead of the auto named dump files. If the @samp{-@var{options}}
6628 form is used, @var{options} is a list of @samp{-} separated options
6629 which control the details of the dump. Not all options are applicable
6630 to all dumps; those that are not meaningful are ignored. The
6631 following options are available
6635 Print the address of each node. Usually this is not meaningful as it
6636 changes according to the environment and source file. Its primary use
6637 is for tying up a dump file with a debug environment.
6639 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6640 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6641 use working backward from mangled names in the assembly file.
6643 When dumping front-end intermediate representations, inhibit dumping
6644 of members of a scope or body of a function merely because that scope
6645 has been reached. Only dump such items when they are directly reachable
6648 When dumping pretty-printed trees, this option inhibits dumping the
6649 bodies of control structures.
6651 When dumping RTL, print the RTL in slim (condensed) form instead of
6652 the default LISP-like representation.
6654 Print a raw representation of the tree. By default, trees are
6655 pretty-printed into a C-like representation.
6657 Enable more detailed dumps (not honored by every dump option). Also
6658 include information from the optimization passes.
6660 Enable dumping various statistics about the pass (not honored by every dump
6663 Enable showing basic block boundaries (disabled in raw dumps).
6665 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6666 dump a representation of the control flow graph suitable for viewing with
6667 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6668 the file is pretty-printed as a subgraph, so that GraphViz can render them
6669 all in a single plot.
6671 This option currently only works for RTL dumps, and the RTL is always
6672 dumped in slim form.
6674 Enable showing virtual operands for every statement.
6676 Enable showing line numbers for statements.
6678 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6680 Enable showing the tree dump for each statement.
6682 Enable showing the EH region number holding each statement.
6684 Enable showing scalar evolution analysis details.
6686 Enable showing optimization information (only available in certain
6689 Enable showing missed optimization information (only available in certain
6692 Enable other detailed optimization information (only available in
6694 @item =@var{filename}
6695 Instead of an auto named dump file, output into the given file
6696 name. The file names @file{stdout} and @file{stderr} are treated
6697 specially and are considered already open standard streams. For
6701 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6702 -fdump-tree-pre=stderr file.c
6705 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6706 output on to @file{stderr}. If two conflicting dump filenames are
6707 given for the same pass, then the latter option overrides the earlier
6711 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6712 and @option{lineno}.
6715 Turn on all optimization options, i.e., @option{optimized},
6716 @option{missed}, and @option{note}.
6719 The following tree dumps are possible:
6723 @opindex fdump-tree-original
6724 Dump before any tree based optimization, to @file{@var{file}.original}.
6727 @opindex fdump-tree-optimized
6728 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6731 @opindex fdump-tree-gimple
6732 Dump each function before and after the gimplification pass to a file. The
6733 file name is made by appending @file{.gimple} to the source file name.
6736 @opindex fdump-tree-cfg
6737 Dump the control flow graph of each function to a file. The file name is
6738 made by appending @file{.cfg} to the source file name.
6741 @opindex fdump-tree-ch
6742 Dump each function after copying loop headers. The file name is made by
6743 appending @file{.ch} to the source file name.
6746 @opindex fdump-tree-ssa
6747 Dump SSA related information to a file. The file name is made by appending
6748 @file{.ssa} to the source file name.
6751 @opindex fdump-tree-alias
6752 Dump aliasing information for each function. The file name is made by
6753 appending @file{.alias} to the source file name.
6756 @opindex fdump-tree-ccp
6757 Dump each function after CCP@. The file name is made by appending
6758 @file{.ccp} to the source file name.
6761 @opindex fdump-tree-storeccp
6762 Dump each function after STORE-CCP@. The file name is made by appending
6763 @file{.storeccp} to the source file name.
6766 @opindex fdump-tree-pre
6767 Dump trees after partial redundancy elimination. The file name is made
6768 by appending @file{.pre} to the source file name.
6771 @opindex fdump-tree-fre
6772 Dump trees after full redundancy elimination. The file name is made
6773 by appending @file{.fre} to the source file name.
6776 @opindex fdump-tree-copyprop
6777 Dump trees after copy propagation. The file name is made
6778 by appending @file{.copyprop} to the source file name.
6780 @item store_copyprop
6781 @opindex fdump-tree-store_copyprop
6782 Dump trees after store copy-propagation. The file name is made
6783 by appending @file{.store_copyprop} to the source file name.
6786 @opindex fdump-tree-dce
6787 Dump each function after dead code elimination. The file name is made by
6788 appending @file{.dce} to the source file name.
6791 @opindex fdump-tree-sra
6792 Dump each function after performing scalar replacement of aggregates. The
6793 file name is made by appending @file{.sra} to the source file name.
6796 @opindex fdump-tree-sink
6797 Dump each function after performing code sinking. The file name is made
6798 by appending @file{.sink} to the source file name.
6801 @opindex fdump-tree-dom
6802 Dump each function after applying dominator tree optimizations. The file
6803 name is made by appending @file{.dom} to the source file name.
6806 @opindex fdump-tree-dse
6807 Dump each function after applying dead store elimination. The file
6808 name is made by appending @file{.dse} to the source file name.
6811 @opindex fdump-tree-phiopt
6812 Dump each function after optimizing PHI nodes into straightline code. The file
6813 name is made by appending @file{.phiopt} to the source file name.
6816 @opindex fdump-tree-forwprop
6817 Dump each function after forward propagating single use variables. The file
6818 name is made by appending @file{.forwprop} to the source file name.
6821 @opindex fdump-tree-copyrename
6822 Dump each function after applying the copy rename optimization. The file
6823 name is made by appending @file{.copyrename} to the source file name.
6826 @opindex fdump-tree-nrv
6827 Dump each function after applying the named return value optimization on
6828 generic trees. The file name is made by appending @file{.nrv} to the source
6832 @opindex fdump-tree-vect
6833 Dump each function after applying vectorization of loops. The file name is
6834 made by appending @file{.vect} to the source file name.
6837 @opindex fdump-tree-slp
6838 Dump each function after applying vectorization of basic blocks. The file name
6839 is made by appending @file{.slp} to the source file name.
6842 @opindex fdump-tree-vrp
6843 Dump each function after Value Range Propagation (VRP). The file name
6844 is made by appending @file{.vrp} to the source file name.
6847 @opindex fdump-tree-all
6848 Enable all the available tree dumps with the flags provided in this option.
6852 @itemx -fopt-info-@var{options}
6853 @itemx -fopt-info-@var{options}=@var{filename}
6855 Controls optimization dumps from various optimization passes. If the
6856 @samp{-@var{options}} form is used, @var{options} is a list of
6857 @samp{-} separated option keywords to select the dump details and
6860 The @var{options} can be divided into two groups: options describing the
6861 verbosity of the dump, and options describing which optimizations
6862 should be included. The options from both the groups can be freely
6863 mixed as they are non-overlapping. However, in case of any conflicts,
6864 the later options override the earlier options on the command
6867 The following options control the dump verbosity:
6871 Print information when an optimization is successfully applied. It is
6872 up to a pass to decide which information is relevant. For example, the
6873 vectorizer passes print the source location of loops which are
6874 successfully vectorized.
6876 Print information about missed optimizations. Individual passes
6877 control which information to include in the output.
6879 Print verbose information about optimizations, such as certain
6880 transformations, more detailed messages about decisions etc.
6882 Print detailed optimization information. This includes
6883 @samp{optimized}, @samp{missed}, and @samp{note}.
6886 One or more of the following option keywords can be used to describe a
6887 group of optimizations:
6891 Enable dumps from all interprocedural optimizations.
6893 Enable dumps from all loop optimizations.
6895 Enable dumps from all inlining optimizations.
6897 Enable dumps from all vectorization optimizations.
6899 Enable dumps from all optimizations. This is a superset of
6900 the optimization groups listed above.
6904 omitted, it defaults to @samp{optimized-optall}, which means to dump all
6905 info about successful optimizations from all the passes.
6907 If the @var{filename} is provided, then the dumps from all the
6908 applicable optimizations are concatenated into the @var{filename}.
6909 Otherwise the dump is output onto @file{stderr}. Though multiple
6910 @option{-fopt-info} options are accepted, only one of them can include
6911 a @var{filename}. If other filenames are provided then all but the
6912 first such option are ignored.
6914 Note that the output @var{filename} is overwritten
6915 in case of multiple translation units. If a combined output from
6916 multiple translation units is desired, @file{stderr} should be used
6919 In the following example, the optimization info is output to
6928 gcc -O3 -fopt-info-missed=missed.all
6932 outputs missed optimization report from all the passes into
6933 @file{missed.all}, and this one:
6936 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
6940 prints information about missed optimization opportunities from
6941 vectorization passes on @file{stderr}.
6942 Note that @option{-fopt-info-vec-missed} is equivalent to
6943 @option{-fopt-info-missed-vec}.
6947 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
6951 outputs information about missed optimizations as well as
6952 optimized locations from all the inlining passes into
6958 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
6962 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
6963 in conflict since only one output file is allowed. In this case, only
6964 the first option takes effect and the subsequent options are
6965 ignored. Thus only @file{vec.miss} is produced which contains
6966 dumps from the vectorizer about missed opportunities.
6968 @item -frandom-seed=@var{number}
6969 @opindex frandom-seed
6970 This option provides a seed that GCC uses in place of
6971 random numbers in generating certain symbol names
6972 that have to be different in every compiled file. It is also used to
6973 place unique stamps in coverage data files and the object files that
6974 produce them. You can use the @option{-frandom-seed} option to produce
6975 reproducibly identical object files.
6977 The @var{number} should be different for every file you compile.
6979 @item -fsched-verbose=@var{n}
6980 @opindex fsched-verbose
6981 On targets that use instruction scheduling, this option controls the
6982 amount of debugging output the scheduler prints. This information is
6983 written to standard error, unless @option{-fdump-rtl-sched1} or
6984 @option{-fdump-rtl-sched2} is specified, in which case it is output
6985 to the usual dump listing file, @file{.sched1} or @file{.sched2}
6986 respectively. However for @var{n} greater than nine, the output is
6987 always printed to standard error.
6989 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
6990 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
6991 For @var{n} greater than one, it also output basic block probabilities,
6992 detailed ready list information and unit/insn info. For @var{n} greater
6993 than two, it includes RTL at abort point, control-flow and regions info.
6994 And for @var{n} over four, @option{-fsched-verbose} also includes
6998 @itemx -save-temps=cwd
7000 Store the usual ``temporary'' intermediate files permanently; place them
7001 in the current directory and name them based on the source file. Thus,
7002 compiling @file{foo.c} with @option{-c -save-temps} produces files
7003 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7004 preprocessed @file{foo.i} output file even though the compiler now
7005 normally uses an integrated preprocessor.
7007 When used in combination with the @option{-x} command-line option,
7008 @option{-save-temps} is sensible enough to avoid over writing an
7009 input source file with the same extension as an intermediate file.
7010 The corresponding intermediate file may be obtained by renaming the
7011 source file before using @option{-save-temps}.
7013 If you invoke GCC in parallel, compiling several different source
7014 files that share a common base name in different subdirectories or the
7015 same source file compiled for multiple output destinations, it is
7016 likely that the different parallel compilers will interfere with each
7017 other, and overwrite the temporary files. For instance:
7020 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7021 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7024 may result in @file{foo.i} and @file{foo.o} being written to
7025 simultaneously by both compilers.
7027 @item -save-temps=obj
7028 @opindex save-temps=obj
7029 Store the usual ``temporary'' intermediate files permanently. If the
7030 @option{-o} option is used, the temporary files are based on the
7031 object file. If the @option{-o} option is not used, the
7032 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7037 gcc -save-temps=obj -c foo.c
7038 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7039 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7043 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7044 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7045 @file{dir2/yfoobar.o}.
7047 @item -time@r{[}=@var{file}@r{]}
7049 Report the CPU time taken by each subprocess in the compilation
7050 sequence. For C source files, this is the compiler proper and assembler
7051 (plus the linker if linking is done).
7053 Without the specification of an output file, the output looks like this:
7060 The first number on each line is the ``user time'', that is time spent
7061 executing the program itself. The second number is ``system time'',
7062 time spent executing operating system routines on behalf of the program.
7063 Both numbers are in seconds.
7065 With the specification of an output file, the output is appended to the
7066 named file, and it looks like this:
7069 0.12 0.01 cc1 @var{options}
7070 0.00 0.01 as @var{options}
7073 The ``user time'' and the ``system time'' are moved before the program
7074 name, and the options passed to the program are displayed, so that one
7075 can later tell what file was being compiled, and with which options.
7077 @item -fvar-tracking
7078 @opindex fvar-tracking
7079 Run variable tracking pass. It computes where variables are stored at each
7080 position in code. Better debugging information is then generated
7081 (if the debugging information format supports this information).
7083 It is enabled by default when compiling with optimization (@option{-Os},
7084 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7085 the debug info format supports it.
7087 @item -fvar-tracking-assignments
7088 @opindex fvar-tracking-assignments
7089 @opindex fno-var-tracking-assignments
7090 Annotate assignments to user variables early in the compilation and
7091 attempt to carry the annotations over throughout the compilation all the
7092 way to the end, in an attempt to improve debug information while
7093 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7095 It can be enabled even if var-tracking is disabled, in which case
7096 annotations are created and maintained, but discarded at the end.
7097 By default, this flag is enabled together with @option{-fvar-tracking},
7098 except when selective scheduling is enabled.
7100 @item -fvar-tracking-assignments-toggle
7101 @opindex fvar-tracking-assignments-toggle
7102 @opindex fno-var-tracking-assignments-toggle
7103 Toggle @option{-fvar-tracking-assignments}, in the same way that
7104 @option{-gtoggle} toggles @option{-g}.
7106 @item -print-file-name=@var{library}
7107 @opindex print-file-name
7108 Print the full absolute name of the library file @var{library} that
7109 would be used when linking---and don't do anything else. With this
7110 option, GCC does not compile or link anything; it just prints the
7113 @item -print-multi-directory
7114 @opindex print-multi-directory
7115 Print the directory name corresponding to the multilib selected by any
7116 other switches present in the command line. This directory is supposed
7117 to exist in @env{GCC_EXEC_PREFIX}.
7119 @item -print-multi-lib
7120 @opindex print-multi-lib
7121 Print the mapping from multilib directory names to compiler switches
7122 that enable them. The directory name is separated from the switches by
7123 @samp{;}, and each switch starts with an @samp{@@} instead of the
7124 @samp{-}, without spaces between multiple switches. This is supposed to
7125 ease shell processing.
7127 @item -print-multi-os-directory
7128 @opindex print-multi-os-directory
7129 Print the path to OS libraries for the selected
7130 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7131 present in the @file{lib} subdirectory and no multilibs are used, this is
7132 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7133 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7134 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7135 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7137 @item -print-multiarch
7138 @opindex print-multiarch
7139 Print the path to OS libraries for the selected multiarch,
7140 relative to some @file{lib} subdirectory.
7142 @item -print-prog-name=@var{program}
7143 @opindex print-prog-name
7144 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7146 @item -print-libgcc-file-name
7147 @opindex print-libgcc-file-name
7148 Same as @option{-print-file-name=libgcc.a}.
7150 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7151 but you do want to link with @file{libgcc.a}. You can do:
7154 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7157 @item -print-search-dirs
7158 @opindex print-search-dirs
7159 Print the name of the configured installation directory and a list of
7160 program and library directories @command{gcc} searches---and don't do anything else.
7162 This is useful when @command{gcc} prints the error message
7163 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7164 To resolve this you either need to put @file{cpp0} and the other compiler
7165 components where @command{gcc} expects to find them, or you can set the environment
7166 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7167 Don't forget the trailing @samp{/}.
7168 @xref{Environment Variables}.
7170 @item -print-sysroot
7171 @opindex print-sysroot
7172 Print the target sysroot directory that is used during
7173 compilation. This is the target sysroot specified either at configure
7174 time or using the @option{--sysroot} option, possibly with an extra
7175 suffix that depends on compilation options. If no target sysroot is
7176 specified, the option prints nothing.
7178 @item -print-sysroot-headers-suffix
7179 @opindex print-sysroot-headers-suffix
7180 Print the suffix added to the target sysroot when searching for
7181 headers, or give an error if the compiler is not configured with such
7182 a suffix---and don't do anything else.
7185 @opindex dumpmachine
7186 Print the compiler's target machine (for example,
7187 @samp{i686-pc-linux-gnu})---and don't do anything else.
7190 @opindex dumpversion
7191 Print the compiler version (for example, @code{3.0})---and don't do
7196 Print the compiler's built-in specs---and don't do anything else. (This
7197 is used when GCC itself is being built.) @xref{Spec Files}.
7199 @item -fno-eliminate-unused-debug-types
7200 @opindex feliminate-unused-debug-types
7201 @opindex fno-eliminate-unused-debug-types
7202 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7203 output for types that are nowhere used in the source file being compiled.
7204 Sometimes it is useful to have GCC emit debugging
7205 information for all types declared in a compilation
7206 unit, regardless of whether or not they are actually used
7207 in that compilation unit, for example
7208 if, in the debugger, you want to cast a value to a type that is
7209 not actually used in your program (but is declared). More often,
7210 however, this results in a significant amount of wasted space.
7213 @node Optimize Options
7214 @section Options That Control Optimization
7215 @cindex optimize options
7216 @cindex options, optimization
7218 These options control various sorts of optimizations.
7220 Without any optimization option, the compiler's goal is to reduce the
7221 cost of compilation and to make debugging produce the expected
7222 results. Statements are independent: if you stop the program with a
7223 breakpoint between statements, you can then assign a new value to any
7224 variable or change the program counter to any other statement in the
7225 function and get exactly the results you expect from the source
7228 Turning on optimization flags makes the compiler attempt to improve
7229 the performance and/or code size at the expense of compilation time
7230 and possibly the ability to debug the program.
7232 The compiler performs optimization based on the knowledge it has of the
7233 program. Compiling multiple files at once to a single output file mode allows
7234 the compiler to use information gained from all of the files when compiling
7237 Not all optimizations are controlled directly by a flag. Only
7238 optimizations that have a flag are listed in this section.
7240 Most optimizations are only enabled if an @option{-O} level is set on
7241 the command line. Otherwise they are disabled, even if individual
7242 optimization flags are specified.
7244 Depending on the target and how GCC was configured, a slightly different
7245 set of optimizations may be enabled at each @option{-O} level than
7246 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7247 to find out the exact set of optimizations that are enabled at each level.
7248 @xref{Overall Options}, for examples.
7255 Optimize. Optimizing compilation takes somewhat more time, and a lot
7256 more memory for a large function.
7258 With @option{-O}, the compiler tries to reduce code size and execution
7259 time, without performing any optimizations that take a great deal of
7262 @option{-O} turns on the following optimization flags:
7265 -fbranch-count-reg @gol
7266 -fcombine-stack-adjustments @gol
7268 -fcprop-registers @gol
7271 -fdelayed-branch @gol
7273 -fforward-propagate @gol
7274 -fguess-branch-probability @gol
7275 -fif-conversion2 @gol
7276 -fif-conversion @gol
7277 -finline-functions-called-once @gol
7278 -fipa-pure-const @gol
7280 -fipa-reference @gol
7281 -fmerge-constants @gol
7282 -fmove-loop-invariants @gol
7284 -fsplit-wide-types @gol
7289 -ftree-copy-prop @gol
7290 -ftree-copyrename @gol
7292 -ftree-dominator-opts @gol
7294 -ftree-forwprop @gol
7304 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7305 where doing so does not interfere with debugging.
7309 Optimize even more. GCC performs nearly all supported optimizations
7310 that do not involve a space-speed tradeoff.
7311 As compared to @option{-O}, this option increases both compilation time
7312 and the performance of the generated code.
7314 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7315 also turns on the following optimization flags:
7316 @gccoptlist{-fthread-jumps @gol
7317 -falign-functions -falign-jumps @gol
7318 -falign-loops -falign-labels @gol
7321 -fcse-follow-jumps -fcse-skip-blocks @gol
7322 -fdelete-null-pointer-checks @gol
7323 -fdevirtualize -fdevirtualize-speculatively @gol
7324 -fexpensive-optimizations @gol
7325 -fgcse -fgcse-lm @gol
7326 -fhoist-adjacent-loads @gol
7327 -finline-small-functions @gol
7328 -findirect-inlining @gol
7330 -fipa-cp-alignment @gol
7333 -fisolate-erroneous-paths-dereference @gol
7335 -foptimize-sibling-calls @gol
7336 -foptimize-strlen @gol
7337 -fpartial-inlining @gol
7339 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7340 -frerun-cse-after-loop @gol
7341 -fsched-interblock -fsched-spec @gol
7342 -fschedule-insns -fschedule-insns2 @gol
7343 -fstrict-aliasing -fstrict-overflow @gol
7344 -ftree-builtin-call-dce @gol
7345 -ftree-switch-conversion -ftree-tail-merge @gol
7350 Please note the warning under @option{-fgcse} about
7351 invoking @option{-O2} on programs that use computed gotos.
7355 Optimize yet more. @option{-O3} turns on all optimizations specified
7356 by @option{-O2} and also turns on the @option{-finline-functions},
7357 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7358 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7359 @option{-ftree-loop-distribute-patterns},
7360 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7361 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7365 Reduce compilation time and make debugging produce the expected
7366 results. This is the default.
7370 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7371 do not typically increase code size. It also performs further
7372 optimizations designed to reduce code size.
7374 @option{-Os} disables the following optimization flags:
7375 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7376 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7377 -fprefetch-loop-arrays}
7381 Disregard strict standards compliance. @option{-Ofast} enables all
7382 @option{-O3} optimizations. It also enables optimizations that are not
7383 valid for all standard-compliant programs.
7384 It turns on @option{-ffast-math} and the Fortran-specific
7385 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7389 Optimize debugging experience. @option{-Og} enables optimizations
7390 that do not interfere with debugging. It should be the optimization
7391 level of choice for the standard edit-compile-debug cycle, offering
7392 a reasonable level of optimization while maintaining fast compilation
7393 and a good debugging experience.
7395 If you use multiple @option{-O} options, with or without level numbers,
7396 the last such option is the one that is effective.
7399 Options of the form @option{-f@var{flag}} specify machine-independent
7400 flags. Most flags have both positive and negative forms; the negative
7401 form of @option{-ffoo} is @option{-fno-foo}. In the table
7402 below, only one of the forms is listed---the one you typically
7403 use. You can figure out the other form by either removing @samp{no-}
7406 The following options control specific optimizations. They are either
7407 activated by @option{-O} options or are related to ones that are. You
7408 can use the following flags in the rare cases when ``fine-tuning'' of
7409 optimizations to be performed is desired.
7412 @item -fno-defer-pop
7413 @opindex fno-defer-pop
7414 Always pop the arguments to each function call as soon as that function
7415 returns. For machines that must pop arguments after a function call,
7416 the compiler normally lets arguments accumulate on the stack for several
7417 function calls and pops them all at once.
7419 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7421 @item -fforward-propagate
7422 @opindex fforward-propagate
7423 Perform a forward propagation pass on RTL@. The pass tries to combine two
7424 instructions and checks if the result can be simplified. If loop unrolling
7425 is active, two passes are performed and the second is scheduled after
7428 This option is enabled by default at optimization levels @option{-O},
7429 @option{-O2}, @option{-O3}, @option{-Os}.
7431 @item -ffp-contract=@var{style}
7432 @opindex ffp-contract
7433 @option{-ffp-contract=off} disables floating-point expression contraction.
7434 @option{-ffp-contract=fast} enables floating-point expression contraction
7435 such as forming of fused multiply-add operations if the target has
7436 native support for them.
7437 @option{-ffp-contract=on} enables floating-point expression contraction
7438 if allowed by the language standard. This is currently not implemented
7439 and treated equal to @option{-ffp-contract=off}.
7441 The default is @option{-ffp-contract=fast}.
7443 @item -fomit-frame-pointer
7444 @opindex fomit-frame-pointer
7445 Don't keep the frame pointer in a register for functions that
7446 don't need one. This avoids the instructions to save, set up and
7447 restore frame pointers; it also makes an extra register available
7448 in many functions. @strong{It also makes debugging impossible on
7451 On some machines, such as the VAX, this flag has no effect, because
7452 the standard calling sequence automatically handles the frame pointer
7453 and nothing is saved by pretending it doesn't exist. The
7454 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7455 whether a target machine supports this flag. @xref{Registers,,Register
7456 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7458 Starting with GCC version 4.6, the default setting (when not optimizing for
7459 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets has been changed to
7460 @option{-fomit-frame-pointer}. The default can be reverted to
7461 @option{-fno-omit-frame-pointer} by configuring GCC with the
7462 @option{--enable-frame-pointer} configure option.
7464 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7466 @item -foptimize-sibling-calls
7467 @opindex foptimize-sibling-calls
7468 Optimize sibling and tail recursive calls.
7470 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7472 @item -foptimize-strlen
7473 @opindex foptimize-strlen
7474 Optimize various standard C string functions (e.g. @code{strlen},
7475 @code{strchr} or @code{strcpy}) and
7476 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7478 Enabled at levels @option{-O2}, @option{-O3}.
7482 Do not expand any functions inline apart from those marked with
7483 the @code{always_inline} attribute. This is the default when not
7486 Single functions can be exempted from inlining by marking them
7487 with the @code{noinline} attribute.
7489 @item -finline-small-functions
7490 @opindex finline-small-functions
7491 Integrate functions into their callers when their body is smaller than expected
7492 function call code (so overall size of program gets smaller). The compiler
7493 heuristically decides which functions are simple enough to be worth integrating
7494 in this way. This inlining applies to all functions, even those not declared
7497 Enabled at level @option{-O2}.
7499 @item -findirect-inlining
7500 @opindex findirect-inlining
7501 Inline also indirect calls that are discovered to be known at compile
7502 time thanks to previous inlining. This option has any effect only
7503 when inlining itself is turned on by the @option{-finline-functions}
7504 or @option{-finline-small-functions} options.
7506 Enabled at level @option{-O2}.
7508 @item -finline-functions
7509 @opindex finline-functions
7510 Consider all functions for inlining, even if they are not declared inline.
7511 The compiler heuristically decides which functions are worth integrating
7514 If all calls to a given function are integrated, and the function is
7515 declared @code{static}, then the function is normally not output as
7516 assembler code in its own right.
7518 Enabled at level @option{-O3}.
7520 @item -finline-functions-called-once
7521 @opindex finline-functions-called-once
7522 Consider all @code{static} functions called once for inlining into their
7523 caller even if they are not marked @code{inline}. If a call to a given
7524 function is integrated, then the function is not output as assembler code
7527 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7529 @item -fearly-inlining
7530 @opindex fearly-inlining
7531 Inline functions marked by @code{always_inline} and functions whose body seems
7532 smaller than the function call overhead early before doing
7533 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7534 makes profiling significantly cheaper and usually inlining faster on programs
7535 having large chains of nested wrapper functions.
7541 Perform interprocedural scalar replacement of aggregates, removal of
7542 unused parameters and replacement of parameters passed by reference
7543 by parameters passed by value.
7545 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7547 @item -finline-limit=@var{n}
7548 @opindex finline-limit
7549 By default, GCC limits the size of functions that can be inlined. This flag
7550 allows coarse control of this limit. @var{n} is the size of functions that
7551 can be inlined in number of pseudo instructions.
7553 Inlining is actually controlled by a number of parameters, which may be
7554 specified individually by using @option{--param @var{name}=@var{value}}.
7555 The @option{-finline-limit=@var{n}} option sets some of these parameters
7559 @item max-inline-insns-single
7560 is set to @var{n}/2.
7561 @item max-inline-insns-auto
7562 is set to @var{n}/2.
7565 See below for a documentation of the individual
7566 parameters controlling inlining and for the defaults of these parameters.
7568 @emph{Note:} there may be no value to @option{-finline-limit} that results
7569 in default behavior.
7571 @emph{Note:} pseudo instruction represents, in this particular context, an
7572 abstract measurement of function's size. In no way does it represent a count
7573 of assembly instructions and as such its exact meaning might change from one
7574 release to an another.
7576 @item -fno-keep-inline-dllexport
7577 @opindex fno-keep-inline-dllexport
7578 This is a more fine-grained version of @option{-fkeep-inline-functions},
7579 which applies only to functions that are declared using the @code{dllexport}
7580 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7583 @item -fkeep-inline-functions
7584 @opindex fkeep-inline-functions
7585 In C, emit @code{static} functions that are declared @code{inline}
7586 into the object file, even if the function has been inlined into all
7587 of its callers. This switch does not affect functions using the
7588 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7589 inline functions into the object file.
7591 @item -fkeep-static-consts
7592 @opindex fkeep-static-consts
7593 Emit variables declared @code{static const} when optimization isn't turned
7594 on, even if the variables aren't referenced.
7596 GCC enables this option by default. If you want to force the compiler to
7597 check if a variable is referenced, regardless of whether or not
7598 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7600 @item -fmerge-constants
7601 @opindex fmerge-constants
7602 Attempt to merge identical constants (string constants and floating-point
7603 constants) across compilation units.
7605 This option is the default for optimized compilation if the assembler and
7606 linker support it. Use @option{-fno-merge-constants} to inhibit this
7609 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7611 @item -fmerge-all-constants
7612 @opindex fmerge-all-constants
7613 Attempt to merge identical constants and identical variables.
7615 This option implies @option{-fmerge-constants}. In addition to
7616 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7617 arrays or initialized constant variables with integral or floating-point
7618 types. Languages like C or C++ require each variable, including multiple
7619 instances of the same variable in recursive calls, to have distinct locations,
7620 so using this option results in non-conforming
7623 @item -fmodulo-sched
7624 @opindex fmodulo-sched
7625 Perform swing modulo scheduling immediately before the first scheduling
7626 pass. This pass looks at innermost loops and reorders their
7627 instructions by overlapping different iterations.
7629 @item -fmodulo-sched-allow-regmoves
7630 @opindex fmodulo-sched-allow-regmoves
7631 Perform more aggressive SMS-based modulo scheduling with register moves
7632 allowed. By setting this flag certain anti-dependences edges are
7633 deleted, which triggers the generation of reg-moves based on the
7634 life-range analysis. This option is effective only with
7635 @option{-fmodulo-sched} enabled.
7637 @item -fno-branch-count-reg
7638 @opindex fno-branch-count-reg
7639 Do not use ``decrement and branch'' instructions on a count register,
7640 but instead generate a sequence of instructions that decrement a
7641 register, compare it against zero, then branch based upon the result.
7642 This option is only meaningful on architectures that support such
7643 instructions, which include x86, PowerPC, IA-64 and S/390.
7645 Enabled by default at @option{-O1} and higher.
7647 The default is @option{-fbranch-count-reg}.
7649 @item -fno-function-cse
7650 @opindex fno-function-cse
7651 Do not put function addresses in registers; make each instruction that
7652 calls a constant function contain the function's address explicitly.
7654 This option results in less efficient code, but some strange hacks
7655 that alter the assembler output may be confused by the optimizations
7656 performed when this option is not used.
7658 The default is @option{-ffunction-cse}
7660 @item -fno-zero-initialized-in-bss
7661 @opindex fno-zero-initialized-in-bss
7662 If the target supports a BSS section, GCC by default puts variables that
7663 are initialized to zero into BSS@. This can save space in the resulting
7666 This option turns off this behavior because some programs explicitly
7667 rely on variables going to the data section---e.g., so that the
7668 resulting executable can find the beginning of that section and/or make
7669 assumptions based on that.
7671 The default is @option{-fzero-initialized-in-bss}.
7673 @item -fthread-jumps
7674 @opindex fthread-jumps
7675 Perform optimizations that check to see if a jump branches to a
7676 location where another comparison subsumed by the first is found. If
7677 so, the first branch is redirected to either the destination of the
7678 second branch or a point immediately following it, depending on whether
7679 the condition is known to be true or false.
7681 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7683 @item -fsplit-wide-types
7684 @opindex fsplit-wide-types
7685 When using a type that occupies multiple registers, such as @code{long
7686 long} on a 32-bit system, split the registers apart and allocate them
7687 independently. This normally generates better code for those types,
7688 but may make debugging more difficult.
7690 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7693 @item -fcse-follow-jumps
7694 @opindex fcse-follow-jumps
7695 In common subexpression elimination (CSE), scan through jump instructions
7696 when the target of the jump is not reached by any other path. For
7697 example, when CSE encounters an @code{if} statement with an
7698 @code{else} clause, CSE follows the jump when the condition
7701 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7703 @item -fcse-skip-blocks
7704 @opindex fcse-skip-blocks
7705 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7706 follow jumps that conditionally skip over blocks. When CSE
7707 encounters a simple @code{if} statement with no else clause,
7708 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7709 body of the @code{if}.
7711 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7713 @item -frerun-cse-after-loop
7714 @opindex frerun-cse-after-loop
7715 Re-run common subexpression elimination after loop optimizations are
7718 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7722 Perform a global common subexpression elimination pass.
7723 This pass also performs global constant and copy propagation.
7725 @emph{Note:} When compiling a program using computed gotos, a GCC
7726 extension, you may get better run-time performance if you disable
7727 the global common subexpression elimination pass by adding
7728 @option{-fno-gcse} to the command line.
7730 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7734 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7735 attempts to move loads that are only killed by stores into themselves. This
7736 allows a loop containing a load/store sequence to be changed to a load outside
7737 the loop, and a copy/store within the loop.
7739 Enabled by default when @option{-fgcse} is enabled.
7743 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7744 global common subexpression elimination. This pass attempts to move
7745 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7746 loops containing a load/store sequence can be changed to a load before
7747 the loop and a store after the loop.
7749 Not enabled at any optimization level.
7753 When @option{-fgcse-las} is enabled, the global common subexpression
7754 elimination pass eliminates redundant loads that come after stores to the
7755 same memory location (both partial and full redundancies).
7757 Not enabled at any optimization level.
7759 @item -fgcse-after-reload
7760 @opindex fgcse-after-reload
7761 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7762 pass is performed after reload. The purpose of this pass is to clean up
7765 @item -faggressive-loop-optimizations
7766 @opindex faggressive-loop-optimizations
7767 This option tells the loop optimizer to use language constraints to
7768 derive bounds for the number of iterations of a loop. This assumes that
7769 loop code does not invoke undefined behavior by for example causing signed
7770 integer overflows or out-of-bound array accesses. The bounds for the
7771 number of iterations of a loop are used to guide loop unrolling and peeling
7772 and loop exit test optimizations.
7773 This option is enabled by default.
7775 @item -funsafe-loop-optimizations
7776 @opindex funsafe-loop-optimizations
7777 This option tells the loop optimizer to assume that loop indices do not
7778 overflow, and that loops with nontrivial exit condition are not
7779 infinite. This enables a wider range of loop optimizations even if
7780 the loop optimizer itself cannot prove that these assumptions are valid.
7781 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7782 if it finds this kind of loop.
7784 @item -fcrossjumping
7785 @opindex fcrossjumping
7786 Perform cross-jumping transformation.
7787 This transformation unifies equivalent code and saves code size. The
7788 resulting code may or may not perform better than without cross-jumping.
7790 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7792 @item -fauto-inc-dec
7793 @opindex fauto-inc-dec
7794 Combine increments or decrements of addresses with memory accesses.
7795 This pass is always skipped on architectures that do not have
7796 instructions to support this. Enabled by default at @option{-O} and
7797 higher on architectures that support this.
7801 Perform dead code elimination (DCE) on RTL@.
7802 Enabled by default at @option{-O} and higher.
7806 Perform dead store elimination (DSE) on RTL@.
7807 Enabled by default at @option{-O} and higher.
7809 @item -fif-conversion
7810 @opindex fif-conversion
7811 Attempt to transform conditional jumps into branch-less equivalents. This
7812 includes use of conditional moves, min, max, set flags and abs instructions, and
7813 some tricks doable by standard arithmetics. The use of conditional execution
7814 on chips where it is available is controlled by @option{-fif-conversion2}.
7816 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7818 @item -fif-conversion2
7819 @opindex fif-conversion2
7820 Use conditional execution (where available) to transform conditional jumps into
7821 branch-less equivalents.
7823 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7825 @item -fdeclone-ctor-dtor
7826 @opindex fdeclone-ctor-dtor
7827 The C++ ABI requires multiple entry points for constructors and
7828 destructors: one for a base subobject, one for a complete object, and
7829 one for a virtual destructor that calls operator delete afterwards.
7830 For a hierarchy with virtual bases, the base and complete variants are
7831 clones, which means two copies of the function. With this option, the
7832 base and complete variants are changed to be thunks that call a common
7835 Enabled by @option{-Os}.
7837 @item -fdelete-null-pointer-checks
7838 @opindex fdelete-null-pointer-checks
7839 Assume that programs cannot safely dereference null pointers, and that
7840 no code or data element resides there. This enables simple constant
7841 folding optimizations at all optimization levels. In addition, other
7842 optimization passes in GCC use this flag to control global dataflow
7843 analyses that eliminate useless checks for null pointers; these assume
7844 that if a pointer is checked after it has already been dereferenced,
7847 Note however that in some environments this assumption is not true.
7848 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
7849 for programs that depend on that behavior.
7851 Some targets, especially embedded ones, disable this option at all levels.
7852 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
7853 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
7854 are enabled independently at different optimization levels.
7856 @item -fdevirtualize
7857 @opindex fdevirtualize
7858 Attempt to convert calls to virtual functions to direct calls. This
7859 is done both within a procedure and interprocedurally as part of
7860 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
7861 propagation (@option{-fipa-cp}).
7862 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7864 @item -fdevirtualize-speculatively
7865 @opindex fdevirtualize-speculatively
7866 Attempt to convert calls to virtual functions to speculative direct calls.
7867 Based on the analysis of the type inheritance graph, determine for a given call
7868 the set of likely targets. If the set is small, preferably of size 1, change
7869 the call into a conditional deciding between direct and indirect calls. The
7870 speculative calls enable more optimizations, such as inlining. When they seem
7871 useless after further optimization, they are converted back into original form.
7873 @item -fdevirtualize-at-ltrans
7874 @opindex fdevirtualize-at-ltrans
7875 Stream extra information needed for aggressive devirtualization when running
7876 the link-time optimizer in local transformation mode.
7877 This option enables more devirtualization but
7878 significantly increases the size of streamed data. For this reason it is
7879 disabled by default.
7881 @item -fexpensive-optimizations
7882 @opindex fexpensive-optimizations
7883 Perform a number of minor optimizations that are relatively expensive.
7885 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7889 Attempt to remove redundant extension instructions. This is especially
7890 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
7891 registers after writing to their lower 32-bit half.
7893 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
7894 @option{-O3}, @option{-Os}.
7896 @item -fno-lifetime-dse
7897 @opindex fno-lifetime-dse
7898 In C++ the value of an object is only affected by changes within its
7899 lifetime: when the constructor begins, the object has an indeterminate
7900 value, and any changes during the lifetime of the object are dead when
7901 the object is destroyed. Normally dead store elimination will take
7902 advantage of this; if your code relies on the value of the object
7903 storage persisting beyond the lifetime of the object, you can use this
7904 flag to disable this optimization.
7906 @item -flive-range-shrinkage
7907 @opindex flive-range-shrinkage
7908 Attempt to decrease register pressure through register live range
7909 shrinkage. This is helpful for fast processors with small or moderate
7912 @item -fira-algorithm=@var{algorithm}
7913 @opindex fira-algorithm
7914 Use the specified coloring algorithm for the integrated register
7915 allocator. The @var{algorithm} argument can be @samp{priority}, which
7916 specifies Chow's priority coloring, or @samp{CB}, which specifies
7917 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
7918 for all architectures, but for those targets that do support it, it is
7919 the default because it generates better code.
7921 @item -fira-region=@var{region}
7922 @opindex fira-region
7923 Use specified regions for the integrated register allocator. The
7924 @var{region} argument should be one of the following:
7929 Use all loops as register allocation regions.
7930 This can give the best results for machines with a small and/or
7931 irregular register set.
7934 Use all loops except for loops with small register pressure
7935 as the regions. This value usually gives
7936 the best results in most cases and for most architectures,
7937 and is enabled by default when compiling with optimization for speed
7938 (@option{-O}, @option{-O2}, @dots{}).
7941 Use all functions as a single region.
7942 This typically results in the smallest code size, and is enabled by default for
7943 @option{-Os} or @option{-O0}.
7947 @item -fira-hoist-pressure
7948 @opindex fira-hoist-pressure
7949 Use IRA to evaluate register pressure in the code hoisting pass for
7950 decisions to hoist expressions. This option usually results in smaller
7951 code, but it can slow the compiler down.
7953 This option is enabled at level @option{-Os} for all targets.
7955 @item -fira-loop-pressure
7956 @opindex fira-loop-pressure
7957 Use IRA to evaluate register pressure in loops for decisions to move
7958 loop invariants. This option usually results in generation
7959 of faster and smaller code on machines with large register files (>= 32
7960 registers), but it can slow the compiler down.
7962 This option is enabled at level @option{-O3} for some targets.
7964 @item -fno-ira-share-save-slots
7965 @opindex fno-ira-share-save-slots
7966 Disable sharing of stack slots used for saving call-used hard
7967 registers living through a call. Each hard register gets a
7968 separate stack slot, and as a result function stack frames are
7971 @item -fno-ira-share-spill-slots
7972 @opindex fno-ira-share-spill-slots
7973 Disable sharing of stack slots allocated for pseudo-registers. Each
7974 pseudo-register that does not get a hard register gets a separate
7975 stack slot, and as a result function stack frames are larger.
7977 @item -fira-verbose=@var{n}
7978 @opindex fira-verbose
7979 Control the verbosity of the dump file for the integrated register allocator.
7980 The default value is 5. If the value @var{n} is greater or equal to 10,
7981 the dump output is sent to stderr using the same format as @var{n} minus 10.
7985 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7986 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7987 values if it is profitable.
7989 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7991 @item -fdelayed-branch
7992 @opindex fdelayed-branch
7993 If supported for the target machine, attempt to reorder instructions
7994 to exploit instruction slots available after delayed branch
7997 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7999 @item -fschedule-insns
8000 @opindex fschedule-insns
8001 If supported for the target machine, attempt to reorder instructions to
8002 eliminate execution stalls due to required data being unavailable. This
8003 helps machines that have slow floating point or memory load instructions
8004 by allowing other instructions to be issued until the result of the load
8005 or floating-point instruction is required.
8007 Enabled at levels @option{-O2}, @option{-O3}.
8009 @item -fschedule-insns2
8010 @opindex fschedule-insns2
8011 Similar to @option{-fschedule-insns}, but requests an additional pass of
8012 instruction scheduling after register allocation has been done. This is
8013 especially useful on machines with a relatively small number of
8014 registers and where memory load instructions take more than one cycle.
8016 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8018 @item -fno-sched-interblock
8019 @opindex fno-sched-interblock
8020 Don't schedule instructions across basic blocks. This is normally
8021 enabled by default when scheduling before register allocation, i.e.@:
8022 with @option{-fschedule-insns} or at @option{-O2} or higher.
8024 @item -fno-sched-spec
8025 @opindex fno-sched-spec
8026 Don't allow speculative motion of non-load instructions. This is normally
8027 enabled by default when scheduling before register allocation, i.e.@:
8028 with @option{-fschedule-insns} or at @option{-O2} or higher.
8030 @item -fsched-pressure
8031 @opindex fsched-pressure
8032 Enable register pressure sensitive insn scheduling before register
8033 allocation. This only makes sense when scheduling before register
8034 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8035 @option{-O2} or higher. Usage of this option can improve the
8036 generated code and decrease its size by preventing register pressure
8037 increase above the number of available hard registers and subsequent
8038 spills in register allocation.
8040 @item -fsched-spec-load
8041 @opindex fsched-spec-load
8042 Allow speculative motion of some load instructions. This only makes
8043 sense when scheduling before register allocation, i.e.@: with
8044 @option{-fschedule-insns} or at @option{-O2} or higher.
8046 @item -fsched-spec-load-dangerous
8047 @opindex fsched-spec-load-dangerous
8048 Allow speculative motion of more load instructions. This only makes
8049 sense when scheduling before register allocation, i.e.@: with
8050 @option{-fschedule-insns} or at @option{-O2} or higher.
8052 @item -fsched-stalled-insns
8053 @itemx -fsched-stalled-insns=@var{n}
8054 @opindex fsched-stalled-insns
8055 Define how many insns (if any) can be moved prematurely from the queue
8056 of stalled insns into the ready list during the second scheduling pass.
8057 @option{-fno-sched-stalled-insns} means that no insns are moved
8058 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8059 on how many queued insns can be moved prematurely.
8060 @option{-fsched-stalled-insns} without a value is equivalent to
8061 @option{-fsched-stalled-insns=1}.
8063 @item -fsched-stalled-insns-dep
8064 @itemx -fsched-stalled-insns-dep=@var{n}
8065 @opindex fsched-stalled-insns-dep
8066 Define how many insn groups (cycles) are examined for a dependency
8067 on a stalled insn that is a candidate for premature removal from the queue
8068 of stalled insns. This has an effect only during the second scheduling pass,
8069 and only if @option{-fsched-stalled-insns} is used.
8070 @option{-fno-sched-stalled-insns-dep} is equivalent to
8071 @option{-fsched-stalled-insns-dep=0}.
8072 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8073 @option{-fsched-stalled-insns-dep=1}.
8075 @item -fsched2-use-superblocks
8076 @opindex fsched2-use-superblocks
8077 When scheduling after register allocation, use superblock scheduling.
8078 This allows motion across basic block boundaries,
8079 resulting in faster schedules. This option is experimental, as not all machine
8080 descriptions used by GCC model the CPU closely enough to avoid unreliable
8081 results from the algorithm.
8083 This only makes sense when scheduling after register allocation, i.e.@: with
8084 @option{-fschedule-insns2} or at @option{-O2} or higher.
8086 @item -fsched-group-heuristic
8087 @opindex fsched-group-heuristic
8088 Enable the group heuristic in the scheduler. This heuristic favors
8089 the instruction that belongs to a schedule group. This is enabled
8090 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8091 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8093 @item -fsched-critical-path-heuristic
8094 @opindex fsched-critical-path-heuristic
8095 Enable the critical-path heuristic in the scheduler. This heuristic favors
8096 instructions on the critical path. This is enabled by default when
8097 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8098 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8100 @item -fsched-spec-insn-heuristic
8101 @opindex fsched-spec-insn-heuristic
8102 Enable the speculative instruction heuristic in the scheduler. This
8103 heuristic favors speculative instructions with greater dependency weakness.
8104 This is enabled by default when scheduling is enabled, i.e.@:
8105 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8106 or at @option{-O2} or higher.
8108 @item -fsched-rank-heuristic
8109 @opindex fsched-rank-heuristic
8110 Enable the rank heuristic in the scheduler. This heuristic favors
8111 the instruction belonging to a basic block with greater size or frequency.
8112 This is enabled by default when scheduling is enabled, i.e.@:
8113 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8114 at @option{-O2} or higher.
8116 @item -fsched-last-insn-heuristic
8117 @opindex fsched-last-insn-heuristic
8118 Enable the last-instruction heuristic in the scheduler. This heuristic
8119 favors the instruction that is less dependent on the last instruction
8120 scheduled. This is enabled by default when scheduling is enabled,
8121 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8122 at @option{-O2} or higher.
8124 @item -fsched-dep-count-heuristic
8125 @opindex fsched-dep-count-heuristic
8126 Enable the dependent-count heuristic in the scheduler. This heuristic
8127 favors the instruction that has more instructions depending on it.
8128 This is enabled by default when scheduling is enabled, i.e.@:
8129 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8130 at @option{-O2} or higher.
8132 @item -freschedule-modulo-scheduled-loops
8133 @opindex freschedule-modulo-scheduled-loops
8134 Modulo scheduling is performed before traditional scheduling. If a loop
8135 is modulo scheduled, later scheduling passes may change its schedule.
8136 Use this option to control that behavior.
8138 @item -fselective-scheduling
8139 @opindex fselective-scheduling
8140 Schedule instructions using selective scheduling algorithm. Selective
8141 scheduling runs instead of the first scheduler pass.
8143 @item -fselective-scheduling2
8144 @opindex fselective-scheduling2
8145 Schedule instructions using selective scheduling algorithm. Selective
8146 scheduling runs instead of the second scheduler pass.
8148 @item -fsel-sched-pipelining
8149 @opindex fsel-sched-pipelining
8150 Enable software pipelining of innermost loops during selective scheduling.
8151 This option has no effect unless one of @option{-fselective-scheduling} or
8152 @option{-fselective-scheduling2} is turned on.
8154 @item -fsel-sched-pipelining-outer-loops
8155 @opindex fsel-sched-pipelining-outer-loops
8156 When pipelining loops during selective scheduling, also pipeline outer loops.
8157 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8159 @item -fsemantic-interposition
8160 @opindex fsemantic-interposition
8161 Some object formats, like ELF, allow interposing of symbols by the
8163 This means that for symbols exported from the DSO, the compiler cannot perform
8164 interprocedural propagation, inlining and other optimizations in anticipation
8165 that the function or variable in question may change. While this feature is
8166 useful, for example, to rewrite memory allocation functions by a debugging
8167 implementation, it is expensive in the terms of code quality.
8168 With @option{-fno-semantic-interposition} the compiler assumes that
8169 if interposition happens for functions the overwriting function will have
8170 precisely the same semantics (and side effects).
8171 Similarly if interposition happens
8172 for variables, the constructor of the variable will be the same. The flag
8173 has no effect for functions explicitly declared inline
8174 (where it is never allowed for interposition to change semantics)
8175 and for symbols explicitly declared weak.
8178 @opindex fshrink-wrap
8179 Emit function prologues only before parts of the function that need it,
8180 rather than at the top of the function. This flag is enabled by default at
8181 @option{-O} and higher.
8183 @item -fcaller-saves
8184 @opindex fcaller-saves
8185 Enable allocation of values to registers that are clobbered by
8186 function calls, by emitting extra instructions to save and restore the
8187 registers around such calls. Such allocation is done only when it
8188 seems to result in better code.
8190 This option is always enabled by default on certain machines, usually
8191 those which have no call-preserved registers to use instead.
8193 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8195 @item -fcombine-stack-adjustments
8196 @opindex fcombine-stack-adjustments
8197 Tracks stack adjustments (pushes and pops) and stack memory references
8198 and then tries to find ways to combine them.
8200 Enabled by default at @option{-O1} and higher.
8204 Use caller save registers for allocation if those registers are not used by
8205 any called function. In that case it is not necessary to save and restore
8206 them around calls. This is only possible if called functions are part of
8207 same compilation unit as current function and they are compiled before it.
8209 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8211 @item -fconserve-stack
8212 @opindex fconserve-stack
8213 Attempt to minimize stack usage. The compiler attempts to use less
8214 stack space, even if that makes the program slower. This option
8215 implies setting the @option{large-stack-frame} parameter to 100
8216 and the @option{large-stack-frame-growth} parameter to 400.
8218 @item -ftree-reassoc
8219 @opindex ftree-reassoc
8220 Perform reassociation on trees. This flag is enabled by default
8221 at @option{-O} and higher.
8225 Perform partial redundancy elimination (PRE) on trees. This flag is
8226 enabled by default at @option{-O2} and @option{-O3}.
8228 @item -ftree-partial-pre
8229 @opindex ftree-partial-pre
8230 Make partial redundancy elimination (PRE) more aggressive. This flag is
8231 enabled by default at @option{-O3}.
8233 @item -ftree-forwprop
8234 @opindex ftree-forwprop
8235 Perform forward propagation on trees. This flag is enabled by default
8236 at @option{-O} and higher.
8240 Perform full redundancy elimination (FRE) on trees. The difference
8241 between FRE and PRE is that FRE only considers expressions
8242 that are computed on all paths leading to the redundant computation.
8243 This analysis is faster than PRE, though it exposes fewer redundancies.
8244 This flag is enabled by default at @option{-O} and higher.
8246 @item -ftree-phiprop
8247 @opindex ftree-phiprop
8248 Perform hoisting of loads from conditional pointers on trees. This
8249 pass is enabled by default at @option{-O} and higher.
8251 @item -fhoist-adjacent-loads
8252 @opindex fhoist-adjacent-loads
8253 Speculatively hoist loads from both branches of an if-then-else if the
8254 loads are from adjacent locations in the same structure and the target
8255 architecture has a conditional move instruction. This flag is enabled
8256 by default at @option{-O2} and higher.
8258 @item -ftree-copy-prop
8259 @opindex ftree-copy-prop
8260 Perform copy propagation on trees. This pass eliminates unnecessary
8261 copy operations. This flag is enabled by default at @option{-O} and
8264 @item -fipa-pure-const
8265 @opindex fipa-pure-const
8266 Discover which functions are pure or constant.
8267 Enabled by default at @option{-O} and higher.
8269 @item -fipa-reference
8270 @opindex fipa-reference
8271 Discover which static variables do not escape the
8273 Enabled by default at @option{-O} and higher.
8277 Perform interprocedural pointer analysis and interprocedural modification
8278 and reference analysis. This option can cause excessive memory and
8279 compile-time usage on large compilation units. It is not enabled by
8280 default at any optimization level.
8283 @opindex fipa-profile
8284 Perform interprocedural profile propagation. The functions called only from
8285 cold functions are marked as cold. Also functions executed once (such as
8286 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8287 functions and loop less parts of functions executed once are then optimized for
8289 Enabled by default at @option{-O} and higher.
8293 Perform interprocedural constant propagation.
8294 This optimization analyzes the program to determine when values passed
8295 to functions are constants and then optimizes accordingly.
8296 This optimization can substantially increase performance
8297 if the application has constants passed to functions.
8298 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8300 @item -fipa-cp-clone
8301 @opindex fipa-cp-clone
8302 Perform function cloning to make interprocedural constant propagation stronger.
8303 When enabled, interprocedural constant propagation performs function cloning
8304 when externally visible function can be called with constant arguments.
8305 Because this optimization can create multiple copies of functions,
8306 it may significantly increase code size
8307 (see @option{--param ipcp-unit-growth=@var{value}}).
8308 This flag is enabled by default at @option{-O3}.
8310 @item -fipa-cp-alignment
8311 @opindex -fipa-cp-alignment
8312 When enabled, this optimization propagates alignment of function
8313 parameters to support better vectorization and string operations.
8315 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8316 requires that @option{-fipa-cp} is enabled.
8320 Perform Identical Code Folding for functions and read-only variables.
8321 The optimization reduces code size and may disturb unwind stacks by replacing
8322 a function by equivalent one with a different name. The optimization works
8323 more effectively with link time optimization enabled.
8325 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8326 works on different levels and thus the optimizations are not same - there are
8327 equivalences that are found only by GCC and equivalences found only by Gold.
8329 This flag is enabled by default at @option{-O2} and @option{-Os}.
8331 @item -fisolate-erroneous-paths-dereference
8332 @opindex fisolate-erroneous-paths-dereference
8333 Detect paths that trigger erroneous or undefined behavior due to
8334 dereferencing a null pointer. Isolate those paths from the main control
8335 flow and turn the statement with erroneous or undefined behavior into a trap.
8336 This flag is enabled by default at @option{-O2} and higher.
8338 @item -fisolate-erroneous-paths-attribute
8339 @opindex fisolate-erroneous-paths-attribute
8340 Detect paths that trigger erroneous or undefined behavior due a null value
8341 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8342 attribute. Isolate those paths from the main control flow and turn the
8343 statement with erroneous or undefined behavior into a trap. This is not
8344 currently enabled, but may be enabled by @option{-O2} in the future.
8348 Perform forward store motion on trees. This flag is
8349 enabled by default at @option{-O} and higher.
8351 @item -ftree-bit-ccp
8352 @opindex ftree-bit-ccp
8353 Perform sparse conditional bit constant propagation on trees and propagate
8354 pointer alignment information.
8355 This pass only operates on local scalar variables and is enabled by default
8356 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8360 Perform sparse conditional constant propagation (CCP) on trees. This
8361 pass only operates on local scalar variables and is enabled by default
8362 at @option{-O} and higher.
8365 @opindex fssa-phiopt
8366 Perform pattern matching on SSA PHI nodes to optimize conditional
8367 code. This pass is enabled by default at @option{-O} and higher.
8369 @item -ftree-switch-conversion
8370 @opindex ftree-switch-conversion
8371 Perform conversion of simple initializations in a switch to
8372 initializations from a scalar array. This flag is enabled by default
8373 at @option{-O2} and higher.
8375 @item -ftree-tail-merge
8376 @opindex ftree-tail-merge
8377 Look for identical code sequences. When found, replace one with a jump to the
8378 other. This optimization is known as tail merging or cross jumping. This flag
8379 is enabled by default at @option{-O2} and higher. The compilation time
8381 be limited using @option{max-tail-merge-comparisons} parameter and
8382 @option{max-tail-merge-iterations} parameter.
8386 Perform dead code elimination (DCE) on trees. This flag is enabled by
8387 default at @option{-O} and higher.
8389 @item -ftree-builtin-call-dce
8390 @opindex ftree-builtin-call-dce
8391 Perform conditional dead code elimination (DCE) for calls to built-in functions
8392 that may set @code{errno} but are otherwise side-effect free. This flag is
8393 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8396 @item -ftree-dominator-opts
8397 @opindex ftree-dominator-opts
8398 Perform a variety of simple scalar cleanups (constant/copy
8399 propagation, redundancy elimination, range propagation and expression
8400 simplification) based on a dominator tree traversal. This also
8401 performs jump threading (to reduce jumps to jumps). This flag is
8402 enabled by default at @option{-O} and higher.
8406 Perform dead store elimination (DSE) on trees. A dead store is a store into
8407 a memory location that is later overwritten by another store without
8408 any intervening loads. In this case the earlier store can be deleted. This
8409 flag is enabled by default at @option{-O} and higher.
8413 Perform loop header copying on trees. This is beneficial since it increases
8414 effectiveness of code motion optimizations. It also saves one jump. This flag
8415 is enabled by default at @option{-O} and higher. It is not enabled
8416 for @option{-Os}, since it usually increases code size.
8418 @item -ftree-loop-optimize
8419 @opindex ftree-loop-optimize
8420 Perform loop optimizations on trees. This flag is enabled by default
8421 at @option{-O} and higher.
8423 @item -ftree-loop-linear
8424 @opindex ftree-loop-linear
8425 Perform loop interchange transformations on tree. Same as
8426 @option{-floop-interchange}. To use this code transformation, GCC has
8427 to be configured with @option{--with-isl} to enable the Graphite loop
8428 transformation infrastructure.
8430 @item -floop-interchange
8431 @opindex floop-interchange
8432 Perform loop interchange transformations on loops. Interchanging two
8433 nested loops switches the inner and outer loops. For example, given a
8438 A(J, I) = A(J, I) * C
8443 loop interchange transforms the loop as if it were written:
8447 A(J, I) = A(J, I) * C
8451 which can be beneficial when @code{N} is larger than the caches,
8452 because in Fortran, the elements of an array are stored in memory
8453 contiguously by column, and the original loop iterates over rows,
8454 potentially creating at each access a cache miss. This optimization
8455 applies to all the languages supported by GCC and is not limited to
8456 Fortran. To use this code transformation, GCC has to be configured
8457 with @option{--with-isl} to enable the Graphite loop transformation
8460 @item -floop-strip-mine
8461 @opindex floop-strip-mine
8462 Perform loop strip mining transformations on loops. Strip mining
8463 splits a loop into two nested loops. The outer loop has strides
8464 equal to the strip size and the inner loop has strides of the
8465 original loop within a strip. The strip length can be changed
8466 using the @option{loop-block-tile-size} parameter. For example,
8474 loop strip mining transforms the loop as if it were written:
8477 DO I = II, min (II + 50, N)
8482 This optimization applies to all the languages supported by GCC and is
8483 not limited to Fortran. To use this code transformation, GCC has to
8484 be configured with @option{--with-isl} to enable the Graphite loop
8485 transformation infrastructure.
8488 @opindex floop-block
8489 Perform loop blocking transformations on loops. Blocking strip mines
8490 each loop in the loop nest such that the memory accesses of the
8491 element loops fit inside caches. The strip length can be changed
8492 using the @option{loop-block-tile-size} parameter. For example, given
8497 A(J, I) = B(I) + C(J)
8502 loop blocking transforms the loop as if it were written:
8506 DO I = II, min (II + 50, N)
8507 DO J = JJ, min (JJ + 50, M)
8508 A(J, I) = B(I) + C(J)
8514 which can be beneficial when @code{M} is larger than the caches,
8515 because the innermost loop iterates over a smaller amount of data
8516 which can be kept in the caches. This optimization applies to all the
8517 languages supported by GCC and is not limited to Fortran. To use this
8518 code transformation, GCC has to be configured with @option{--with-isl}
8519 to enable the Graphite loop transformation infrastructure.
8521 @item -fgraphite-identity
8522 @opindex fgraphite-identity
8523 Enable the identity transformation for graphite. For every SCoP we generate
8524 the polyhedral representation and transform it back to gimple. Using
8525 @option{-fgraphite-identity} we can check the costs or benefits of the
8526 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8527 are also performed by the code generator ISL, like index splitting and
8528 dead code elimination in loops.
8530 @item -floop-nest-optimize
8531 @opindex floop-nest-optimize
8532 Enable the ISL based loop nest optimizer. This is a generic loop nest
8533 optimizer based on the Pluto optimization algorithms. It calculates a loop
8534 structure optimized for data-locality and parallelism. This option
8537 @item -floop-unroll-and-jam
8538 @opindex floop-unroll-and-jam
8539 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8540 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8541 The unrolled dimension (counting from the most inner one) can be changed
8542 using the @option{loop-unroll-jam-depth} parameter. .
8544 @item -floop-parallelize-all
8545 @opindex floop-parallelize-all
8546 Use the Graphite data dependence analysis to identify loops that can
8547 be parallelized. Parallelize all the loops that can be analyzed to
8548 not contain loop carried dependences without checking that it is
8549 profitable to parallelize the loops.
8551 @item -fcheck-data-deps
8552 @opindex fcheck-data-deps
8553 Compare the results of several data dependence analyzers. This option
8554 is used for debugging the data dependence analyzers.
8556 @item -ftree-loop-if-convert
8557 @opindex ftree-loop-if-convert
8558 Attempt to transform conditional jumps in the innermost loops to
8559 branch-less equivalents. The intent is to remove control-flow from
8560 the innermost loops in order to improve the ability of the
8561 vectorization pass to handle these loops. This is enabled by default
8562 if vectorization is enabled.
8564 @item -ftree-loop-if-convert-stores
8565 @opindex ftree-loop-if-convert-stores
8566 Attempt to also if-convert conditional jumps containing memory writes.
8567 This transformation can be unsafe for multi-threaded programs as it
8568 transforms conditional memory writes into unconditional memory writes.
8571 for (i = 0; i < N; i++)
8577 for (i = 0; i < N; i++)
8578 A[i] = cond ? expr : A[i];
8580 potentially producing data races.
8582 @item -ftree-loop-distribution
8583 @opindex ftree-loop-distribution
8584 Perform loop distribution. This flag can improve cache performance on
8585 big loop bodies and allow further loop optimizations, like
8586 parallelization or vectorization, to take place. For example, the loop
8603 @item -ftree-loop-distribute-patterns
8604 @opindex ftree-loop-distribute-patterns
8605 Perform loop distribution of patterns that can be code generated with
8606 calls to a library. This flag is enabled by default at @option{-O3}.
8608 This pass distributes the initialization loops and generates a call to
8609 memset zero. For example, the loop
8625 and the initialization loop is transformed into a call to memset zero.
8627 @item -ftree-loop-im
8628 @opindex ftree-loop-im
8629 Perform loop invariant motion on trees. This pass moves only invariants that
8630 are hard to handle at RTL level (function calls, operations that expand to
8631 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8632 operands of conditions that are invariant out of the loop, so that we can use
8633 just trivial invariantness analysis in loop unswitching. The pass also includes
8636 @item -ftree-loop-ivcanon
8637 @opindex ftree-loop-ivcanon
8638 Create a canonical counter for number of iterations in loops for which
8639 determining number of iterations requires complicated analysis. Later
8640 optimizations then may determine the number easily. Useful especially
8641 in connection with unrolling.
8645 Perform induction variable optimizations (strength reduction, induction
8646 variable merging and induction variable elimination) on trees.
8648 @item -ftree-parallelize-loops=n
8649 @opindex ftree-parallelize-loops
8650 Parallelize loops, i.e., split their iteration space to run in n threads.
8651 This is only possible for loops whose iterations are independent
8652 and can be arbitrarily reordered. The optimization is only
8653 profitable on multiprocessor machines, for loops that are CPU-intensive,
8654 rather than constrained e.g.@: by memory bandwidth. This option
8655 implies @option{-pthread}, and thus is only supported on targets
8656 that have support for @option{-pthread}.
8660 Perform function-local points-to analysis on trees. This flag is
8661 enabled by default at @option{-O} and higher.
8665 Perform scalar replacement of aggregates. This pass replaces structure
8666 references with scalars to prevent committing structures to memory too
8667 early. This flag is enabled by default at @option{-O} and higher.
8669 @item -ftree-copyrename
8670 @opindex ftree-copyrename
8671 Perform copy renaming on trees. This pass attempts to rename compiler
8672 temporaries to other variables at copy locations, usually resulting in
8673 variable names which more closely resemble the original variables. This flag
8674 is enabled by default at @option{-O} and higher.
8676 @item -ftree-coalesce-inlined-vars
8677 @opindex ftree-coalesce-inlined-vars
8678 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8679 combine small user-defined variables too, but only if they are inlined
8680 from other functions. It is a more limited form of
8681 @option{-ftree-coalesce-vars}. This may harm debug information of such
8682 inlined variables, but it keeps variables of the inlined-into
8683 function apart from each other, such that they are more likely to
8684 contain the expected values in a debugging session. This was the
8685 default in GCC versions older than 4.7.
8687 @item -ftree-coalesce-vars
8688 @opindex ftree-coalesce-vars
8689 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8690 combine small user-defined variables too, instead of just compiler
8691 temporaries. This may severely limit the ability to debug an optimized
8692 program compiled with @option{-fno-var-tracking-assignments}. In the
8693 negated form, this flag prevents SSA coalescing of user variables,
8694 including inlined ones. This option is enabled by default.
8698 Perform temporary expression replacement during the SSA->normal phase. Single
8699 use/single def temporaries are replaced at their use location with their
8700 defining expression. This results in non-GIMPLE code, but gives the expanders
8701 much more complex trees to work on resulting in better RTL generation. This is
8702 enabled by default at @option{-O} and higher.
8706 Perform straight-line strength reduction on trees. This recognizes related
8707 expressions involving multiplications and replaces them by less expensive
8708 calculations when possible. This is enabled by default at @option{-O} and
8711 @item -ftree-vectorize
8712 @opindex ftree-vectorize
8713 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8714 and @option{-ftree-slp-vectorize} if not explicitly specified.
8716 @item -ftree-loop-vectorize
8717 @opindex ftree-loop-vectorize
8718 Perform loop vectorization on trees. This flag is enabled by default at
8719 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8721 @item -ftree-slp-vectorize
8722 @opindex ftree-slp-vectorize
8723 Perform basic block vectorization on trees. This flag is enabled by default at
8724 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8726 @item -fvect-cost-model=@var{model}
8727 @opindex fvect-cost-model
8728 Alter the cost model used for vectorization. The @var{model} argument
8729 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8730 With the @samp{unlimited} model the vectorized code-path is assumed
8731 to be profitable while with the @samp{dynamic} model a runtime check
8732 guards the vectorized code-path to enable it only for iteration
8733 counts that will likely execute faster than when executing the original
8734 scalar loop. The @samp{cheap} model disables vectorization of
8735 loops where doing so would be cost prohibitive for example due to
8736 required runtime checks for data dependence or alignment but otherwise
8737 is equal to the @samp{dynamic} model.
8738 The default cost model depends on other optimization flags and is
8739 either @samp{dynamic} or @samp{cheap}.
8741 @item -fsimd-cost-model=@var{model}
8742 @opindex fsimd-cost-model
8743 Alter the cost model used for vectorization of loops marked with the OpenMP
8744 or Cilk Plus simd directive. The @var{model} argument should be one of
8745 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8746 have the same meaning as described in @option{-fvect-cost-model} and by
8747 default a cost model defined with @option{-fvect-cost-model} is used.
8751 Perform Value Range Propagation on trees. This is similar to the
8752 constant propagation pass, but instead of values, ranges of values are
8753 propagated. This allows the optimizers to remove unnecessary range
8754 checks like array bound checks and null pointer checks. This is
8755 enabled by default at @option{-O2} and higher. Null pointer check
8756 elimination is only done if @option{-fdelete-null-pointer-checks} is
8759 @item -fsplit-ivs-in-unroller
8760 @opindex fsplit-ivs-in-unroller
8761 Enables expression of values of induction variables in later iterations
8762 of the unrolled loop using the value in the first iteration. This breaks
8763 long dependency chains, thus improving efficiency of the scheduling passes.
8765 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8766 same effect. However, that is not reliable in cases where the loop body
8767 is more complicated than a single basic block. It also does not work at all
8768 on some architectures due to restrictions in the CSE pass.
8770 This optimization is enabled by default.
8772 @item -fvariable-expansion-in-unroller
8773 @opindex fvariable-expansion-in-unroller
8774 With this option, the compiler creates multiple copies of some
8775 local variables when unrolling a loop, which can result in superior code.
8777 @item -fpartial-inlining
8778 @opindex fpartial-inlining
8779 Inline parts of functions. This option has any effect only
8780 when inlining itself is turned on by the @option{-finline-functions}
8781 or @option{-finline-small-functions} options.
8783 Enabled at level @option{-O2}.
8785 @item -fpredictive-commoning
8786 @opindex fpredictive-commoning
8787 Perform predictive commoning optimization, i.e., reusing computations
8788 (especially memory loads and stores) performed in previous
8789 iterations of loops.
8791 This option is enabled at level @option{-O3}.
8793 @item -fprefetch-loop-arrays
8794 @opindex fprefetch-loop-arrays
8795 If supported by the target machine, generate instructions to prefetch
8796 memory to improve the performance of loops that access large arrays.
8798 This option may generate better or worse code; results are highly
8799 dependent on the structure of loops within the source code.
8801 Disabled at level @option{-Os}.
8804 @itemx -fno-peephole2
8805 @opindex fno-peephole
8806 @opindex fno-peephole2
8807 Disable any machine-specific peephole optimizations. The difference
8808 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8809 are implemented in the compiler; some targets use one, some use the
8810 other, a few use both.
8812 @option{-fpeephole} is enabled by default.
8813 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8815 @item -fno-guess-branch-probability
8816 @opindex fno-guess-branch-probability
8817 Do not guess branch probabilities using heuristics.
8819 GCC uses heuristics to guess branch probabilities if they are
8820 not provided by profiling feedback (@option{-fprofile-arcs}). These
8821 heuristics are based on the control flow graph. If some branch probabilities
8822 are specified by @code{__builtin_expect}, then the heuristics are
8823 used to guess branch probabilities for the rest of the control flow graph,
8824 taking the @code{__builtin_expect} info into account. The interactions
8825 between the heuristics and @code{__builtin_expect} can be complex, and in
8826 some cases, it may be useful to disable the heuristics so that the effects
8827 of @code{__builtin_expect} are easier to understand.
8829 The default is @option{-fguess-branch-probability} at levels
8830 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8832 @item -freorder-blocks
8833 @opindex freorder-blocks
8834 Reorder basic blocks in the compiled function in order to reduce number of
8835 taken branches and improve code locality.
8837 Enabled at levels @option{-O2}, @option{-O3}.
8839 @item -freorder-blocks-and-partition
8840 @opindex freorder-blocks-and-partition
8841 In addition to reordering basic blocks in the compiled function, in order
8842 to reduce number of taken branches, partitions hot and cold basic blocks
8843 into separate sections of the assembly and .o files, to improve
8844 paging and cache locality performance.
8846 This optimization is automatically turned off in the presence of
8847 exception handling, for linkonce sections, for functions with a user-defined
8848 section attribute and on any architecture that does not support named
8851 Enabled for x86 at levels @option{-O2}, @option{-O3}.
8853 @item -freorder-functions
8854 @opindex freorder-functions
8855 Reorder functions in the object file in order to
8856 improve code locality. This is implemented by using special
8857 subsections @code{.text.hot} for most frequently executed functions and
8858 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
8859 the linker so object file format must support named sections and linker must
8860 place them in a reasonable way.
8862 Also profile feedback must be available to make this option effective. See
8863 @option{-fprofile-arcs} for details.
8865 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8867 @item -fstrict-aliasing
8868 @opindex fstrict-aliasing
8869 Allow the compiler to assume the strictest aliasing rules applicable to
8870 the language being compiled. For C (and C++), this activates
8871 optimizations based on the type of expressions. In particular, an
8872 object of one type is assumed never to reside at the same address as an
8873 object of a different type, unless the types are almost the same. For
8874 example, an @code{unsigned int} can alias an @code{int}, but not a
8875 @code{void*} or a @code{double}. A character type may alias any other
8878 @anchor{Type-punning}Pay special attention to code like this:
8891 The practice of reading from a different union member than the one most
8892 recently written to (called ``type-punning'') is common. Even with
8893 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
8894 is accessed through the union type. So, the code above works as
8895 expected. @xref{Structures unions enumerations and bit-fields
8896 implementation}. However, this code might not:
8907 Similarly, access by taking the address, casting the resulting pointer
8908 and dereferencing the result has undefined behavior, even if the cast
8909 uses a union type, e.g.:
8913 return ((union a_union *) &d)->i;
8917 The @option{-fstrict-aliasing} option is enabled at levels
8918 @option{-O2}, @option{-O3}, @option{-Os}.
8920 @item -fstrict-overflow
8921 @opindex fstrict-overflow
8922 Allow the compiler to assume strict signed overflow rules, depending
8923 on the language being compiled. For C (and C++) this means that
8924 overflow when doing arithmetic with signed numbers is undefined, which
8925 means that the compiler may assume that it does not happen. This
8926 permits various optimizations. For example, the compiler assumes
8927 that an expression like @code{i + 10 > i} is always true for
8928 signed @code{i}. This assumption is only valid if signed overflow is
8929 undefined, as the expression is false if @code{i + 10} overflows when
8930 using twos complement arithmetic. When this option is in effect any
8931 attempt to determine whether an operation on signed numbers
8932 overflows must be written carefully to not actually involve overflow.
8934 This option also allows the compiler to assume strict pointer
8935 semantics: given a pointer to an object, if adding an offset to that
8936 pointer does not produce a pointer to the same object, the addition is
8937 undefined. This permits the compiler to conclude that @code{p + u >
8938 p} is always true for a pointer @code{p} and unsigned integer
8939 @code{u}. This assumption is only valid because pointer wraparound is
8940 undefined, as the expression is false if @code{p + u} overflows using
8941 twos complement arithmetic.
8943 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
8944 that integer signed overflow is fully defined: it wraps. When
8945 @option{-fwrapv} is used, there is no difference between
8946 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
8947 integers. With @option{-fwrapv} certain types of overflow are
8948 permitted. For example, if the compiler gets an overflow when doing
8949 arithmetic on constants, the overflowed value can still be used with
8950 @option{-fwrapv}, but not otherwise.
8952 The @option{-fstrict-overflow} option is enabled at levels
8953 @option{-O2}, @option{-O3}, @option{-Os}.
8955 @item -falign-functions
8956 @itemx -falign-functions=@var{n}
8957 @opindex falign-functions
8958 Align the start of functions to the next power-of-two greater than
8959 @var{n}, skipping up to @var{n} bytes. For instance,
8960 @option{-falign-functions=32} aligns functions to the next 32-byte
8961 boundary, but @option{-falign-functions=24} aligns to the next
8962 32-byte boundary only if this can be done by skipping 23 bytes or less.
8964 @option{-fno-align-functions} and @option{-falign-functions=1} are
8965 equivalent and mean that functions are not aligned.
8967 Some assemblers only support this flag when @var{n} is a power of two;
8968 in that case, it is rounded up.
8970 If @var{n} is not specified or is zero, use a machine-dependent default.
8972 Enabled at levels @option{-O2}, @option{-O3}.
8974 @item -falign-labels
8975 @itemx -falign-labels=@var{n}
8976 @opindex falign-labels
8977 Align all branch targets to a power-of-two boundary, skipping up to
8978 @var{n} bytes like @option{-falign-functions}. This option can easily
8979 make code slower, because it must insert dummy operations for when the
8980 branch target is reached in the usual flow of the code.
8982 @option{-fno-align-labels} and @option{-falign-labels=1} are
8983 equivalent and mean that labels are not aligned.
8985 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
8986 are greater than this value, then their values are used instead.
8988 If @var{n} is not specified or is zero, use a machine-dependent default
8989 which is very likely to be @samp{1}, meaning no alignment.
8991 Enabled at levels @option{-O2}, @option{-O3}.
8994 @itemx -falign-loops=@var{n}
8995 @opindex falign-loops
8996 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
8997 like @option{-falign-functions}. If the loops are
8998 executed many times, this makes up for any execution of the dummy
9001 @option{-fno-align-loops} and @option{-falign-loops=1} are
9002 equivalent and mean that loops are not aligned.
9004 If @var{n} is not specified or is zero, use a machine-dependent default.
9006 Enabled at levels @option{-O2}, @option{-O3}.
9009 @itemx -falign-jumps=@var{n}
9010 @opindex falign-jumps
9011 Align branch targets to a power-of-two boundary, for branch targets
9012 where the targets can only be reached by jumping, skipping up to @var{n}
9013 bytes like @option{-falign-functions}. In this case, no dummy operations
9016 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9017 equivalent and mean that loops are not aligned.
9019 If @var{n} is not specified or is zero, use a machine-dependent default.
9021 Enabled at levels @option{-O2}, @option{-O3}.
9023 @item -funit-at-a-time
9024 @opindex funit-at-a-time
9025 This option is left for compatibility reasons. @option{-funit-at-a-time}
9026 has no effect, while @option{-fno-unit-at-a-time} implies
9027 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9031 @item -fno-toplevel-reorder
9032 @opindex fno-toplevel-reorder
9033 Do not reorder top-level functions, variables, and @code{asm}
9034 statements. Output them in the same order that they appear in the
9035 input file. When this option is used, unreferenced static variables
9036 are not removed. This option is intended to support existing code
9037 that relies on a particular ordering. For new code, it is better to
9038 use attributes when possible.
9040 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9041 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9046 Constructs webs as commonly used for register allocation purposes and assign
9047 each web individual pseudo register. This allows the register allocation pass
9048 to operate on pseudos directly, but also strengthens several other optimization
9049 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9050 however, make debugging impossible, since variables no longer stay in a
9053 Enabled by default with @option{-funroll-loops}.
9055 @item -fwhole-program
9056 @opindex fwhole-program
9057 Assume that the current compilation unit represents the whole program being
9058 compiled. All public functions and variables with the exception of @code{main}
9059 and those merged by attribute @code{externally_visible} become static functions
9060 and in effect are optimized more aggressively by interprocedural optimizers.
9062 This option should not be used in combination with @option{-flto}.
9063 Instead relying on a linker plugin should provide safer and more precise
9066 @item -flto[=@var{n}]
9068 This option runs the standard link-time optimizer. When invoked
9069 with source code, it generates GIMPLE (one of GCC's internal
9070 representations) and writes it to special ELF sections in the object
9071 file. When the object files are linked together, all the function
9072 bodies are read from these ELF sections and instantiated as if they
9073 had been part of the same translation unit.
9075 To use the link-time optimizer, @option{-flto} and optimization
9076 options should be specified at compile time and during the final link.
9080 gcc -c -O2 -flto foo.c
9081 gcc -c -O2 -flto bar.c
9082 gcc -o myprog -flto -O2 foo.o bar.o
9085 The first two invocations to GCC save a bytecode representation
9086 of GIMPLE into special ELF sections inside @file{foo.o} and
9087 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9088 @file{foo.o} and @file{bar.o}, merges the two files into a single
9089 internal image, and compiles the result as usual. Since both
9090 @file{foo.o} and @file{bar.o} are merged into a single image, this
9091 causes all the interprocedural analyses and optimizations in GCC to
9092 work across the two files as if they were a single one. This means,
9093 for example, that the inliner is able to inline functions in
9094 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9096 Another (simpler) way to enable link-time optimization is:
9099 gcc -o myprog -flto -O2 foo.c bar.c
9102 The above generates bytecode for @file{foo.c} and @file{bar.c},
9103 merges them together into a single GIMPLE representation and optimizes
9104 them as usual to produce @file{myprog}.
9106 The only important thing to keep in mind is that to enable link-time
9107 optimizations you need to use the GCC driver to perform the link-step.
9108 GCC then automatically performs link-time optimization if any of the
9109 objects involved were compiled with the @option{-flto}. You generally
9110 should specify the optimization options to be used for link-time
9111 optimization though GCC tries to be clever at guessing an
9112 optimization level to use from the options used at compile-time
9113 if you fail to specify one at link-time. You can always override
9114 the automatic decision to do link-time optimization at link-time
9115 by passing @option{-fno-lto} to the link command.
9117 To make whole program optimization effective, it is necessary to make
9118 certain whole program assumptions. The compiler needs to know
9119 what functions and variables can be accessed by libraries and runtime
9120 outside of the link-time optimized unit. When supported by the linker,
9121 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9122 to the compiler about used and externally visible symbols. When
9123 the linker plugin is not available, @option{-fwhole-program} should be
9124 used to allow the compiler to make these assumptions, which leads
9125 to more aggressive optimization decisions.
9127 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9128 compiled with @option{-flto}, the generated object file is larger than
9129 a regular object file because it contains GIMPLE bytecodes and the usual
9130 final code (see @option{-ffat-lto-objects}. This means that
9131 object files with LTO information can be linked as normal object
9132 files; if @option{-fno-lto} is passed to the linker, no
9133 interprocedural optimizations are applied. Note that when
9134 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9135 but you cannot perform a regular, non-LTO link on them.
9137 Additionally, the optimization flags used to compile individual files
9138 are not necessarily related to those used at link time. For instance,
9141 gcc -c -O0 -ffat-lto-objects -flto foo.c
9142 gcc -c -O0 -ffat-lto-objects -flto bar.c
9143 gcc -o myprog -O3 foo.o bar.o
9146 This produces individual object files with unoptimized assembler
9147 code, but the resulting binary @file{myprog} is optimized at
9148 @option{-O3}. If, instead, the final binary is generated with
9149 @option{-fno-lto}, then @file{myprog} is not optimized.
9151 When producing the final binary, GCC only
9152 applies link-time optimizations to those files that contain bytecode.
9153 Therefore, you can mix and match object files and libraries with
9154 GIMPLE bytecodes and final object code. GCC automatically selects
9155 which files to optimize in LTO mode and which files to link without
9158 There are some code generation flags preserved by GCC when
9159 generating bytecodes, as they need to be used during the final link
9160 stage. Generally options specified at link-time override those
9161 specified at compile-time.
9163 If you do not specify an optimization level option @option{-O} at
9164 link-time then GCC computes one based on the optimization levels
9165 used when compiling the object files. The highest optimization
9168 Currently, the following options and their setting are take from
9169 the first object file that explicitely specified it:
9170 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9171 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9172 and all the @option{-m} target flags.
9174 Certain ABI changing flags are required to match in all compilation-units
9175 and trying to override this at link-time with a conflicting value
9176 is ignored. This includes options such as @option{-freg-struct-return}
9177 and @option{-fpcc-struct-return}.
9179 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9180 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9181 are passed through to the link stage and merged conservatively for
9182 conflicting translation units. Specifically
9183 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9184 precedence and for example @option{-ffp-contract=off} takes precedence
9185 over @option{-ffp-contract=fast}. You can override them at linke-time.
9187 It is recommended that you compile all the files participating in the
9188 same link with the same options and also specify those options at
9191 If LTO encounters objects with C linkage declared with incompatible
9192 types in separate translation units to be linked together (undefined
9193 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9194 issued. The behavior is still undefined at run time. Similar
9195 diagnostics may be raised for other languages.
9197 Another feature of LTO is that it is possible to apply interprocedural
9198 optimizations on files written in different languages:
9203 gfortran -c -flto baz.f90
9204 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9207 Notice that the final link is done with @command{g++} to get the C++
9208 runtime libraries and @option{-lgfortran} is added to get the Fortran
9209 runtime libraries. In general, when mixing languages in LTO mode, you
9210 should use the same link command options as when mixing languages in a
9211 regular (non-LTO) compilation.
9213 If object files containing GIMPLE bytecode are stored in a library archive, say
9214 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9215 are using a linker with plugin support. To create static libraries suitable
9216 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9217 and @command{ranlib};
9218 to show the symbols of object files with GIMPLE bytecode, use
9219 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9220 and @command{nm} have been compiled with plugin support. At link time, use the the
9221 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9222 the LTO optimization process:
9225 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9228 With the linker plugin enabled, the linker extracts the needed
9229 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9230 to make them part of the aggregated GIMPLE image to be optimized.
9232 If you are not using a linker with plugin support and/or do not
9233 enable the linker plugin, then the objects inside @file{libfoo.a}
9234 are extracted and linked as usual, but they do not participate
9235 in the LTO optimization process. In order to make a static library suitable
9236 for both LTO optimization and usual linkage, compile its object files with
9237 @option{-flto} @option{-ffat-lto-objects}.
9239 Link-time optimizations do not require the presence of the whole program to
9240 operate. If the program does not require any symbols to be exported, it is
9241 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9242 the interprocedural optimizers to use more aggressive assumptions which may
9243 lead to improved optimization opportunities.
9244 Use of @option{-fwhole-program} is not needed when linker plugin is
9245 active (see @option{-fuse-linker-plugin}).
9247 The current implementation of LTO makes no
9248 attempt to generate bytecode that is portable between different
9249 types of hosts. The bytecode files are versioned and there is a
9250 strict version check, so bytecode files generated in one version of
9251 GCC do not work with an older or newer version of GCC.
9253 Link-time optimization does not work well with generation of debugging
9254 information. Combining @option{-flto} with
9255 @option{-g} is currently experimental and expected to produce unexpected
9258 If you specify the optional @var{n}, the optimization and code
9259 generation done at link time is executed in parallel using @var{n}
9260 parallel jobs by utilizing an installed @command{make} program. The
9261 environment variable @env{MAKE} may be used to override the program
9262 used. The default value for @var{n} is 1.
9264 You can also specify @option{-flto=jobserver} to use GNU make's
9265 job server mode to determine the number of parallel jobs. This
9266 is useful when the Makefile calling GCC is already executing in parallel.
9267 You must prepend a @samp{+} to the command recipe in the parent Makefile
9268 for this to work. This option likely only works if @env{MAKE} is
9271 @item -flto-partition=@var{alg}
9272 @opindex flto-partition
9273 Specify the partitioning algorithm used by the link-time optimizer.
9274 The value is either @samp{1to1} to specify a partitioning mirroring
9275 the original source files or @samp{balanced} to specify partitioning
9276 into equally sized chunks (whenever possible) or @samp{max} to create
9277 new partition for every symbol where possible. Specifying @samp{none}
9278 as an algorithm disables partitioning and streaming completely.
9279 The default value is @samp{balanced}. While @samp{1to1} can be used
9280 as an workaround for various code ordering issues, the @samp{max}
9281 partitioning is intended for internal testing only.
9282 The value @samp{one} specifies that exactly one partition should be
9283 used while the value @samp{none} bypasses partitioning and executes
9284 the link-time optimization step directly from the WPA phase.
9286 @item -flto-odr-type-merging
9287 @opindex flto-odr-type-merging
9288 Enable streaming of mangled types names of C++ types and their unification
9289 at linktime. This increases size of LTO object files, but enable
9290 diagnostics about One Definition Rule violations.
9292 @item -flto-compression-level=@var{n}
9293 @opindex flto-compression-level
9294 This option specifies the level of compression used for intermediate
9295 language written to LTO object files, and is only meaningful in
9296 conjunction with LTO mode (@option{-flto}). Valid
9297 values are 0 (no compression) to 9 (maximum compression). Values
9298 outside this range are clamped to either 0 or 9. If the option is not
9299 given, a default balanced compression setting is used.
9302 @opindex flto-report
9303 Prints a report with internal details on the workings of the link-time
9304 optimizer. The contents of this report vary from version to version.
9305 It is meant to be useful to GCC developers when processing object
9306 files in LTO mode (via @option{-flto}).
9308 Disabled by default.
9310 @item -flto-report-wpa
9311 @opindex flto-report-wpa
9312 Like @option{-flto-report}, but only print for the WPA phase of Link
9315 @item -fuse-linker-plugin
9316 @opindex fuse-linker-plugin
9317 Enables the use of a linker plugin during link-time optimization. This
9318 option relies on plugin support in the linker, which is available in gold
9319 or in GNU ld 2.21 or newer.
9321 This option enables the extraction of object files with GIMPLE bytecode out
9322 of library archives. This improves the quality of optimization by exposing
9323 more code to the link-time optimizer. This information specifies what
9324 symbols can be accessed externally (by non-LTO object or during dynamic
9325 linking). Resulting code quality improvements on binaries (and shared
9326 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9327 See @option{-flto} for a description of the effect of this flag and how to
9330 This option is enabled by default when LTO support in GCC is enabled
9331 and GCC was configured for use with
9332 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9334 @item -ffat-lto-objects
9335 @opindex ffat-lto-objects
9336 Fat LTO objects are object files that contain both the intermediate language
9337 and the object code. This makes them usable for both LTO linking and normal
9338 linking. This option is effective only when compiling with @option{-flto}
9339 and is ignored at link time.
9341 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9342 requires the complete toolchain to be aware of LTO. It requires a linker with
9343 linker plugin support for basic functionality. Additionally,
9344 @command{nm}, @command{ar} and @command{ranlib}
9345 need to support linker plugins to allow a full-featured build environment
9346 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9347 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9348 to these tools. With non fat LTO makefiles need to be modified to use them.
9350 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9353 @item -fcompare-elim
9354 @opindex fcompare-elim
9355 After register allocation and post-register allocation instruction splitting,
9356 identify arithmetic instructions that compute processor flags similar to a
9357 comparison operation based on that arithmetic. If possible, eliminate the
9358 explicit comparison operation.
9360 This pass only applies to certain targets that cannot explicitly represent
9361 the comparison operation before register allocation is complete.
9363 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9365 @item -fcprop-registers
9366 @opindex fcprop-registers
9367 After register allocation and post-register allocation instruction splitting,
9368 perform a copy-propagation pass to try to reduce scheduling dependencies
9369 and occasionally eliminate the copy.
9371 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9373 @item -fprofile-correction
9374 @opindex fprofile-correction
9375 Profiles collected using an instrumented binary for multi-threaded programs may
9376 be inconsistent due to missed counter updates. When this option is specified,
9377 GCC uses heuristics to correct or smooth out such inconsistencies. By
9378 default, GCC emits an error message when an inconsistent profile is detected.
9380 @item -fprofile-dir=@var{path}
9381 @opindex fprofile-dir
9383 Set the directory to search for the profile data files in to @var{path}.
9384 This option affects only the profile data generated by
9385 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9386 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9387 and its related options. Both absolute and relative paths can be used.
9388 By default, GCC uses the current directory as @var{path}, thus the
9389 profile data file appears in the same directory as the object file.
9391 @item -fprofile-generate
9392 @itemx -fprofile-generate=@var{path}
9393 @opindex fprofile-generate
9395 Enable options usually used for instrumenting application to produce
9396 profile useful for later recompilation with profile feedback based
9397 optimization. You must use @option{-fprofile-generate} both when
9398 compiling and when linking your program.
9400 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9402 If @var{path} is specified, GCC looks at the @var{path} to find
9403 the profile feedback data files. See @option{-fprofile-dir}.
9406 @itemx -fprofile-use=@var{path}
9407 @opindex fprofile-use
9408 Enable profile feedback-directed optimizations,
9409 and the following optimizations
9410 which are generally profitable only with profile feedback available:
9411 @option{-fbranch-probabilities}, @option{-fvpt},
9412 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9413 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9415 By default, GCC emits an error message if the feedback profiles do not
9416 match the source code. This error can be turned into a warning by using
9417 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9420 If @var{path} is specified, GCC looks at the @var{path} to find
9421 the profile feedback data files. See @option{-fprofile-dir}.
9423 @item -fauto-profile
9424 @itemx -fauto-profile=@var{path}
9425 @opindex fauto-profile
9426 Enable sampling-based feedback-directed optimizations,
9427 and the following optimizations
9428 which are generally profitable only with profile feedback available:
9429 @option{-fbranch-probabilities}, @option{-fvpt},
9430 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9431 @option{-ftree-vectorize},
9432 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9433 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9434 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9436 @var{path} is the name of a file containing AutoFDO profile information.
9437 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9439 Producing an AutoFDO profile data file requires running your program
9440 with the @command{perf} utility on a supported GNU/Linux target system.
9441 For more information, see @uref{https://perf.wiki.kernel.org/}.
9445 perf record -e br_inst_retired:near_taken -b -o perf.data \
9449 Then use the @command{create_gcov} tool to convert the raw profile data
9450 to a format that can be used by GCC.@ You must also supply the
9451 unstripped binary for your program to this tool.
9452 See @uref{https://github.com/google/autofdo}.
9456 create_gcov --binary=your_program.unstripped --profile=perf.data \
9461 The following options control compiler behavior regarding floating-point
9462 arithmetic. These options trade off between speed and
9463 correctness. All must be specifically enabled.
9467 @opindex ffloat-store
9468 Do not store floating-point variables in registers, and inhibit other
9469 options that might change whether a floating-point value is taken from a
9472 @cindex floating-point precision
9473 This option prevents undesirable excess precision on machines such as
9474 the 68000 where the floating registers (of the 68881) keep more
9475 precision than a @code{double} is supposed to have. Similarly for the
9476 x86 architecture. For most programs, the excess precision does only
9477 good, but a few programs rely on the precise definition of IEEE floating
9478 point. Use @option{-ffloat-store} for such programs, after modifying
9479 them to store all pertinent intermediate computations into variables.
9481 @item -fexcess-precision=@var{style}
9482 @opindex fexcess-precision
9483 This option allows further control over excess precision on machines
9484 where floating-point registers have more precision than the IEEE
9485 @code{float} and @code{double} types and the processor does not
9486 support operations rounding to those types. By default,
9487 @option{-fexcess-precision=fast} is in effect; this means that
9488 operations are carried out in the precision of the registers and that
9489 it is unpredictable when rounding to the types specified in the source
9490 code takes place. When compiling C, if
9491 @option{-fexcess-precision=standard} is specified then excess
9492 precision follows the rules specified in ISO C99; in particular,
9493 both casts and assignments cause values to be rounded to their
9494 semantic types (whereas @option{-ffloat-store} only affects
9495 assignments). This option is enabled by default for C if a strict
9496 conformance option such as @option{-std=c99} is used.
9499 @option{-fexcess-precision=standard} is not implemented for languages
9500 other than C, and has no effect if
9501 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9502 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9503 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9504 semantics apply without excess precision, and in the latter, rounding
9509 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9510 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9511 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9513 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9515 This option is not turned on by any @option{-O} option besides
9516 @option{-Ofast} since it can result in incorrect output for programs
9517 that depend on an exact implementation of IEEE or ISO rules/specifications
9518 for math functions. It may, however, yield faster code for programs
9519 that do not require the guarantees of these specifications.
9521 @item -fno-math-errno
9522 @opindex fno-math-errno
9523 Do not set @code{errno} after calling math functions that are executed
9524 with a single instruction, e.g., @code{sqrt}. A program that relies on
9525 IEEE exceptions for math error handling may want to use this flag
9526 for speed while maintaining IEEE arithmetic compatibility.
9528 This option is not turned on by any @option{-O} option since
9529 it can result in incorrect output for programs that depend on
9530 an exact implementation of IEEE or ISO rules/specifications for
9531 math functions. It may, however, yield faster code for programs
9532 that do not require the guarantees of these specifications.
9534 The default is @option{-fmath-errno}.
9536 On Darwin systems, the math library never sets @code{errno}. There is
9537 therefore no reason for the compiler to consider the possibility that
9538 it might, and @option{-fno-math-errno} is the default.
9540 @item -funsafe-math-optimizations
9541 @opindex funsafe-math-optimizations
9543 Allow optimizations for floating-point arithmetic that (a) assume
9544 that arguments and results are valid and (b) may violate IEEE or
9545 ANSI standards. When used at link-time, it may include libraries
9546 or startup files that change the default FPU control word or other
9547 similar optimizations.
9549 This option is not turned on by any @option{-O} option since
9550 it can result in incorrect output for programs that depend on
9551 an exact implementation of IEEE or ISO rules/specifications for
9552 math functions. It may, however, yield faster code for programs
9553 that do not require the guarantees of these specifications.
9554 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9555 @option{-fassociative-math} and @option{-freciprocal-math}.
9557 The default is @option{-fno-unsafe-math-optimizations}.
9559 @item -fassociative-math
9560 @opindex fassociative-math
9562 Allow re-association of operands in series of floating-point operations.
9563 This violates the ISO C and C++ language standard by possibly changing
9564 computation result. NOTE: re-ordering may change the sign of zero as
9565 well as ignore NaNs and inhibit or create underflow or overflow (and
9566 thus cannot be used on code that relies on rounding behavior like
9567 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9568 and thus may not be used when ordered comparisons are required.
9569 This option requires that both @option{-fno-signed-zeros} and
9570 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9571 much sense with @option{-frounding-math}. For Fortran the option
9572 is automatically enabled when both @option{-fno-signed-zeros} and
9573 @option{-fno-trapping-math} are in effect.
9575 The default is @option{-fno-associative-math}.
9577 @item -freciprocal-math
9578 @opindex freciprocal-math
9580 Allow the reciprocal of a value to be used instead of dividing by
9581 the value if this enables optimizations. For example @code{x / y}
9582 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9583 is subject to common subexpression elimination. Note that this loses
9584 precision and increases the number of flops operating on the value.
9586 The default is @option{-fno-reciprocal-math}.
9588 @item -ffinite-math-only
9589 @opindex ffinite-math-only
9590 Allow optimizations for floating-point arithmetic that assume
9591 that arguments and results are not NaNs or +-Infs.
9593 This option is not turned on by any @option{-O} option since
9594 it can result in incorrect output for programs that depend on
9595 an exact implementation of IEEE or ISO rules/specifications for
9596 math functions. It may, however, yield faster code for programs
9597 that do not require the guarantees of these specifications.
9599 The default is @option{-fno-finite-math-only}.
9601 @item -fno-signed-zeros
9602 @opindex fno-signed-zeros
9603 Allow optimizations for floating-point arithmetic that ignore the
9604 signedness of zero. IEEE arithmetic specifies the behavior of
9605 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9606 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9607 This option implies that the sign of a zero result isn't significant.
9609 The default is @option{-fsigned-zeros}.
9611 @item -fno-trapping-math
9612 @opindex fno-trapping-math
9613 Compile code assuming that floating-point operations cannot generate
9614 user-visible traps. These traps include division by zero, overflow,
9615 underflow, inexact result and invalid operation. This option requires
9616 that @option{-fno-signaling-nans} be in effect. Setting this option may
9617 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9619 This option should never be turned on by any @option{-O} option since
9620 it can result in incorrect output for programs that depend on
9621 an exact implementation of IEEE or ISO rules/specifications for
9624 The default is @option{-ftrapping-math}.
9626 @item -frounding-math
9627 @opindex frounding-math
9628 Disable transformations and optimizations that assume default floating-point
9629 rounding behavior. This is round-to-zero for all floating point
9630 to integer conversions, and round-to-nearest for all other arithmetic
9631 truncations. This option should be specified for programs that change
9632 the FP rounding mode dynamically, or that may be executed with a
9633 non-default rounding mode. This option disables constant folding of
9634 floating-point expressions at compile time (which may be affected by
9635 rounding mode) and arithmetic transformations that are unsafe in the
9636 presence of sign-dependent rounding modes.
9638 The default is @option{-fno-rounding-math}.
9640 This option is experimental and does not currently guarantee to
9641 disable all GCC optimizations that are affected by rounding mode.
9642 Future versions of GCC may provide finer control of this setting
9643 using C99's @code{FENV_ACCESS} pragma. This command-line option
9644 will be used to specify the default state for @code{FENV_ACCESS}.
9646 @item -fsignaling-nans
9647 @opindex fsignaling-nans
9648 Compile code assuming that IEEE signaling NaNs may generate user-visible
9649 traps during floating-point operations. Setting this option disables
9650 optimizations that may change the number of exceptions visible with
9651 signaling NaNs. This option implies @option{-ftrapping-math}.
9653 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9656 The default is @option{-fno-signaling-nans}.
9658 This option is experimental and does not currently guarantee to
9659 disable all GCC optimizations that affect signaling NaN behavior.
9661 @item -fsingle-precision-constant
9662 @opindex fsingle-precision-constant
9663 Treat floating-point constants as single precision instead of
9664 implicitly converting them to double-precision constants.
9666 @item -fcx-limited-range
9667 @opindex fcx-limited-range
9668 When enabled, this option states that a range reduction step is not
9669 needed when performing complex division. Also, there is no checking
9670 whether the result of a complex multiplication or division is @code{NaN
9671 + I*NaN}, with an attempt to rescue the situation in that case. The
9672 default is @option{-fno-cx-limited-range}, but is enabled by
9673 @option{-ffast-math}.
9675 This option controls the default setting of the ISO C99
9676 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9679 @item -fcx-fortran-rules
9680 @opindex fcx-fortran-rules
9681 Complex multiplication and division follow Fortran rules. Range
9682 reduction is done as part of complex division, but there is no checking
9683 whether the result of a complex multiplication or division is @code{NaN
9684 + I*NaN}, with an attempt to rescue the situation in that case.
9686 The default is @option{-fno-cx-fortran-rules}.
9690 The following options control optimizations that may improve
9691 performance, but are not enabled by any @option{-O} options. This
9692 section includes experimental options that may produce broken code.
9695 @item -fbranch-probabilities
9696 @opindex fbranch-probabilities
9697 After running a program compiled with @option{-fprofile-arcs}
9698 (@pxref{Debugging Options,, Options for Debugging Your Program or
9699 @command{gcc}}), you can compile it a second time using
9700 @option{-fbranch-probabilities}, to improve optimizations based on
9701 the number of times each branch was taken. When a program
9702 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9703 counts to a file called @file{@var{sourcename}.gcda} for each source
9704 file. The information in this data file is very dependent on the
9705 structure of the generated code, so you must use the same source code
9706 and the same optimization options for both compilations.
9708 With @option{-fbranch-probabilities}, GCC puts a
9709 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9710 These can be used to improve optimization. Currently, they are only
9711 used in one place: in @file{reorg.c}, instead of guessing which path a
9712 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9713 exactly determine which path is taken more often.
9715 @item -fprofile-values
9716 @opindex fprofile-values
9717 If combined with @option{-fprofile-arcs}, it adds code so that some
9718 data about values of expressions in the program is gathered.
9720 With @option{-fbranch-probabilities}, it reads back the data gathered
9721 from profiling values of expressions for usage in optimizations.
9723 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9725 @item -fprofile-reorder-functions
9726 @opindex fprofile-reorder-functions
9727 Function reordering based on profile instrumentation collects
9728 first time of execution of a function and orders these functions
9731 Enabled with @option{-fprofile-use}.
9735 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9736 to add code to gather information about values of expressions.
9738 With @option{-fbranch-probabilities}, it reads back the data gathered
9739 and actually performs the optimizations based on them.
9740 Currently the optimizations include specialization of division operations
9741 using the knowledge about the value of the denominator.
9743 @item -frename-registers
9744 @opindex frename-registers
9745 Attempt to avoid false dependencies in scheduled code by making use
9746 of registers left over after register allocation. This optimization
9747 most benefits processors with lots of registers. Depending on the
9748 debug information format adopted by the target, however, it can
9749 make debugging impossible, since variables no longer stay in
9750 a ``home register''.
9752 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9754 @item -fschedule-fusion
9755 @opindex fschedule-fusion
9756 Performs a target dependent pass over the instruction stream to schedule
9757 instructions of same type together because target machine can execute them
9758 more efficiently if they are adjacent to each other in the instruction flow.
9760 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9764 Perform tail duplication to enlarge superblock size. This transformation
9765 simplifies the control flow of the function allowing other optimizations to do
9768 Enabled with @option{-fprofile-use}.
9770 @item -funroll-loops
9771 @opindex funroll-loops
9772 Unroll loops whose number of iterations can be determined at compile time or
9773 upon entry to the loop. @option{-funroll-loops} implies
9774 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9775 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9776 a small constant number of iterations). This option makes code larger, and may
9777 or may not make it run faster.
9779 Enabled with @option{-fprofile-use}.
9781 @item -funroll-all-loops
9782 @opindex funroll-all-loops
9783 Unroll all loops, even if their number of iterations is uncertain when
9784 the loop is entered. This usually makes programs run more slowly.
9785 @option{-funroll-all-loops} implies the same options as
9786 @option{-funroll-loops}.
9789 @opindex fpeel-loops
9790 Peels loops for which there is enough information that they do not
9791 roll much (from profile feedback). It also turns on complete loop peeling
9792 (i.e.@: complete removal of loops with small constant number of iterations).
9794 Enabled with @option{-fprofile-use}.
9796 @item -fmove-loop-invariants
9797 @opindex fmove-loop-invariants
9798 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9799 at level @option{-O1}
9801 @item -funswitch-loops
9802 @opindex funswitch-loops
9803 Move branches with loop invariant conditions out of the loop, with duplicates
9804 of the loop on both branches (modified according to result of the condition).
9806 @item -ffunction-sections
9807 @itemx -fdata-sections
9808 @opindex ffunction-sections
9809 @opindex fdata-sections
9810 Place each function or data item into its own section in the output
9811 file if the target supports arbitrary sections. The name of the
9812 function or the name of the data item determines the section's name
9815 Use these options on systems where the linker can perform optimizations
9816 to improve locality of reference in the instruction space. Most systems
9817 using the ELF object format and SPARC processors running Solaris 2 have
9818 linkers with such optimizations. AIX may have these optimizations in
9821 Only use these options when there are significant benefits from doing
9822 so. When you specify these options, the assembler and linker
9823 create larger object and executable files and are also slower.
9824 You cannot use @command{gprof} on all systems if you
9825 specify this option, and you may have problems with debugging if
9826 you specify both this option and @option{-g}.
9828 @item -fbranch-target-load-optimize
9829 @opindex fbranch-target-load-optimize
9830 Perform branch target register load optimization before prologue / epilogue
9832 The use of target registers can typically be exposed only during reload,
9833 thus hoisting loads out of loops and doing inter-block scheduling needs
9834 a separate optimization pass.
9836 @item -fbranch-target-load-optimize2
9837 @opindex fbranch-target-load-optimize2
9838 Perform branch target register load optimization after prologue / epilogue
9841 @item -fbtr-bb-exclusive
9842 @opindex fbtr-bb-exclusive
9843 When performing branch target register load optimization, don't reuse
9844 branch target registers within any basic block.
9846 @item -fstack-protector
9847 @opindex fstack-protector
9848 Emit extra code to check for buffer overflows, such as stack smashing
9849 attacks. This is done by adding a guard variable to functions with
9850 vulnerable objects. This includes functions that call @code{alloca}, and
9851 functions with buffers larger than 8 bytes. The guards are initialized
9852 when a function is entered and then checked when the function exits.
9853 If a guard check fails, an error message is printed and the program exits.
9855 @item -fstack-protector-all
9856 @opindex fstack-protector-all
9857 Like @option{-fstack-protector} except that all functions are protected.
9859 @item -fstack-protector-strong
9860 @opindex fstack-protector-strong
9861 Like @option{-fstack-protector} but includes additional functions to
9862 be protected --- those that have local array definitions, or have
9863 references to local frame addresses.
9865 @item -fstack-protector-explicit
9866 @opindex fstack-protector-explicit
9867 Like @option{-fstack-protector} but only protects those functions which
9868 have the @code{stack_protect} attribute
9871 @opindex fstdarg-opt
9872 Optimize the prologue of variadic argument functions with respect to usage of
9875 @item -fsection-anchors
9876 @opindex fsection-anchors
9877 Try to reduce the number of symbolic address calculations by using
9878 shared ``anchor'' symbols to address nearby objects. This transformation
9879 can help to reduce the number of GOT entries and GOT accesses on some
9882 For example, the implementation of the following function @code{foo}:
9886 int foo (void) @{ return a + b + c; @}
9890 usually calculates the addresses of all three variables, but if you
9891 compile it with @option{-fsection-anchors}, it accesses the variables
9892 from a common anchor point instead. The effect is similar to the
9893 following pseudocode (which isn't valid C):
9898 register int *xr = &x;
9899 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
9903 Not all targets support this option.
9905 @item --param @var{name}=@var{value}
9907 In some places, GCC uses various constants to control the amount of
9908 optimization that is done. For example, GCC does not inline functions
9909 that contain more than a certain number of instructions. You can
9910 control some of these constants on the command line using the
9911 @option{--param} option.
9913 The names of specific parameters, and the meaning of the values, are
9914 tied to the internals of the compiler, and are subject to change
9915 without notice in future releases.
9917 In each case, the @var{value} is an integer. The allowable choices for
9921 @item predictable-branch-outcome
9922 When branch is predicted to be taken with probability lower than this threshold
9923 (in percent), then it is considered well predictable. The default is 10.
9925 @item max-crossjump-edges
9926 The maximum number of incoming edges to consider for cross-jumping.
9927 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
9928 the number of edges incoming to each block. Increasing values mean
9929 more aggressive optimization, making the compilation time increase with
9930 probably small improvement in executable size.
9932 @item min-crossjump-insns
9933 The minimum number of instructions that must be matched at the end
9934 of two blocks before cross-jumping is performed on them. This
9935 value is ignored in the case where all instructions in the block being
9936 cross-jumped from are matched. The default value is 5.
9938 @item max-grow-copy-bb-insns
9939 The maximum code size expansion factor when copying basic blocks
9940 instead of jumping. The expansion is relative to a jump instruction.
9941 The default value is 8.
9943 @item max-goto-duplication-insns
9944 The maximum number of instructions to duplicate to a block that jumps
9945 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
9946 passes, GCC factors computed gotos early in the compilation process,
9947 and unfactors them as late as possible. Only computed jumps at the
9948 end of a basic blocks with no more than max-goto-duplication-insns are
9949 unfactored. The default value is 8.
9951 @item max-delay-slot-insn-search
9952 The maximum number of instructions to consider when looking for an
9953 instruction to fill a delay slot. If more than this arbitrary number of
9954 instructions are searched, the time savings from filling the delay slot
9955 are minimal, so stop searching. Increasing values mean more
9956 aggressive optimization, making the compilation time increase with probably
9957 small improvement in execution time.
9959 @item max-delay-slot-live-search
9960 When trying to fill delay slots, the maximum number of instructions to
9961 consider when searching for a block with valid live register
9962 information. Increasing this arbitrarily chosen value means more
9963 aggressive optimization, increasing the compilation time. This parameter
9964 should be removed when the delay slot code is rewritten to maintain the
9967 @item max-gcse-memory
9968 The approximate maximum amount of memory that can be allocated in
9969 order to perform the global common subexpression elimination
9970 optimization. If more memory than specified is required, the
9971 optimization is not done.
9973 @item max-gcse-insertion-ratio
9974 If the ratio of expression insertions to deletions is larger than this value
9975 for any expression, then RTL PRE inserts or removes the expression and thus
9976 leaves partially redundant computations in the instruction stream. The default value is 20.
9978 @item max-pending-list-length
9979 The maximum number of pending dependencies scheduling allows
9980 before flushing the current state and starting over. Large functions
9981 with few branches or calls can create excessively large lists which
9982 needlessly consume memory and resources.
9984 @item max-modulo-backtrack-attempts
9985 The maximum number of backtrack attempts the scheduler should make
9986 when modulo scheduling a loop. Larger values can exponentially increase
9989 @item max-inline-insns-single
9990 Several parameters control the tree inliner used in GCC@.
9991 This number sets the maximum number of instructions (counted in GCC's
9992 internal representation) in a single function that the tree inliner
9993 considers for inlining. This only affects functions declared
9994 inline and methods implemented in a class declaration (C++).
9995 The default value is 400.
9997 @item max-inline-insns-auto
9998 When you use @option{-finline-functions} (included in @option{-O3}),
9999 a lot of functions that would otherwise not be considered for inlining
10000 by the compiler are investigated. To those functions, a different
10001 (more restrictive) limit compared to functions declared inline can
10003 The default value is 40.
10005 @item inline-min-speedup
10006 When estimated performance improvement of caller + callee runtime exceeds this
10007 threshold (in precent), the function can be inlined regardless the limit on
10008 @option{--param max-inline-insns-single} and @option{--param
10009 max-inline-insns-auto}.
10011 @item large-function-insns
10012 The limit specifying really large functions. For functions larger than this
10013 limit after inlining, inlining is constrained by
10014 @option{--param large-function-growth}. This parameter is useful primarily
10015 to avoid extreme compilation time caused by non-linear algorithms used by the
10017 The default value is 2700.
10019 @item large-function-growth
10020 Specifies maximal growth of large function caused by inlining in percents.
10021 The default value is 100 which limits large function growth to 2.0 times
10024 @item large-unit-insns
10025 The limit specifying large translation unit. Growth caused by inlining of
10026 units larger than this limit is limited by @option{--param inline-unit-growth}.
10027 For small units this might be too tight.
10028 For example, consider a unit consisting of function A
10029 that is inline and B that just calls A three times. If B is small relative to
10030 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10031 large units consisting of small inlineable functions, however, the overall unit
10032 growth limit is needed to avoid exponential explosion of code size. Thus for
10033 smaller units, the size is increased to @option{--param large-unit-insns}
10034 before applying @option{--param inline-unit-growth}. The default is 10000.
10036 @item inline-unit-growth
10037 Specifies maximal overall growth of the compilation unit caused by inlining.
10038 The default value is 15 which limits unit growth to 1.15 times the original
10039 size. Cold functions (either marked cold via an attribute or by profile
10040 feedback) are not accounted into the unit size.
10042 @item ipcp-unit-growth
10043 Specifies maximal overall growth of the compilation unit caused by
10044 interprocedural constant propagation. The default value is 10 which limits
10045 unit growth to 1.1 times the original size.
10047 @item large-stack-frame
10048 The limit specifying large stack frames. While inlining the algorithm is trying
10049 to not grow past this limit too much. The default value is 256 bytes.
10051 @item large-stack-frame-growth
10052 Specifies maximal growth of large stack frames caused by inlining in percents.
10053 The default value is 1000 which limits large stack frame growth to 11 times
10056 @item max-inline-insns-recursive
10057 @itemx max-inline-insns-recursive-auto
10058 Specifies the maximum number of instructions an out-of-line copy of a
10059 self-recursive inline
10060 function can grow into by performing recursive inlining.
10062 @option{--param max-inline-insns-recursive} applies to functions
10064 For functions not declared inline, recursive inlining
10065 happens only when @option{-finline-functions} (included in @option{-O3}) is
10066 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10067 default value is 450.
10069 @item max-inline-recursive-depth
10070 @itemx max-inline-recursive-depth-auto
10071 Specifies the maximum recursion depth used for recursive inlining.
10073 @option{--param max-inline-recursive-depth} applies to functions
10074 declared inline. For functions not declared inline, recursive inlining
10075 happens only when @option{-finline-functions} (included in @option{-O3}) is
10076 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10077 default value is 8.
10079 @item min-inline-recursive-probability
10080 Recursive inlining is profitable only for function having deep recursion
10081 in average and can hurt for function having little recursion depth by
10082 increasing the prologue size or complexity of function body to other
10085 When profile feedback is available (see @option{-fprofile-generate}) the actual
10086 recursion depth can be guessed from probability that function recurses via a
10087 given call expression. This parameter limits inlining only to call expressions
10088 whose probability exceeds the given threshold (in percents).
10089 The default value is 10.
10091 @item early-inlining-insns
10092 Specify growth that the early inliner can make. In effect it increases
10093 the amount of inlining for code having a large abstraction penalty.
10094 The default value is 14.
10096 @item max-early-inliner-iterations
10097 Limit of iterations of the early inliner. This basically bounds
10098 the number of nested indirect calls the early inliner can resolve.
10099 Deeper chains are still handled by late inlining.
10101 @item comdat-sharing-probability
10102 Probability (in percent) that C++ inline function with comdat visibility
10103 are shared across multiple compilation units. The default value is 20.
10105 @item profile-func-internal-id
10106 A parameter to control whether to use function internal id in profile
10107 database lookup. If the value is 0, the compiler uses an id that
10108 is based on function assembler name and filename, which makes old profile
10109 data more tolerant to source changes such as function reordering etc.
10110 The default value is 0.
10112 @item min-vect-loop-bound
10113 The minimum number of iterations under which loops are not vectorized
10114 when @option{-ftree-vectorize} is used. The number of iterations after
10115 vectorization needs to be greater than the value specified by this option
10116 to allow vectorization. The default value is 0.
10118 @item gcse-cost-distance-ratio
10119 Scaling factor in calculation of maximum distance an expression
10120 can be moved by GCSE optimizations. This is currently supported only in the
10121 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10122 is with simple expressions, i.e., the expressions that have cost
10123 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10124 hoisting of simple expressions. The default value is 10.
10126 @item gcse-unrestricted-cost
10127 Cost, roughly measured as the cost of a single typical machine
10128 instruction, at which GCSE optimizations do not constrain
10129 the distance an expression can travel. This is currently
10130 supported only in the code hoisting pass. The lesser the cost,
10131 the more aggressive code hoisting is. Specifying 0
10132 allows all expressions to travel unrestricted distances.
10133 The default value is 3.
10135 @item max-hoist-depth
10136 The depth of search in the dominator tree for expressions to hoist.
10137 This is used to avoid quadratic behavior in hoisting algorithm.
10138 The value of 0 does not limit on the search, but may slow down compilation
10139 of huge functions. The default value is 30.
10141 @item max-tail-merge-comparisons
10142 The maximum amount of similar bbs to compare a bb with. This is used to
10143 avoid quadratic behavior in tree tail merging. The default value is 10.
10145 @item max-tail-merge-iterations
10146 The maximum amount of iterations of the pass over the function. This is used to
10147 limit compilation time in tree tail merging. The default value is 2.
10149 @item max-unrolled-insns
10150 The maximum number of instructions that a loop may have to be unrolled.
10151 If a loop is unrolled, this parameter also determines how many times
10152 the loop code is unrolled.
10154 @item max-average-unrolled-insns
10155 The maximum number of instructions biased by probabilities of their execution
10156 that a loop may have to be unrolled. If a loop is unrolled,
10157 this parameter also determines how many times the loop code is unrolled.
10159 @item max-unroll-times
10160 The maximum number of unrollings of a single loop.
10162 @item max-peeled-insns
10163 The maximum number of instructions that a loop may have to be peeled.
10164 If a loop is peeled, this parameter also determines how many times
10165 the loop code is peeled.
10167 @item max-peel-times
10168 The maximum number of peelings of a single loop.
10170 @item max-peel-branches
10171 The maximum number of branches on the hot path through the peeled sequence.
10173 @item max-completely-peeled-insns
10174 The maximum number of insns of a completely peeled loop.
10176 @item max-completely-peel-times
10177 The maximum number of iterations of a loop to be suitable for complete peeling.
10179 @item max-completely-peel-loop-nest-depth
10180 The maximum depth of a loop nest suitable for complete peeling.
10182 @item max-unswitch-insns
10183 The maximum number of insns of an unswitched loop.
10185 @item max-unswitch-level
10186 The maximum number of branches unswitched in a single loop.
10188 @item lim-expensive
10189 The minimum cost of an expensive expression in the loop invariant motion.
10191 @item iv-consider-all-candidates-bound
10192 Bound on number of candidates for induction variables, below which
10193 all candidates are considered for each use in induction variable
10194 optimizations. If there are more candidates than this,
10195 only the most relevant ones are considered to avoid quadratic time complexity.
10197 @item iv-max-considered-uses
10198 The induction variable optimizations give up on loops that contain more
10199 induction variable uses.
10201 @item iv-always-prune-cand-set-bound
10202 If the number of candidates in the set is smaller than this value,
10203 always try to remove unnecessary ivs from the set
10204 when adding a new one.
10206 @item scev-max-expr-size
10207 Bound on size of expressions used in the scalar evolutions analyzer.
10208 Large expressions slow the analyzer.
10210 @item scev-max-expr-complexity
10211 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10212 Complex expressions slow the analyzer.
10214 @item omega-max-vars
10215 The maximum number of variables in an Omega constraint system.
10216 The default value is 128.
10218 @item omega-max-geqs
10219 The maximum number of inequalities in an Omega constraint system.
10220 The default value is 256.
10222 @item omega-max-eqs
10223 The maximum number of equalities in an Omega constraint system.
10224 The default value is 128.
10226 @item omega-max-wild-cards
10227 The maximum number of wildcard variables that the Omega solver is
10228 able to insert. The default value is 18.
10230 @item omega-hash-table-size
10231 The size of the hash table in the Omega solver. The default value is
10234 @item omega-max-keys
10235 The maximal number of keys used by the Omega solver. The default
10238 @item omega-eliminate-redundant-constraints
10239 When set to 1, use expensive methods to eliminate all redundant
10240 constraints. The default value is 0.
10242 @item vect-max-version-for-alignment-checks
10243 The maximum number of run-time checks that can be performed when
10244 doing loop versioning for alignment in the vectorizer.
10246 @item vect-max-version-for-alias-checks
10247 The maximum number of run-time checks that can be performed when
10248 doing loop versioning for alias in the vectorizer.
10250 @item vect-max-peeling-for-alignment
10251 The maximum number of loop peels to enhance access alignment
10252 for vectorizer. Value -1 means 'no limit'.
10254 @item max-iterations-to-track
10255 The maximum number of iterations of a loop the brute-force algorithm
10256 for analysis of the number of iterations of the loop tries to evaluate.
10258 @item hot-bb-count-ws-permille
10259 A basic block profile count is considered hot if it contributes to
10260 the given permillage (i.e. 0...1000) of the entire profiled execution.
10262 @item hot-bb-frequency-fraction
10263 Select fraction of the entry block frequency of executions of basic block in
10264 function given basic block needs to have to be considered hot.
10266 @item max-predicted-iterations
10267 The maximum number of loop iterations we predict statically. This is useful
10268 in cases where a function contains a single loop with known bound and
10269 another loop with unknown bound.
10270 The known number of iterations is predicted correctly, while
10271 the unknown number of iterations average to roughly 10. This means that the
10272 loop without bounds appears artificially cold relative to the other one.
10274 @item builtin-expect-probability
10275 Control the probability of the expression having the specified value. This
10276 parameter takes a percentage (i.e. 0 ... 100) as input.
10277 The default probability of 90 is obtained empirically.
10279 @item align-threshold
10281 Select fraction of the maximal frequency of executions of a basic block in
10282 a function to align the basic block.
10284 @item align-loop-iterations
10286 A loop expected to iterate at least the selected number of iterations is
10289 @item tracer-dynamic-coverage
10290 @itemx tracer-dynamic-coverage-feedback
10292 This value is used to limit superblock formation once the given percentage of
10293 executed instructions is covered. This limits unnecessary code size
10296 The @option{tracer-dynamic-coverage-feedback} is used only when profile
10297 feedback is available. The real profiles (as opposed to statically estimated
10298 ones) are much less balanced allowing the threshold to be larger value.
10300 @item tracer-max-code-growth
10301 Stop tail duplication once code growth has reached given percentage. This is
10302 a rather artificial limit, as most of the duplicates are eliminated later in
10303 cross jumping, so it may be set to much higher values than is the desired code
10306 @item tracer-min-branch-ratio
10308 Stop reverse growth when the reverse probability of best edge is less than this
10309 threshold (in percent).
10311 @item tracer-min-branch-ratio
10312 @itemx tracer-min-branch-ratio-feedback
10314 Stop forward growth if the best edge has probability lower than this
10317 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10318 compilation for profile feedback and one for compilation without. The value
10319 for compilation with profile feedback needs to be more conservative (higher) in
10320 order to make tracer effective.
10322 @item max-cse-path-length
10324 The maximum number of basic blocks on path that CSE considers.
10327 @item max-cse-insns
10328 The maximum number of instructions CSE processes before flushing.
10329 The default is 1000.
10331 @item ggc-min-expand
10333 GCC uses a garbage collector to manage its own memory allocation. This
10334 parameter specifies the minimum percentage by which the garbage
10335 collector's heap should be allowed to expand between collections.
10336 Tuning this may improve compilation speed; it has no effect on code
10339 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10340 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10341 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10342 GCC is not able to calculate RAM on a particular platform, the lower
10343 bound of 30% is used. Setting this parameter and
10344 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10345 every opportunity. This is extremely slow, but can be useful for
10348 @item ggc-min-heapsize
10350 Minimum size of the garbage collector's heap before it begins bothering
10351 to collect garbage. The first collection occurs after the heap expands
10352 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10353 tuning this may improve compilation speed, and has no effect on code
10356 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10357 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10358 with a lower bound of 4096 (four megabytes) and an upper bound of
10359 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10360 particular platform, the lower bound is used. Setting this parameter
10361 very large effectively disables garbage collection. Setting this
10362 parameter and @option{ggc-min-expand} to zero causes a full collection
10363 to occur at every opportunity.
10365 @item max-reload-search-insns
10366 The maximum number of instruction reload should look backward for equivalent
10367 register. Increasing values mean more aggressive optimization, making the
10368 compilation time increase with probably slightly better performance.
10369 The default value is 100.
10371 @item max-cselib-memory-locations
10372 The maximum number of memory locations cselib should take into account.
10373 Increasing values mean more aggressive optimization, making the compilation time
10374 increase with probably slightly better performance. The default value is 500.
10376 @item reorder-blocks-duplicate
10377 @itemx reorder-blocks-duplicate-feedback
10379 Used by the basic block reordering pass to decide whether to use unconditional
10380 branch or duplicate the code on its destination. Code is duplicated when its
10381 estimated size is smaller than this value multiplied by the estimated size of
10382 unconditional jump in the hot spots of the program.
10384 The @option{reorder-block-duplicate-feedback} is used only when profile
10385 feedback is available. It may be set to higher values than
10386 @option{reorder-block-duplicate} since information about the hot spots is more
10389 @item max-sched-ready-insns
10390 The maximum number of instructions ready to be issued the scheduler should
10391 consider at any given time during the first scheduling pass. Increasing
10392 values mean more thorough searches, making the compilation time increase
10393 with probably little benefit. The default value is 100.
10395 @item max-sched-region-blocks
10396 The maximum number of blocks in a region to be considered for
10397 interblock scheduling. The default value is 10.
10399 @item max-pipeline-region-blocks
10400 The maximum number of blocks in a region to be considered for
10401 pipelining in the selective scheduler. The default value is 15.
10403 @item max-sched-region-insns
10404 The maximum number of insns in a region to be considered for
10405 interblock scheduling. The default value is 100.
10407 @item max-pipeline-region-insns
10408 The maximum number of insns in a region to be considered for
10409 pipelining in the selective scheduler. The default value is 200.
10411 @item min-spec-prob
10412 The minimum probability (in percents) of reaching a source block
10413 for interblock speculative scheduling. The default value is 40.
10415 @item max-sched-extend-regions-iters
10416 The maximum number of iterations through CFG to extend regions.
10417 A value of 0 (the default) disables region extensions.
10419 @item max-sched-insn-conflict-delay
10420 The maximum conflict delay for an insn to be considered for speculative motion.
10421 The default value is 3.
10423 @item sched-spec-prob-cutoff
10424 The minimal probability of speculation success (in percents), so that
10425 speculative insns are scheduled.
10426 The default value is 40.
10428 @item sched-spec-state-edge-prob-cutoff
10429 The minimum probability an edge must have for the scheduler to save its
10431 The default value is 10.
10433 @item sched-mem-true-dep-cost
10434 Minimal distance (in CPU cycles) between store and load targeting same
10435 memory locations. The default value is 1.
10437 @item selsched-max-lookahead
10438 The maximum size of the lookahead window of selective scheduling. It is a
10439 depth of search for available instructions.
10440 The default value is 50.
10442 @item selsched-max-sched-times
10443 The maximum number of times that an instruction is scheduled during
10444 selective scheduling. This is the limit on the number of iterations
10445 through which the instruction may be pipelined. The default value is 2.
10447 @item selsched-max-insns-to-rename
10448 The maximum number of best instructions in the ready list that are considered
10449 for renaming in the selective scheduler. The default value is 2.
10452 The minimum value of stage count that swing modulo scheduler
10453 generates. The default value is 2.
10455 @item max-last-value-rtl
10456 The maximum size measured as number of RTLs that can be recorded in an expression
10457 in combiner for a pseudo register as last known value of that register. The default
10460 @item max-combine-insns
10461 The maximum number of instructions the RTL combiner tries to combine.
10462 The default value is 2 at @option{-Og} and 4 otherwise.
10464 @item integer-share-limit
10465 Small integer constants can use a shared data structure, reducing the
10466 compiler's memory usage and increasing its speed. This sets the maximum
10467 value of a shared integer constant. The default value is 256.
10469 @item ssp-buffer-size
10470 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10471 protection when @option{-fstack-protection} is used.
10473 @item min-size-for-stack-sharing
10474 The minimum size of variables taking part in stack slot sharing when not
10475 optimizing. The default value is 32.
10477 @item max-jump-thread-duplication-stmts
10478 Maximum number of statements allowed in a block that needs to be
10479 duplicated when threading jumps.
10481 @item max-fields-for-field-sensitive
10482 Maximum number of fields in a structure treated in
10483 a field sensitive manner during pointer analysis. The default is zero
10484 for @option{-O0} and @option{-O1},
10485 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10487 @item prefetch-latency
10488 Estimate on average number of instructions that are executed before
10489 prefetch finishes. The distance prefetched ahead is proportional
10490 to this constant. Increasing this number may also lead to less
10491 streams being prefetched (see @option{simultaneous-prefetches}).
10493 @item simultaneous-prefetches
10494 Maximum number of prefetches that can run at the same time.
10496 @item l1-cache-line-size
10497 The size of cache line in L1 cache, in bytes.
10499 @item l1-cache-size
10500 The size of L1 cache, in kilobytes.
10502 @item l2-cache-size
10503 The size of L2 cache, in kilobytes.
10505 @item min-insn-to-prefetch-ratio
10506 The minimum ratio between the number of instructions and the
10507 number of prefetches to enable prefetching in a loop.
10509 @item prefetch-min-insn-to-mem-ratio
10510 The minimum ratio between the number of instructions and the
10511 number of memory references to enable prefetching in a loop.
10513 @item use-canonical-types
10514 Whether the compiler should use the ``canonical'' type system. By
10515 default, this should always be 1, which uses a more efficient internal
10516 mechanism for comparing types in C++ and Objective-C++. However, if
10517 bugs in the canonical type system are causing compilation failures,
10518 set this value to 0 to disable canonical types.
10520 @item switch-conversion-max-branch-ratio
10521 Switch initialization conversion refuses to create arrays that are
10522 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10523 branches in the switch.
10525 @item max-partial-antic-length
10526 Maximum length of the partial antic set computed during the tree
10527 partial redundancy elimination optimization (@option{-ftree-pre}) when
10528 optimizing at @option{-O3} and above. For some sorts of source code
10529 the enhanced partial redundancy elimination optimization can run away,
10530 consuming all of the memory available on the host machine. This
10531 parameter sets a limit on the length of the sets that are computed,
10532 which prevents the runaway behavior. Setting a value of 0 for
10533 this parameter allows an unlimited set length.
10535 @item sccvn-max-scc-size
10536 Maximum size of a strongly connected component (SCC) during SCCVN
10537 processing. If this limit is hit, SCCVN processing for the whole
10538 function is not done and optimizations depending on it are
10539 disabled. The default maximum SCC size is 10000.
10541 @item sccvn-max-alias-queries-per-access
10542 Maximum number of alias-oracle queries we perform when looking for
10543 redundancies for loads and stores. If this limit is hit the search
10544 is aborted and the load or store is not considered redundant. The
10545 number of queries is algorithmically limited to the number of
10546 stores on all paths from the load to the function entry.
10547 The default maxmimum number of queries is 1000.
10549 @item ira-max-loops-num
10550 IRA uses regional register allocation by default. If a function
10551 contains more loops than the number given by this parameter, only at most
10552 the given number of the most frequently-executed loops form regions
10553 for regional register allocation. The default value of the
10556 @item ira-max-conflict-table-size
10557 Although IRA uses a sophisticated algorithm to compress the conflict
10558 table, the table can still require excessive amounts of memory for
10559 huge functions. If the conflict table for a function could be more
10560 than the size in MB given by this parameter, the register allocator
10561 instead uses a faster, simpler, and lower-quality
10562 algorithm that does not require building a pseudo-register conflict table.
10563 The default value of the parameter is 2000.
10565 @item ira-loop-reserved-regs
10566 IRA can be used to evaluate more accurate register pressure in loops
10567 for decisions to move loop invariants (see @option{-O3}). The number
10568 of available registers reserved for some other purposes is given
10569 by this parameter. The default value of the parameter is 2, which is
10570 the minimal number of registers needed by typical instructions.
10571 This value is the best found from numerous experiments.
10573 @item lra-inheritance-ebb-probability-cutoff
10574 LRA tries to reuse values reloaded in registers in subsequent insns.
10575 This optimization is called inheritance. EBB is used as a region to
10576 do this optimization. The parameter defines a minimal fall-through
10577 edge probability in percentage used to add BB to inheritance EBB in
10578 LRA. The default value of the parameter is 40. The value was chosen
10579 from numerous runs of SPEC2000 on x86-64.
10581 @item loop-invariant-max-bbs-in-loop
10582 Loop invariant motion can be very expensive, both in compilation time and
10583 in amount of needed compile-time memory, with very large loops. Loops
10584 with more basic blocks than this parameter won't have loop invariant
10585 motion optimization performed on them. The default value of the
10586 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10588 @item loop-max-datarefs-for-datadeps
10589 Building data dapendencies is expensive for very large loops. This
10590 parameter limits the number of data references in loops that are
10591 considered for data dependence analysis. These large loops are no
10592 handled by the optimizations using loop data dependencies.
10593 The default value is 1000.
10595 @item max-vartrack-size
10596 Sets a maximum number of hash table slots to use during variable
10597 tracking dataflow analysis of any function. If this limit is exceeded
10598 with variable tracking at assignments enabled, analysis for that
10599 function is retried without it, after removing all debug insns from
10600 the function. If the limit is exceeded even without debug insns, var
10601 tracking analysis is completely disabled for the function. Setting
10602 the parameter to zero makes it unlimited.
10604 @item max-vartrack-expr-depth
10605 Sets a maximum number of recursion levels when attempting to map
10606 variable names or debug temporaries to value expressions. This trades
10607 compilation time for more complete debug information. If this is set too
10608 low, value expressions that are available and could be represented in
10609 debug information may end up not being used; setting this higher may
10610 enable the compiler to find more complex debug expressions, but compile
10611 time and memory use may grow. The default is 12.
10613 @item min-nondebug-insn-uid
10614 Use uids starting at this parameter for nondebug insns. The range below
10615 the parameter is reserved exclusively for debug insns created by
10616 @option{-fvar-tracking-assignments}, but debug insns may get
10617 (non-overlapping) uids above it if the reserved range is exhausted.
10619 @item ipa-sra-ptr-growth-factor
10620 IPA-SRA replaces a pointer to an aggregate with one or more new
10621 parameters only when their cumulative size is less or equal to
10622 @option{ipa-sra-ptr-growth-factor} times the size of the original
10625 @item sra-max-scalarization-size-Ospeed
10626 @item sra-max-scalarization-size-Osize
10627 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10628 replace scalar parts of aggregates with uses of independent scalar
10629 variables. These parameters control the maximum size, in storage units,
10630 of aggregate which is considered for replacement when compiling for
10632 (@option{sra-max-scalarization-size-Ospeed}) or size
10633 (@option{sra-max-scalarization-size-Osize}) respectively.
10635 @item tm-max-aggregate-size
10636 When making copies of thread-local variables in a transaction, this
10637 parameter specifies the size in bytes after which variables are
10638 saved with the logging functions as opposed to save/restore code
10639 sequence pairs. This option only applies when using
10642 @item graphite-max-nb-scop-params
10643 To avoid exponential effects in the Graphite loop transforms, the
10644 number of parameters in a Static Control Part (SCoP) is bounded. The
10645 default value is 10 parameters. A variable whose value is unknown at
10646 compilation time and defined outside a SCoP is a parameter of the SCoP.
10648 @item graphite-max-bbs-per-function
10649 To avoid exponential effects in the detection of SCoPs, the size of
10650 the functions analyzed by Graphite is bounded. The default value is
10653 @item loop-block-tile-size
10654 Loop blocking or strip mining transforms, enabled with
10655 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10656 loop in the loop nest by a given number of iterations. The strip
10657 length can be changed using the @option{loop-block-tile-size}
10658 parameter. The default value is 51 iterations.
10660 @item loop-unroll-jam-size
10661 Specify the unroll factor for the @option{-floop-unroll-and-jam}. The
10662 default value is 4.
10664 @item loop-unroll-jam-depth
10665 Specify the dimension to be unrolled (counting from the most inner loop)
10666 for the @option{-floop-unroll-and-jam}. The default value is 2.
10668 @item ipa-cp-value-list-size
10669 IPA-CP attempts to track all possible values and types passed to a function's
10670 parameter in order to propagate them and perform devirtualization.
10671 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10672 stores per one formal parameter of a function.
10674 @item ipa-cp-eval-threshold
10675 IPA-CP calculates its own score of cloning profitability heuristics
10676 and performs those cloning opportunities with scores that exceed
10677 @option{ipa-cp-eval-threshold}.
10679 @item ipa-max-agg-items
10680 IPA-CP is also capable to propagate a number of scalar values passed
10681 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10682 number of such values per one parameter.
10684 @item ipa-cp-loop-hint-bonus
10685 When IPA-CP determines that a cloning candidate would make the number
10686 of iterations of a loop known, it adds a bonus of
10687 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10690 @item ipa-cp-array-index-hint-bonus
10691 When IPA-CP determines that a cloning candidate would make the index of
10692 an array access known, it adds a bonus of
10693 @option{ipa-cp-array-index-hint-bonus} to the profitability
10694 score of the candidate.
10696 @item ipa-max-aa-steps
10697 During its analysis of function bodies, IPA-CP employs alias analysis
10698 in order to track values pointed to by function parameters. In order
10699 not spend too much time analyzing huge functions, it gives up and
10700 consider all memory clobbered after examining
10701 @option{ipa-max-aa-steps} statements modifying memory.
10703 @item lto-partitions
10704 Specify desired number of partitions produced during WHOPR compilation.
10705 The number of partitions should exceed the number of CPUs used for compilation.
10706 The default value is 32.
10708 @item lto-minpartition
10709 Size of minimal partition for WHOPR (in estimated instructions).
10710 This prevents expenses of splitting very small programs into too many
10713 @item cxx-max-namespaces-for-diagnostic-help
10714 The maximum number of namespaces to consult for suggestions when C++
10715 name lookup fails for an identifier. The default is 1000.
10717 @item sink-frequency-threshold
10718 The maximum relative execution frequency (in percents) of the target block
10719 relative to a statement's original block to allow statement sinking of a
10720 statement. Larger numbers result in more aggressive statement sinking.
10721 The default value is 75. A small positive adjustment is applied for
10722 statements with memory operands as those are even more profitable so sink.
10724 @item max-stores-to-sink
10725 The maximum number of conditional stores paires that can be sunk. Set to 0
10726 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10727 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10729 @item allow-store-data-races
10730 Allow optimizers to introduce new data races on stores.
10731 Set to 1 to allow, otherwise to 0. This option is enabled by default
10732 at optimization level @option{-Ofast}.
10734 @item case-values-threshold
10735 The smallest number of different values for which it is best to use a
10736 jump-table instead of a tree of conditional branches. If the value is
10737 0, use the default for the machine. The default is 0.
10739 @item tree-reassoc-width
10740 Set the maximum number of instructions executed in parallel in
10741 reassociated tree. This parameter overrides target dependent
10742 heuristics used by default if has non zero value.
10744 @item sched-pressure-algorithm
10745 Choose between the two available implementations of
10746 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10747 and is the more likely to prevent instructions from being reordered.
10748 Algorithm 2 was designed to be a compromise between the relatively
10749 conservative approach taken by algorithm 1 and the rather aggressive
10750 approach taken by the default scheduler. It relies more heavily on
10751 having a regular register file and accurate register pressure classes.
10752 See @file{haifa-sched.c} in the GCC sources for more details.
10754 The default choice depends on the target.
10756 @item max-slsr-cand-scan
10757 Set the maximum number of existing candidates that are considered when
10758 seeking a basis for a new straight-line strength reduction candidate.
10761 Enable buffer overflow detection for global objects. This kind
10762 of protection is enabled by default if you are using
10763 @option{-fsanitize=address} option.
10764 To disable global objects protection use @option{--param asan-globals=0}.
10767 Enable buffer overflow detection for stack objects. This kind of
10768 protection is enabled by default when using@option{-fsanitize=address}.
10769 To disable stack protection use @option{--param asan-stack=0} option.
10771 @item asan-instrument-reads
10772 Enable buffer overflow detection for memory reads. This kind of
10773 protection is enabled by default when using @option{-fsanitize=address}.
10774 To disable memory reads protection use
10775 @option{--param asan-instrument-reads=0}.
10777 @item asan-instrument-writes
10778 Enable buffer overflow detection for memory writes. This kind of
10779 protection is enabled by default when using @option{-fsanitize=address}.
10780 To disable memory writes protection use
10781 @option{--param asan-instrument-writes=0} option.
10783 @item asan-memintrin
10784 Enable detection for built-in functions. This kind of protection
10785 is enabled by default when using @option{-fsanitize=address}.
10786 To disable built-in functions protection use
10787 @option{--param asan-memintrin=0}.
10789 @item asan-use-after-return
10790 Enable detection of use-after-return. This kind of protection
10791 is enabled by default when using @option{-fsanitize=address} option.
10792 To disable use-after-return detection use
10793 @option{--param asan-use-after-return=0}.
10795 @item asan-instrumentation-with-call-threshold
10796 If number of memory accesses in function being instrumented
10797 is greater or equal to this number, use callbacks instead of inline checks.
10798 E.g. to disable inline code use
10799 @option{--param asan-instrumentation-with-call-threshold=0}.
10801 @item chkp-max-ctor-size
10802 Static constructors generated by Pointer Bounds Checker may become very
10803 large and significantly increase compile time at optimization level
10804 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10805 in a single generated constructor. Default value is 5000.
10807 @item max-fsm-thread-path-insns
10808 Maximum number of instructions to copy when duplicating blocks on a
10809 finite state automaton jump thread path. The default is 100.
10811 @item max-fsm-thread-length
10812 Maximum number of basic blocks on a finite state automaton jump thread
10813 path. The default is 10.
10815 @item max-fsm-thread-paths
10816 Maximum number of new jump thread paths to create for a finite state
10817 automaton. The default is 50.
10822 @node Preprocessor Options
10823 @section Options Controlling the Preprocessor
10824 @cindex preprocessor options
10825 @cindex options, preprocessor
10827 These options control the C preprocessor, which is run on each C source
10828 file before actual compilation.
10830 If you use the @option{-E} option, nothing is done except preprocessing.
10831 Some of these options make sense only together with @option{-E} because
10832 they cause the preprocessor output to be unsuitable for actual
10836 @item -Wp,@var{option}
10838 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10839 and pass @var{option} directly through to the preprocessor. If
10840 @var{option} contains commas, it is split into multiple options at the
10841 commas. However, many options are modified, translated or interpreted
10842 by the compiler driver before being passed to the preprocessor, and
10843 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10844 interface is undocumented and subject to change, so whenever possible
10845 you should avoid using @option{-Wp} and let the driver handle the
10848 @item -Xpreprocessor @var{option}
10849 @opindex Xpreprocessor
10850 Pass @var{option} as an option to the preprocessor. You can use this to
10851 supply system-specific preprocessor options that GCC does not
10854 If you want to pass an option that takes an argument, you must use
10855 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10857 @item -no-integrated-cpp
10858 @opindex no-integrated-cpp
10859 Perform preprocessing as a separate pass before compilation.
10860 By default, GCC performs preprocessing as an integrated part of
10861 input tokenization and parsing.
10862 If this option is provided, the appropriate language front end
10863 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10864 and Objective-C, respectively) is instead invoked twice,
10865 once for preprocessing only and once for actual compilation
10866 of the preprocessed input.
10867 This option may be useful in conjunction with the @option{-B} or
10868 @option{-wrapper} options to specify an alternate preprocessor or
10869 perform additional processing of the program source between
10870 normal preprocessing and compilation.
10873 @include cppopts.texi
10875 @node Assembler Options
10876 @section Passing Options to the Assembler
10878 @c prevent bad page break with this line
10879 You can pass options to the assembler.
10882 @item -Wa,@var{option}
10884 Pass @var{option} as an option to the assembler. If @var{option}
10885 contains commas, it is split into multiple options at the commas.
10887 @item -Xassembler @var{option}
10888 @opindex Xassembler
10889 Pass @var{option} as an option to the assembler. You can use this to
10890 supply system-specific assembler options that GCC does not
10893 If you want to pass an option that takes an argument, you must use
10894 @option{-Xassembler} twice, once for the option and once for the argument.
10899 @section Options for Linking
10900 @cindex link options
10901 @cindex options, linking
10903 These options come into play when the compiler links object files into
10904 an executable output file. They are meaningless if the compiler is
10905 not doing a link step.
10909 @item @var{object-file-name}
10910 A file name that does not end in a special recognized suffix is
10911 considered to name an object file or library. (Object files are
10912 distinguished from libraries by the linker according to the file
10913 contents.) If linking is done, these object files are used as input
10922 If any of these options is used, then the linker is not run, and
10923 object file names should not be used as arguments. @xref{Overall
10927 @opindex fuse-ld=bfd
10928 Use the @command{bfd} linker instead of the default linker.
10930 @item -fuse-ld=gold
10931 @opindex fuse-ld=gold
10932 Use the @command{gold} linker instead of the default linker.
10935 @item -l@var{library}
10936 @itemx -l @var{library}
10938 Search the library named @var{library} when linking. (The second
10939 alternative with the library as a separate argument is only for
10940 POSIX compliance and is not recommended.)
10942 It makes a difference where in the command you write this option; the
10943 linker searches and processes libraries and object files in the order they
10944 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10945 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10946 to functions in @samp{z}, those functions may not be loaded.
10948 The linker searches a standard list of directories for the library,
10949 which is actually a file named @file{lib@var{library}.a}. The linker
10950 then uses this file as if it had been specified precisely by name.
10952 The directories searched include several standard system directories
10953 plus any that you specify with @option{-L}.
10955 Normally the files found this way are library files---archive files
10956 whose members are object files. The linker handles an archive file by
10957 scanning through it for members which define symbols that have so far
10958 been referenced but not defined. But if the file that is found is an
10959 ordinary object file, it is linked in the usual fashion. The only
10960 difference between using an @option{-l} option and specifying a file name
10961 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10962 and searches several directories.
10966 You need this special case of the @option{-l} option in order to
10967 link an Objective-C or Objective-C++ program.
10969 @item -nostartfiles
10970 @opindex nostartfiles
10971 Do not use the standard system startup files when linking.
10972 The standard system libraries are used normally, unless @option{-nostdlib}
10973 or @option{-nodefaultlibs} is used.
10975 @item -nodefaultlibs
10976 @opindex nodefaultlibs
10977 Do not use the standard system libraries when linking.
10978 Only the libraries you specify are passed to the linker, and options
10979 specifying linkage of the system libraries, such as @option{-static-libgcc}
10980 or @option{-shared-libgcc}, are ignored.
10981 The standard startup files are used normally, unless @option{-nostartfiles}
10984 The compiler may generate calls to @code{memcmp},
10985 @code{memset}, @code{memcpy} and @code{memmove}.
10986 These entries are usually resolved by entries in
10987 libc. These entry points should be supplied through some other
10988 mechanism when this option is specified.
10992 Do not use the standard system startup files or libraries when linking.
10993 No startup files and only the libraries you specify are passed to
10994 the linker, and options specifying linkage of the system libraries, such as
10995 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
10997 The compiler may generate calls to @code{memcmp}, @code{memset},
10998 @code{memcpy} and @code{memmove}.
10999 These entries are usually resolved by entries in
11000 libc. These entry points should be supplied through some other
11001 mechanism when this option is specified.
11003 @cindex @option{-lgcc}, use with @option{-nostdlib}
11004 @cindex @option{-nostdlib} and unresolved references
11005 @cindex unresolved references and @option{-nostdlib}
11006 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11007 @cindex @option{-nodefaultlibs} and unresolved references
11008 @cindex unresolved references and @option{-nodefaultlibs}
11009 One of the standard libraries bypassed by @option{-nostdlib} and
11010 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11011 which GCC uses to overcome shortcomings of particular machines, or special
11012 needs for some languages.
11013 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11014 Collection (GCC) Internals},
11015 for more discussion of @file{libgcc.a}.)
11016 In most cases, you need @file{libgcc.a} even when you want to avoid
11017 other standard libraries. In other words, when you specify @option{-nostdlib}
11018 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11019 This ensures that you have no unresolved references to internal GCC
11020 library subroutines.
11021 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11022 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11023 GNU Compiler Collection (GCC) Internals}.)
11027 Produce a position independent executable on targets that support it.
11028 For predictable results, you must also specify the same set of options
11029 used for compilation (@option{-fpie}, @option{-fPIE},
11030 or model suboptions) when you specify this linker option.
11034 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11035 that support it. This instructs the linker to add all symbols, not
11036 only used ones, to the dynamic symbol table. This option is needed
11037 for some uses of @code{dlopen} or to allow obtaining backtraces
11038 from within a program.
11042 Remove all symbol table and relocation information from the executable.
11046 On systems that support dynamic linking, this prevents linking with the shared
11047 libraries. On other systems, this option has no effect.
11051 Produce a shared object which can then be linked with other objects to
11052 form an executable. Not all systems support this option. For predictable
11053 results, you must also specify the same set of options used for compilation
11054 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11055 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11056 needs to build supplementary stub code for constructors to work. On
11057 multi-libbed systems, @samp{gcc -shared} must select the correct support
11058 libraries to link against. Failing to supply the correct flags may lead
11059 to subtle defects. Supplying them in cases where they are not necessary
11062 @item -shared-libgcc
11063 @itemx -static-libgcc
11064 @opindex shared-libgcc
11065 @opindex static-libgcc
11066 On systems that provide @file{libgcc} as a shared library, these options
11067 force the use of either the shared or static version, respectively.
11068 If no shared version of @file{libgcc} was built when the compiler was
11069 configured, these options have no effect.
11071 There are several situations in which an application should use the
11072 shared @file{libgcc} instead of the static version. The most common
11073 of these is when the application wishes to throw and catch exceptions
11074 across different shared libraries. In that case, each of the libraries
11075 as well as the application itself should use the shared @file{libgcc}.
11077 Therefore, the G++ and GCJ drivers automatically add
11078 @option{-shared-libgcc} whenever you build a shared library or a main
11079 executable, because C++ and Java programs typically use exceptions, so
11080 this is the right thing to do.
11082 If, instead, you use the GCC driver to create shared libraries, you may
11083 find that they are not always linked with the shared @file{libgcc}.
11084 If GCC finds, at its configuration time, that you have a non-GNU linker
11085 or a GNU linker that does not support option @option{--eh-frame-hdr},
11086 it links the shared version of @file{libgcc} into shared libraries
11087 by default. Otherwise, it takes advantage of the linker and optimizes
11088 away the linking with the shared version of @file{libgcc}, linking with
11089 the static version of libgcc by default. This allows exceptions to
11090 propagate through such shared libraries, without incurring relocation
11091 costs at library load time.
11093 However, if a library or main executable is supposed to throw or catch
11094 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11095 for the languages used in the program, or using the option
11096 @option{-shared-libgcc}, such that it is linked with the shared
11099 @item -static-libasan
11100 @opindex static-libasan
11101 When the @option{-fsanitize=address} option is used to link a program,
11102 the GCC driver automatically links against @option{libasan}. If
11103 @file{libasan} is available as a shared library, and the @option{-static}
11104 option is not used, then this links against the shared version of
11105 @file{libasan}. The @option{-static-libasan} option directs the GCC
11106 driver to link @file{libasan} statically, without necessarily linking
11107 other libraries statically.
11109 @item -static-libtsan
11110 @opindex static-libtsan
11111 When the @option{-fsanitize=thread} option is used to link a program,
11112 the GCC driver automatically links against @option{libtsan}. If
11113 @file{libtsan} is available as a shared library, and the @option{-static}
11114 option is not used, then this links against the shared version of
11115 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11116 driver to link @file{libtsan} statically, without necessarily linking
11117 other libraries statically.
11119 @item -static-liblsan
11120 @opindex static-liblsan
11121 When the @option{-fsanitize=leak} option is used to link a program,
11122 the GCC driver automatically links against @option{liblsan}. If
11123 @file{liblsan} is available as a shared library, and the @option{-static}
11124 option is not used, then this links against the shared version of
11125 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11126 driver to link @file{liblsan} statically, without necessarily linking
11127 other libraries statically.
11129 @item -static-libubsan
11130 @opindex static-libubsan
11131 When the @option{-fsanitize=undefined} option is used to link a program,
11132 the GCC driver automatically links against @option{libubsan}. If
11133 @file{libubsan} is available as a shared library, and the @option{-static}
11134 option is not used, then this links against the shared version of
11135 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11136 driver to link @file{libubsan} statically, without necessarily linking
11137 other libraries statically.
11139 @item -static-libstdc++
11140 @opindex static-libstdc++
11141 When the @command{g++} program is used to link a C++ program, it
11142 normally automatically links against @option{libstdc++}. If
11143 @file{libstdc++} is available as a shared library, and the
11144 @option{-static} option is not used, then this links against the
11145 shared version of @file{libstdc++}. That is normally fine. However, it
11146 is sometimes useful to freeze the version of @file{libstdc++} used by
11147 the program without going all the way to a fully static link. The
11148 @option{-static-libstdc++} option directs the @command{g++} driver to
11149 link @file{libstdc++} statically, without necessarily linking other
11150 libraries statically.
11154 Bind references to global symbols when building a shared object. Warn
11155 about any unresolved references (unless overridden by the link editor
11156 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11159 @item -T @var{script}
11161 @cindex linker script
11162 Use @var{script} as the linker script. This option is supported by most
11163 systems using the GNU linker. On some targets, such as bare-board
11164 targets without an operating system, the @option{-T} option may be required
11165 when linking to avoid references to undefined symbols.
11167 @item -Xlinker @var{option}
11169 Pass @var{option} as an option to the linker. You can use this to
11170 supply system-specific linker options that GCC does not recognize.
11172 If you want to pass an option that takes a separate argument, you must use
11173 @option{-Xlinker} twice, once for the option and once for the argument.
11174 For example, to pass @option{-assert definitions}, you must write
11175 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11176 @option{-Xlinker "-assert definitions"}, because this passes the entire
11177 string as a single argument, which is not what the linker expects.
11179 When using the GNU linker, it is usually more convenient to pass
11180 arguments to linker options using the @option{@var{option}=@var{value}}
11181 syntax than as separate arguments. For example, you can specify
11182 @option{-Xlinker -Map=output.map} rather than
11183 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11184 this syntax for command-line options.
11186 @item -Wl,@var{option}
11188 Pass @var{option} as an option to the linker. If @var{option} contains
11189 commas, it is split into multiple options at the commas. You can use this
11190 syntax to pass an argument to the option.
11191 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11192 linker. When using the GNU linker, you can also get the same effect with
11193 @option{-Wl,-Map=output.map}.
11195 @item -u @var{symbol}
11197 Pretend the symbol @var{symbol} is undefined, to force linking of
11198 library modules to define it. You can use @option{-u} multiple times with
11199 different symbols to force loading of additional library modules.
11201 @item -z @var{keyword}
11203 @option{-z} is passed directly on to the linker along with the keyword
11204 @var{keyword}. See the section in the documentation of your linker for
11205 permitted values and their meanings.
11208 @node Directory Options
11209 @section Options for Directory Search
11210 @cindex directory options
11211 @cindex options, directory search
11212 @cindex search path
11214 These options specify directories to search for header files, for
11215 libraries and for parts of the compiler:
11220 Add the directory @var{dir} to the head of the list of directories to be
11221 searched for header files. This can be used to override a system header
11222 file, substituting your own version, since these directories are
11223 searched before the system header file directories. However, you should
11224 not use this option to add directories that contain vendor-supplied
11225 system header files (use @option{-isystem} for that). If you use more than
11226 one @option{-I} option, the directories are scanned in left-to-right
11227 order; the standard system directories come after.
11229 If a standard system include directory, or a directory specified with
11230 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11231 option is ignored. The directory is still searched but as a
11232 system directory at its normal position in the system include chain.
11233 This is to ensure that GCC's procedure to fix buggy system headers and
11234 the ordering for the @code{include_next} directive are not inadvertently changed.
11235 If you really need to change the search order for system directories,
11236 use the @option{-nostdinc} and/or @option{-isystem} options.
11238 @item -iplugindir=@var{dir}
11239 @opindex iplugindir=
11240 Set the directory to search for plugins that are passed
11241 by @option{-fplugin=@var{name}} instead of
11242 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11243 to be used by the user, but only passed by the driver.
11245 @item -iquote@var{dir}
11247 Add the directory @var{dir} to the head of the list of directories to
11248 be searched for header files only for the case of @code{#include
11249 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11250 otherwise just like @option{-I}.
11254 Add directory @var{dir} to the list of directories to be searched
11257 @item -B@var{prefix}
11259 This option specifies where to find the executables, libraries,
11260 include files, and data files of the compiler itself.
11262 The compiler driver program runs one or more of the subprograms
11263 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11264 @var{prefix} as a prefix for each program it tries to run, both with and
11265 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11267 For each subprogram to be run, the compiler driver first tries the
11268 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11269 is not specified, the driver tries two standard prefixes,
11270 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11271 those results in a file name that is found, the unmodified program
11272 name is searched for using the directories specified in your
11273 @env{PATH} environment variable.
11275 The compiler checks to see if the path provided by the @option{-B}
11276 refers to a directory, and if necessary it adds a directory
11277 separator character at the end of the path.
11279 @option{-B} prefixes that effectively specify directory names also apply
11280 to libraries in the linker, because the compiler translates these
11281 options into @option{-L} options for the linker. They also apply to
11282 include files in the preprocessor, because the compiler translates these
11283 options into @option{-isystem} options for the preprocessor. In this case,
11284 the compiler appends @samp{include} to the prefix.
11286 The runtime support file @file{libgcc.a} can also be searched for using
11287 the @option{-B} prefix, if needed. If it is not found there, the two
11288 standard prefixes above are tried, and that is all. The file is left
11289 out of the link if it is not found by those means.
11291 Another way to specify a prefix much like the @option{-B} prefix is to use
11292 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11295 As a special kludge, if the path provided by @option{-B} is
11296 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11297 9, then it is replaced by @file{[dir/]include}. This is to help
11298 with boot-strapping the compiler.
11300 @item -specs=@var{file}
11302 Process @var{file} after the compiler reads in the standard @file{specs}
11303 file, in order to override the defaults which the @command{gcc} driver
11304 program uses when determining what switches to pass to @command{cc1},
11305 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11306 @option{-specs=@var{file}} can be specified on the command line, and they
11307 are processed in order, from left to right.
11309 @item --sysroot=@var{dir}
11311 Use @var{dir} as the logical root directory for headers and libraries.
11312 For example, if the compiler normally searches for headers in
11313 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11314 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11316 If you use both this option and the @option{-isysroot} option, then
11317 the @option{--sysroot} option applies to libraries, but the
11318 @option{-isysroot} option applies to header files.
11320 The GNU linker (beginning with version 2.16) has the necessary support
11321 for this option. If your linker does not support this option, the
11322 header file aspect of @option{--sysroot} still works, but the
11323 library aspect does not.
11325 @item --no-sysroot-suffix
11326 @opindex no-sysroot-suffix
11327 For some targets, a suffix is added to the root directory specified
11328 with @option{--sysroot}, depending on the other options used, so that
11329 headers may for example be found in
11330 @file{@var{dir}/@var{suffix}/usr/include} instead of
11331 @file{@var{dir}/usr/include}. This option disables the addition of
11336 This option has been deprecated. Please use @option{-iquote} instead for
11337 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
11338 Any directories you specify with @option{-I} options before the @option{-I-}
11339 option are searched only for the case of @code{#include "@var{file}"};
11340 they are not searched for @code{#include <@var{file}>}.
11342 If additional directories are specified with @option{-I} options after
11343 the @option{-I-}, these directories are searched for all @code{#include}
11344 directives. (Ordinarily @emph{all} @option{-I} directories are used
11347 In addition, the @option{-I-} option inhibits the use of the current
11348 directory (where the current input file came from) as the first search
11349 directory for @code{#include "@var{file}"}. There is no way to
11350 override this effect of @option{-I-}. With @option{-I.} you can specify
11351 searching the directory that is current when the compiler is
11352 invoked. That is not exactly the same as what the preprocessor does
11353 by default, but it is often satisfactory.
11355 @option{-I-} does not inhibit the use of the standard system directories
11356 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11363 @section Specifying Subprocesses and the Switches to Pass to Them
11366 @command{gcc} is a driver program. It performs its job by invoking a
11367 sequence of other programs to do the work of compiling, assembling and
11368 linking. GCC interprets its command-line parameters and uses these to
11369 deduce which programs it should invoke, and which command-line options
11370 it ought to place on their command lines. This behavior is controlled
11371 by @dfn{spec strings}. In most cases there is one spec string for each
11372 program that GCC can invoke, but a few programs have multiple spec
11373 strings to control their behavior. The spec strings built into GCC can
11374 be overridden by using the @option{-specs=} command-line switch to specify
11377 @dfn{Spec files} are plaintext files that are used to construct spec
11378 strings. They consist of a sequence of directives separated by blank
11379 lines. The type of directive is determined by the first non-whitespace
11380 character on the line, which can be one of the following:
11383 @item %@var{command}
11384 Issues a @var{command} to the spec file processor. The commands that can
11388 @item %include <@var{file}>
11389 @cindex @code{%include}
11390 Search for @var{file} and insert its text at the current point in the
11393 @item %include_noerr <@var{file}>
11394 @cindex @code{%include_noerr}
11395 Just like @samp{%include}, but do not generate an error message if the include
11396 file cannot be found.
11398 @item %rename @var{old_name} @var{new_name}
11399 @cindex @code{%rename}
11400 Rename the spec string @var{old_name} to @var{new_name}.
11404 @item *[@var{spec_name}]:
11405 This tells the compiler to create, override or delete the named spec
11406 string. All lines after this directive up to the next directive or
11407 blank line are considered to be the text for the spec string. If this
11408 results in an empty string then the spec is deleted. (Or, if the
11409 spec did not exist, then nothing happens.) Otherwise, if the spec
11410 does not currently exist a new spec is created. If the spec does
11411 exist then its contents are overridden by the text of this
11412 directive, unless the first character of that text is the @samp{+}
11413 character, in which case the text is appended to the spec.
11415 @item [@var{suffix}]:
11416 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11417 and up to the next directive or blank line are considered to make up the
11418 spec string for the indicated suffix. When the compiler encounters an
11419 input file with the named suffix, it processes the spec string in
11420 order to work out how to compile that file. For example:
11424 z-compile -input %i
11427 This says that any input file whose name ends in @samp{.ZZ} should be
11428 passed to the program @samp{z-compile}, which should be invoked with the
11429 command-line switch @option{-input} and with the result of performing the
11430 @samp{%i} substitution. (See below.)
11432 As an alternative to providing a spec string, the text following a
11433 suffix directive can be one of the following:
11436 @item @@@var{language}
11437 This says that the suffix is an alias for a known @var{language}. This is
11438 similar to using the @option{-x} command-line switch to GCC to specify a
11439 language explicitly. For example:
11446 Says that .ZZ files are, in fact, C++ source files.
11449 This causes an error messages saying:
11452 @var{name} compiler not installed on this system.
11456 GCC already has an extensive list of suffixes built into it.
11457 This directive adds an entry to the end of the list of suffixes, but
11458 since the list is searched from the end backwards, it is effectively
11459 possible to override earlier entries using this technique.
11463 GCC has the following spec strings built into it. Spec files can
11464 override these strings or create their own. Note that individual
11465 targets can also add their own spec strings to this list.
11468 asm Options to pass to the assembler
11469 asm_final Options to pass to the assembler post-processor
11470 cpp Options to pass to the C preprocessor
11471 cc1 Options to pass to the C compiler
11472 cc1plus Options to pass to the C++ compiler
11473 endfile Object files to include at the end of the link
11474 link Options to pass to the linker
11475 lib Libraries to include on the command line to the linker
11476 libgcc Decides which GCC support library to pass to the linker
11477 linker Sets the name of the linker
11478 predefines Defines to be passed to the C preprocessor
11479 signed_char Defines to pass to CPP to say whether @code{char} is signed
11481 startfile Object files to include at the start of the link
11484 Here is a small example of a spec file:
11487 %rename lib old_lib
11490 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11493 This example renames the spec called @samp{lib} to @samp{old_lib} and
11494 then overrides the previous definition of @samp{lib} with a new one.
11495 The new definition adds in some extra command-line options before
11496 including the text of the old definition.
11498 @dfn{Spec strings} are a list of command-line options to be passed to their
11499 corresponding program. In addition, the spec strings can contain
11500 @samp{%}-prefixed sequences to substitute variable text or to
11501 conditionally insert text into the command line. Using these constructs
11502 it is possible to generate quite complex command lines.
11504 Here is a table of all defined @samp{%}-sequences for spec
11505 strings. Note that spaces are not generated automatically around the
11506 results of expanding these sequences. Therefore you can concatenate them
11507 together or combine them with constant text in a single argument.
11511 Substitute one @samp{%} into the program name or argument.
11514 Substitute the name of the input file being processed.
11517 Substitute the basename of the input file being processed.
11518 This is the substring up to (and not including) the last period
11519 and not including the directory.
11522 This is the same as @samp{%b}, but include the file suffix (text after
11526 Marks the argument containing or following the @samp{%d} as a
11527 temporary file name, so that that file is deleted if GCC exits
11528 successfully. Unlike @samp{%g}, this contributes no text to the
11531 @item %g@var{suffix}
11532 Substitute a file name that has suffix @var{suffix} and is chosen
11533 once per compilation, and mark the argument in the same way as
11534 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11535 name is now chosen in a way that is hard to predict even when previously
11536 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11537 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11538 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11539 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11540 was simply substituted with a file name chosen once per compilation,
11541 without regard to any appended suffix (which was therefore treated
11542 just like ordinary text), making such attacks more likely to succeed.
11544 @item %u@var{suffix}
11545 Like @samp{%g}, but generates a new temporary file name
11546 each time it appears instead of once per compilation.
11548 @item %U@var{suffix}
11549 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11550 new one if there is no such last file name. In the absence of any
11551 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11552 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11553 involves the generation of two distinct file names, one
11554 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11555 simply substituted with a file name chosen for the previous @samp{%u},
11556 without regard to any appended suffix.
11558 @item %j@var{suffix}
11559 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11560 writable, and if @option{-save-temps} is not used;
11561 otherwise, substitute the name
11562 of a temporary file, just like @samp{%u}. This temporary file is not
11563 meant for communication between processes, but rather as a junk
11564 disposal mechanism.
11566 @item %|@var{suffix}
11567 @itemx %m@var{suffix}
11568 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11569 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11570 all. These are the two most common ways to instruct a program that it
11571 should read from standard input or write to standard output. If you
11572 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11573 construct: see for example @file{f/lang-specs.h}.
11575 @item %.@var{SUFFIX}
11576 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11577 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11578 terminated by the next space or %.
11581 Marks the argument containing or following the @samp{%w} as the
11582 designated output file of this compilation. This puts the argument
11583 into the sequence of arguments that @samp{%o} substitutes.
11586 Substitutes the names of all the output files, with spaces
11587 automatically placed around them. You should write spaces
11588 around the @samp{%o} as well or the results are undefined.
11589 @samp{%o} is for use in the specs for running the linker.
11590 Input files whose names have no recognized suffix are not compiled
11591 at all, but they are included among the output files, so they are
11595 Substitutes the suffix for object files. Note that this is
11596 handled specially when it immediately follows @samp{%g, %u, or %U},
11597 because of the need for those to form complete file names. The
11598 handling is such that @samp{%O} is treated exactly as if it had already
11599 been substituted, except that @samp{%g, %u, and %U} do not currently
11600 support additional @var{suffix} characters following @samp{%O} as they do
11601 following, for example, @samp{.o}.
11604 Substitutes the standard macro predefinitions for the
11605 current target machine. Use this when running @command{cpp}.
11608 Like @samp{%p}, but puts @samp{__} before and after the name of each
11609 predefined macro, except for macros that start with @samp{__} or with
11610 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11614 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11615 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11616 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11617 and @option{-imultilib} as necessary.
11620 Current argument is the name of a library or startup file of some sort.
11621 Search for that file in a standard list of directories and substitute
11622 the full name found. The current working directory is included in the
11623 list of directories scanned.
11626 Current argument is the name of a linker script. Search for that file
11627 in the current list of directories to scan for libraries. If the file
11628 is located insert a @option{--script} option into the command line
11629 followed by the full path name found. If the file is not found then
11630 generate an error message. Note: the current working directory is not
11634 Print @var{str} as an error message. @var{str} is terminated by a newline.
11635 Use this when inconsistent options are detected.
11637 @item %(@var{name})
11638 Substitute the contents of spec string @var{name} at this point.
11640 @item %x@{@var{option}@}
11641 Accumulate an option for @samp{%X}.
11644 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11648 Output the accumulated assembler options specified by @option{-Wa}.
11651 Output the accumulated preprocessor options specified by @option{-Wp}.
11654 Process the @code{asm} spec. This is used to compute the
11655 switches to be passed to the assembler.
11658 Process the @code{asm_final} spec. This is a spec string for
11659 passing switches to an assembler post-processor, if such a program is
11663 Process the @code{link} spec. This is the spec for computing the
11664 command line passed to the linker. Typically it makes use of the
11665 @samp{%L %G %S %D and %E} sequences.
11668 Dump out a @option{-L} option for each directory that GCC believes might
11669 contain startup files. If the target supports multilibs then the
11670 current multilib directory is prepended to each of these paths.
11673 Process the @code{lib} spec. This is a spec string for deciding which
11674 libraries are included on the command line to the linker.
11677 Process the @code{libgcc} spec. This is a spec string for deciding
11678 which GCC support library is included on the command line to the linker.
11681 Process the @code{startfile} spec. This is a spec for deciding which
11682 object files are the first ones passed to the linker. Typically
11683 this might be a file named @file{crt0.o}.
11686 Process the @code{endfile} spec. This is a spec string that specifies
11687 the last object files that are passed to the linker.
11690 Process the @code{cpp} spec. This is used to construct the arguments
11691 to be passed to the C preprocessor.
11694 Process the @code{cc1} spec. This is used to construct the options to be
11695 passed to the actual C compiler (@command{cc1}).
11698 Process the @code{cc1plus} spec. This is used to construct the options to be
11699 passed to the actual C++ compiler (@command{cc1plus}).
11702 Substitute the variable part of a matched option. See below.
11703 Note that each comma in the substituted string is replaced by
11707 Remove all occurrences of @code{-S} from the command line. Note---this
11708 command is position dependent. @samp{%} commands in the spec string
11709 before this one see @code{-S}, @samp{%} commands in the spec string
11710 after this one do not.
11712 @item %:@var{function}(@var{args})
11713 Call the named function @var{function}, passing it @var{args}.
11714 @var{args} is first processed as a nested spec string, then split
11715 into an argument vector in the usual fashion. The function returns
11716 a string which is processed as if it had appeared literally as part
11717 of the current spec.
11719 The following built-in spec functions are provided:
11722 @item @code{getenv}
11723 The @code{getenv} spec function takes two arguments: an environment
11724 variable name and a string. If the environment variable is not
11725 defined, a fatal error is issued. Otherwise, the return value is the
11726 value of the environment variable concatenated with the string. For
11727 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11730 %:getenv(TOPDIR /include)
11733 expands to @file{/path/to/top/include}.
11735 @item @code{if-exists}
11736 The @code{if-exists} spec function takes one argument, an absolute
11737 pathname to a file. If the file exists, @code{if-exists} returns the
11738 pathname. Here is a small example of its usage:
11742 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11745 @item @code{if-exists-else}
11746 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11747 spec function, except that it takes two arguments. The first argument is
11748 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11749 returns the pathname. If it does not exist, it returns the second argument.
11750 This way, @code{if-exists-else} can be used to select one file or another,
11751 based on the existence of the first. Here is a small example of its usage:
11755 crt0%O%s %:if-exists(crti%O%s) \
11756 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11759 @item @code{replace-outfile}
11760 The @code{replace-outfile} spec function takes two arguments. It looks for the
11761 first argument in the outfiles array and replaces it with the second argument. Here
11762 is a small example of its usage:
11765 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11768 @item @code{remove-outfile}
11769 The @code{remove-outfile} spec function takes one argument. It looks for the
11770 first argument in the outfiles array and removes it. Here is a small example
11774 %:remove-outfile(-lm)
11777 @item @code{pass-through-libs}
11778 The @code{pass-through-libs} spec function takes any number of arguments. It
11779 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11780 assumes are the names of linker input library archive files) and returns a
11781 result containing all the found arguments each prepended by
11782 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11783 intended to be passed to the LTO linker plugin.
11786 %:pass-through-libs(%G %L %G)
11789 @item @code{print-asm-header}
11790 The @code{print-asm-header} function takes no arguments and simply
11791 prints a banner like:
11797 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
11800 It is used to separate compiler options from assembler options
11801 in the @option{--target-help} output.
11804 @item %@{@code{S}@}
11805 Substitutes the @code{-S} switch, if that switch is given to GCC@.
11806 If that switch is not specified, this substitutes nothing. Note that
11807 the leading dash is omitted when specifying this option, and it is
11808 automatically inserted if the substitution is performed. Thus the spec
11809 string @samp{%@{foo@}} matches the command-line option @option{-foo}
11810 and outputs the command-line option @option{-foo}.
11812 @item %W@{@code{S}@}
11813 Like %@{@code{S}@} but mark last argument supplied within as a file to be
11814 deleted on failure.
11816 @item %@{@code{S}*@}
11817 Substitutes all the switches specified to GCC whose names start
11818 with @code{-S}, but which also take an argument. This is used for
11819 switches like @option{-o}, @option{-D}, @option{-I}, etc.
11820 GCC considers @option{-o foo} as being
11821 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
11822 text, including the space. Thus two arguments are generated.
11824 @item %@{@code{S}*&@code{T}*@}
11825 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
11826 (the order of @code{S} and @code{T} in the spec is not significant).
11827 There can be any number of ampersand-separated variables; for each the
11828 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
11830 @item %@{@code{S}:@code{X}@}
11831 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
11833 @item %@{!@code{S}:@code{X}@}
11834 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
11836 @item %@{@code{S}*:@code{X}@}
11837 Substitutes @code{X} if one or more switches whose names start with
11838 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
11839 once, no matter how many such switches appeared. However, if @code{%*}
11840 appears somewhere in @code{X}, then @code{X} is substituted once
11841 for each matching switch, with the @code{%*} replaced by the part of
11842 that switch matching the @code{*}.
11844 If @code{%*} appears as the last part of a spec sequence then a space
11845 is added after the end of the last substitution. If there is more
11846 text in the sequence, however, then a space is not generated. This
11847 allows the @code{%*} substitution to be used as part of a larger
11848 string. For example, a spec string like this:
11851 %@{mcu=*:--script=%*/memory.ld@}
11855 when matching an option like @option{-mcu=newchip} produces:
11858 --script=newchip/memory.ld
11861 @item %@{.@code{S}:@code{X}@}
11862 Substitutes @code{X}, if processing a file with suffix @code{S}.
11864 @item %@{!.@code{S}:@code{X}@}
11865 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
11867 @item %@{,@code{S}:@code{X}@}
11868 Substitutes @code{X}, if processing a file for language @code{S}.
11870 @item %@{!,@code{S}:@code{X}@}
11871 Substitutes @code{X}, if not processing a file for language @code{S}.
11873 @item %@{@code{S}|@code{P}:@code{X}@}
11874 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
11875 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
11876 @code{*} sequences as well, although they have a stronger binding than
11877 the @samp{|}. If @code{%*} appears in @code{X}, all of the
11878 alternatives must be starred, and only the first matching alternative
11881 For example, a spec string like this:
11884 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
11888 outputs the following command-line options from the following input
11889 command-line options:
11894 -d fred.c -foo -baz -boggle
11895 -d jim.d -bar -baz -boggle
11898 @item %@{S:X; T:Y; :D@}
11900 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
11901 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
11902 be as many clauses as you need. This may be combined with @code{.},
11903 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
11908 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
11909 construct may contain other nested @samp{%} constructs or spaces, or
11910 even newlines. They are processed as usual, as described above.
11911 Trailing white space in @code{X} is ignored. White space may also
11912 appear anywhere on the left side of the colon in these constructs,
11913 except between @code{.} or @code{*} and the corresponding word.
11915 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
11916 handled specifically in these constructs. If another value of
11917 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
11918 @option{-W} switch is found later in the command line, the earlier
11919 switch value is ignored, except with @{@code{S}*@} where @code{S} is
11920 just one letter, which passes all matching options.
11922 The character @samp{|} at the beginning of the predicate text is used to
11923 indicate that a command should be piped to the following command, but
11924 only if @option{-pipe} is specified.
11926 It is built into GCC which switches take arguments and which do not.
11927 (You might think it would be useful to generalize this to allow each
11928 compiler's spec to say which switches take arguments. But this cannot
11929 be done in a consistent fashion. GCC cannot even decide which input
11930 files have been specified without knowing which switches take arguments,
11931 and it must know which input files to compile in order to tell which
11934 GCC also knows implicitly that arguments starting in @option{-l} are to be
11935 treated as compiler output files, and passed to the linker in their
11936 proper position among the other output files.
11938 @c man begin OPTIONS
11940 @node Target Options
11941 @section Specifying Target Machine and Compiler Version
11942 @cindex target options
11943 @cindex cross compiling
11944 @cindex specifying machine version
11945 @cindex specifying compiler version and target machine
11946 @cindex compiler version, specifying
11947 @cindex target machine, specifying
11949 The usual way to run GCC is to run the executable called @command{gcc}, or
11950 @command{@var{machine}-gcc} when cross-compiling, or
11951 @command{@var{machine}-gcc-@var{version}} to run a version other than the
11952 one that was installed last.
11954 @node Submodel Options
11955 @section Hardware Models and Configurations
11956 @cindex submodel options
11957 @cindex specifying hardware config
11958 @cindex hardware models and configurations, specifying
11959 @cindex machine dependent options
11961 Each target machine types can have its own
11962 special options, starting with @samp{-m}, to choose among various
11963 hardware models or configurations---for example, 68010 vs 68020,
11964 floating coprocessor or none. A single installed version of the
11965 compiler can compile for any model or configuration, according to the
11968 Some configurations of the compiler also support additional special
11969 options, usually for compatibility with other compilers on the same
11972 @c This list is ordered alphanumerically by subsection name.
11973 @c It should be the same order and spelling as these options are listed
11974 @c in Machine Dependent Options
11977 * AArch64 Options::
11978 * Adapteva Epiphany Options::
11982 * Blackfin Options::
11987 * DEC Alpha Options::
11990 * GNU/Linux Options::
12000 * MicroBlaze Options::
12003 * MN10300 Options::
12007 * Nios II Options::
12008 * Nvidia PTX Options::
12010 * picoChip Options::
12011 * PowerPC Options::
12013 * RS/6000 and PowerPC Options::
12015 * S/390 and zSeries Options::
12018 * Solaris 2 Options::
12021 * System V Options::
12022 * TILE-Gx Options::
12023 * TILEPro Options::
12028 * VxWorks Options::
12030 * x86 Windows Options::
12031 * Xstormy16 Options::
12033 * zSeries Options::
12036 @node AArch64 Options
12037 @subsection AArch64 Options
12038 @cindex AArch64 Options
12040 These options are defined for AArch64 implementations:
12044 @item -mabi=@var{name}
12046 Generate code for the specified data model. Permissible values
12047 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12048 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12049 but long int and pointer are 64-bit.
12051 The default depends on the specific target configuration. Note that
12052 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12053 entire program with the same ABI, and link with a compatible set of libraries.
12056 @opindex mbig-endian
12057 Generate big-endian code. This is the default when GCC is configured for an
12058 @samp{aarch64_be-*-*} target.
12060 @item -mgeneral-regs-only
12061 @opindex mgeneral-regs-only
12062 Generate code which uses only the general registers.
12064 @item -mlittle-endian
12065 @opindex mlittle-endian
12066 Generate little-endian code. This is the default when GCC is configured for an
12067 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12069 @item -mcmodel=tiny
12070 @opindex mcmodel=tiny
12071 Generate code for the tiny code model. The program and its statically defined
12072 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12073 be statically or dynamically linked. This model is not fully implemented and
12074 mostly treated as @samp{small}.
12076 @item -mcmodel=small
12077 @opindex mcmodel=small
12078 Generate code for the small code model. The program and its statically defined
12079 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12080 be statically or dynamically linked. This is the default code model.
12082 @item -mcmodel=large
12083 @opindex mcmodel=large
12084 Generate code for the large code model. This makes no assumptions about
12085 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12086 statically linked only.
12088 @item -mstrict-align
12089 @opindex mstrict-align
12090 Do not assume that unaligned memory references are handled by the system.
12092 @item -momit-leaf-frame-pointer
12093 @itemx -mno-omit-leaf-frame-pointer
12094 @opindex momit-leaf-frame-pointer
12095 @opindex mno-omit-leaf-frame-pointer
12096 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12099 @item -mtls-dialect=desc
12100 @opindex mtls-dialect=desc
12101 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12102 of TLS variables. This is the default.
12104 @item -mtls-dialect=traditional
12105 @opindex mtls-dialect=traditional
12106 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12109 @item -mfix-cortex-a53-835769
12110 @itemx -mno-fix-cortex-a53-835769
12111 @opindex mfix-cortex-a53-835769
12112 @opindex mno-fix-cortex-a53-835769
12113 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12114 This involves inserting a NOP instruction between memory instructions and
12115 64-bit integer multiply-accumulate instructions.
12117 @item -march=@var{name}
12119 Specify the name of the target architecture, optionally suffixed by one or
12120 more feature modifiers. This option has the form
12121 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12122 only permissible value for @var{arch} is @samp{armv8-a}. The permissible
12123 values for @var{feature} are documented in the sub-section below.
12125 Where conflicting feature modifiers are specified, the right-most feature is
12128 GCC uses this name to determine what kind of instructions it can emit when
12129 generating assembly code.
12131 Where @option{-march} is specified without either of @option{-mtune}
12132 or @option{-mcpu} also being specified, the code is tuned to perform
12133 well across a range of target processors implementing the target
12136 @item -mtune=@var{name}
12138 Specify the name of the target processor for which GCC should tune the
12139 performance of the code. Permissible values for this option are:
12140 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57},
12141 @samp{cortex-a72}, @samp{thunderx}, @samp{xgene1}.
12143 Additionally, this option can specify that GCC should tune the performance
12144 of the code for a big.LITTLE system. Permissible values for this
12145 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12147 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12148 are specified, the code is tuned to perform well across a range
12149 of target processors.
12151 This option cannot be suffixed by feature modifiers.
12153 @item -mcpu=@var{name}
12155 Specify the name of the target processor, optionally suffixed by one or more
12156 feature modifiers. This option has the form
12157 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12158 permissible values for @var{cpu} are the same as those available for
12161 The permissible values for @var{feature} are documented in the sub-section
12164 Where conflicting feature modifiers are specified, the right-most feature is
12167 GCC uses this name to determine what kind of instructions it can emit when
12168 generating assembly code (as if by @option{-march}) and to determine
12169 the target processor for which to tune for performance (as if
12170 by @option{-mtune}). Where this option is used in conjunction
12171 with @option{-march} or @option{-mtune}, those options take precedence
12172 over the appropriate part of this option.
12175 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12176 @cindex @option{-march} feature modifiers
12177 @cindex @option{-mcpu} feature modifiers
12178 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
12183 Enable CRC extension.
12185 Enable Crypto extension. This implies Advanced SIMD is enabled.
12187 Enable floating-point instructions.
12189 Enable Advanced SIMD instructions. This implies floating-point instructions
12190 are enabled. This is the default for all current possible values for options
12191 @option{-march} and @option{-mcpu=}.
12194 @node Adapteva Epiphany Options
12195 @subsection Adapteva Epiphany Options
12197 These @samp{-m} options are defined for Adapteva Epiphany:
12200 @item -mhalf-reg-file
12201 @opindex mhalf-reg-file
12202 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12203 That allows code to run on hardware variants that lack these registers.
12205 @item -mprefer-short-insn-regs
12206 @opindex mprefer-short-insn-regs
12207 Preferrentially allocate registers that allow short instruction generation.
12208 This can result in increased instruction count, so this may either reduce or
12209 increase overall code size.
12211 @item -mbranch-cost=@var{num}
12212 @opindex mbranch-cost
12213 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12214 This cost is only a heuristic and is not guaranteed to produce
12215 consistent results across releases.
12219 Enable the generation of conditional moves.
12221 @item -mnops=@var{num}
12223 Emit @var{num} NOPs before every other generated instruction.
12225 @item -mno-soft-cmpsf
12226 @opindex mno-soft-cmpsf
12227 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12228 and test the flags. This is faster than a software comparison, but can
12229 get incorrect results in the presence of NaNs, or when two different small
12230 numbers are compared such that their difference is calculated as zero.
12231 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12232 software comparisons.
12234 @item -mstack-offset=@var{num}
12235 @opindex mstack-offset
12236 Set the offset between the top of the stack and the stack pointer.
12237 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12238 can be used by leaf functions without stack allocation.
12239 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12240 Note also that this option changes the ABI; compiling a program with a
12241 different stack offset than the libraries have been compiled with
12242 generally does not work.
12243 This option can be useful if you want to evaluate if a different stack
12244 offset would give you better code, but to actually use a different stack
12245 offset to build working programs, it is recommended to configure the
12246 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12248 @item -mno-round-nearest
12249 @opindex mno-round-nearest
12250 Make the scheduler assume that the rounding mode has been set to
12251 truncating. The default is @option{-mround-nearest}.
12254 @opindex mlong-calls
12255 If not otherwise specified by an attribute, assume all calls might be beyond
12256 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12257 function address into a register before performing a (otherwise direct) call.
12258 This is the default.
12260 @item -mshort-calls
12261 @opindex short-calls
12262 If not otherwise specified by an attribute, assume all direct calls are
12263 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12264 for direct calls. The default is @option{-mlong-calls}.
12268 Assume addresses can be loaded as 16-bit unsigned values. This does not
12269 apply to function addresses for which @option{-mlong-calls} semantics
12272 @item -mfp-mode=@var{mode}
12274 Set the prevailing mode of the floating-point unit.
12275 This determines the floating-point mode that is provided and expected
12276 at function call and return time. Making this mode match the mode you
12277 predominantly need at function start can make your programs smaller and
12278 faster by avoiding unnecessary mode switches.
12280 @var{mode} can be set to one the following values:
12284 Any mode at function entry is valid, and retained or restored when
12285 the function returns, and when it calls other functions.
12286 This mode is useful for compiling libraries or other compilation units
12287 you might want to incorporate into different programs with different
12288 prevailing FPU modes, and the convenience of being able to use a single
12289 object file outweighs the size and speed overhead for any extra
12290 mode switching that might be needed, compared with what would be needed
12291 with a more specific choice of prevailing FPU mode.
12294 This is the mode used for floating-point calculations with
12295 truncating (i.e.@: round towards zero) rounding mode. That includes
12296 conversion from floating point to integer.
12298 @item round-nearest
12299 This is the mode used for floating-point calculations with
12300 round-to-nearest-or-even rounding mode.
12303 This is the mode used to perform integer calculations in the FPU, e.g.@:
12304 integer multiply, or integer multiply-and-accumulate.
12307 The default is @option{-mfp-mode=caller}
12309 @item -mnosplit-lohi
12310 @itemx -mno-postinc
12311 @itemx -mno-postmodify
12312 @opindex mnosplit-lohi
12313 @opindex mno-postinc
12314 @opindex mno-postmodify
12315 Code generation tweaks that disable, respectively, splitting of 32-bit
12316 loads, generation of post-increment addresses, and generation of
12317 post-modify addresses. The defaults are @option{msplit-lohi},
12318 @option{-mpost-inc}, and @option{-mpost-modify}.
12320 @item -mnovect-double
12321 @opindex mno-vect-double
12322 Change the preferred SIMD mode to SImode. The default is
12323 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12325 @item -max-vect-align=@var{num}
12326 @opindex max-vect-align
12327 The maximum alignment for SIMD vector mode types.
12328 @var{num} may be 4 or 8. The default is 8.
12329 Note that this is an ABI change, even though many library function
12330 interfaces are unaffected if they don't use SIMD vector modes
12331 in places that affect size and/or alignment of relevant types.
12333 @item -msplit-vecmove-early
12334 @opindex msplit-vecmove-early
12335 Split vector moves into single word moves before reload. In theory this
12336 can give better register allocation, but so far the reverse seems to be
12337 generally the case.
12339 @item -m1reg-@var{reg}
12341 Specify a register to hold the constant @minus{}1, which makes loading small negative
12342 constants and certain bitmasks faster.
12343 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12344 which specify use of that register as a fixed register,
12345 and @samp{none}, which means that no register is used for this
12346 purpose. The default is @option{-m1reg-none}.
12351 @subsection ARC Options
12352 @cindex ARC options
12354 The following options control the architecture variant for which code
12357 @c architecture variants
12360 @item -mbarrel-shifter
12361 @opindex mbarrel-shifter
12362 Generate instructions supported by barrel shifter. This is the default
12363 unless @option{-mcpu=ARC601} is in effect.
12365 @item -mcpu=@var{cpu}
12367 Set architecture type, register usage, and instruction scheduling
12368 parameters for @var{cpu}. There are also shortcut alias options
12369 available for backward compatibility and convenience. Supported
12370 values for @var{cpu} are
12376 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12380 Compile for ARC601. Alias: @option{-mARC601}.
12385 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12386 This is the default when configured with @option{--with-cpu=arc700}@.
12391 @itemx -mdpfp-compact
12392 @opindex mdpfp-compact
12393 FPX: Generate Double Precision FPX instructions, tuned for the compact
12397 @opindex mdpfp-fast
12398 FPX: Generate Double Precision FPX instructions, tuned for the fast
12401 @item -mno-dpfp-lrsr
12402 @opindex mno-dpfp-lrsr
12403 Disable LR and SR instructions from using FPX extension aux registers.
12407 Generate Extended arithmetic instructions. Currently only
12408 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12409 supported. This is always enabled for @option{-mcpu=ARC700}.
12413 Do not generate mpy instructions for ARC700.
12417 Generate 32x16 bit multiply and mac instructions.
12421 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12425 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12430 @itemx -mspfp-compact
12431 @opindex mspfp-compact
12432 FPX: Generate Single Precision FPX instructions, tuned for the compact
12436 @opindex mspfp-fast
12437 FPX: Generate Single Precision FPX instructions, tuned for the fast
12442 Enable generation of ARC SIMD instructions via target-specific
12443 builtins. Only valid for @option{-mcpu=ARC700}.
12446 @opindex msoft-float
12447 This option ignored; it is provided for compatibility purposes only.
12448 Software floating point code is emitted by default, and this default
12449 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12450 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12451 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12455 Generate swap instructions.
12459 The following options are passed through to the assembler, and also
12460 define preprocessor macro symbols.
12462 @c Flags used by the assembler, but for which we define preprocessor
12463 @c macro symbols as well.
12466 @opindex mdsp-packa
12467 Passed down to the assembler to enable the DSP Pack A extensions.
12468 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12472 Passed down to the assembler to enable the dual viterbi butterfly
12473 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12475 @c ARC700 4.10 extension instruction
12478 Passed down to the assembler to enable the Locked Load/Store
12479 Conditional extension. Also sets the preprocessor symbol
12484 Passed down to the assembler. Also sets the preprocessor symbol
12485 @code{__Xxmac_d16}.
12489 Passed down to the assembler. Also sets the preprocessor symbol
12492 @c ARC700 4.10 extension instruction
12495 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12496 extension instruction. Also sets the preprocessor symbol
12499 @c ARC700 4.10 extension instruction
12502 Passed down to the assembler to enable the swap byte ordering
12503 extension instruction. Also sets the preprocessor symbol
12507 @opindex mtelephony
12508 Passed down to the assembler to enable dual and single operand
12509 instructions for telephony. Also sets the preprocessor symbol
12510 @code{__Xtelephony}.
12514 Passed down to the assembler to enable the XY Memory extension. Also
12515 sets the preprocessor symbol @code{__Xxy}.
12519 The following options control how the assembly code is annotated:
12521 @c Assembly annotation options
12525 Annotate assembler instructions with estimated addresses.
12527 @item -mannotate-align
12528 @opindex mannotate-align
12529 Explain what alignment considerations lead to the decision to make an
12530 instruction short or long.
12534 The following options are passed through to the linker:
12536 @c options passed through to the linker
12540 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12541 This option is enabled by default in tool chains built for
12542 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12543 when profiling is not requested.
12545 @item -marclinux_prof
12546 @opindex marclinux_prof
12547 Passed through to the linker, to specify use of the
12548 @code{arclinux_prof} emulation. This option is enabled by default in
12549 tool chains built for @w{@code{arc-linux-uclibc}} and
12550 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12554 The following options control the semantics of generated code:
12556 @c semantically relevant code generation options
12558 @item -mepilogue-cfi
12559 @opindex mepilogue-cfi
12560 Enable generation of call frame information for epilogues.
12562 @item -mno-epilogue-cfi
12563 @opindex mno-epilogue-cfi
12564 Disable generation of call frame information for epilogues.
12567 @opindex mlong-calls
12568 Generate call insns as register indirect calls, thus providing access
12569 to the full 32-bit address range.
12571 @item -mmedium-calls
12572 @opindex mmedium-calls
12573 Don't use less than 25 bit addressing range for calls, which is the
12574 offset available for an unconditional branch-and-link
12575 instruction. Conditional execution of function calls is suppressed, to
12576 allow use of the 25-bit range, rather than the 21-bit range with
12577 conditional branch-and-link. This is the default for tool chains built
12578 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12582 Do not generate sdata references. This is the default for tool chains
12583 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12587 @opindex mucb-mcount
12588 Instrument with mcount calls as used in UCB code. I.e. do the
12589 counting in the callee, not the caller. By default ARC instrumentation
12590 counts in the caller.
12592 @item -mvolatile-cache
12593 @opindex mvolatile-cache
12594 Use ordinarily cached memory accesses for volatile references. This is the
12597 @item -mno-volatile-cache
12598 @opindex mno-volatile-cache
12599 Enable cache bypass for volatile references.
12603 The following options fine tune code generation:
12604 @c code generation tuning options
12607 @opindex malign-call
12608 Do alignment optimizations for call instructions.
12610 @item -mauto-modify-reg
12611 @opindex mauto-modify-reg
12612 Enable the use of pre/post modify with register displacement.
12614 @item -mbbit-peephole
12615 @opindex mbbit-peephole
12616 Enable bbit peephole2.
12620 This option disables a target-specific pass in @file{arc_reorg} to
12621 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12622 generation driven by the combiner pass.
12624 @item -mcase-vector-pcrel
12625 @opindex mcase-vector-pcrel
12626 Use pc-relative switch case tables - this enables case table shortening.
12627 This is the default for @option{-Os}.
12629 @item -mcompact-casesi
12630 @opindex mcompact-casesi
12631 Enable compact casesi pattern.
12632 This is the default for @option{-Os}.
12634 @item -mno-cond-exec
12635 @opindex mno-cond-exec
12636 Disable ARCompact specific pass to generate conditional execution instructions.
12637 Due to delay slot scheduling and interactions between operand numbers,
12638 literal sizes, instruction lengths, and the support for conditional execution,
12639 the target-independent pass to generate conditional execution is often lacking,
12640 so the ARC port has kept a special pass around that tries to find more
12641 conditional execution generating opportunities after register allocation,
12642 branch shortening, and delay slot scheduling have been done. This pass
12643 generally, but not always, improves performance and code size, at the cost of
12644 extra compilation time, which is why there is an option to switch it off.
12645 If you have a problem with call instructions exceeding their allowable
12646 offset range because they are conditionalized, you should consider using
12647 @option{-mmedium-calls} instead.
12649 @item -mearly-cbranchsi
12650 @opindex mearly-cbranchsi
12651 Enable pre-reload use of the cbranchsi pattern.
12653 @item -mexpand-adddi
12654 @opindex mexpand-adddi
12655 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12656 @code{add.f}, @code{adc} etc.
12658 @item -mindexed-loads
12659 @opindex mindexed-loads
12660 Enable the use of indexed loads. This can be problematic because some
12661 optimizers then assume that indexed stores exist, which is not
12666 Enable Local Register Allocation. This is still experimental for ARC,
12667 so by default the compiler uses standard reload
12668 (i.e. @option{-mno-lra}).
12670 @item -mlra-priority-none
12671 @opindex mlra-priority-none
12672 Don't indicate any priority for target registers.
12674 @item -mlra-priority-compact
12675 @opindex mlra-priority-compact
12676 Indicate target register priority for r0..r3 / r12..r15.
12678 @item -mlra-priority-noncompact
12679 @opindex mlra-priority-noncompact
12680 Reduce target regsiter priority for r0..r3 / r12..r15.
12682 @item -mno-millicode
12683 @opindex mno-millicode
12684 When optimizing for size (using @option{-Os}), prologues and epilogues
12685 that have to save or restore a large number of registers are often
12686 shortened by using call to a special function in libgcc; this is
12687 referred to as a @emph{millicode} call. As these calls can pose
12688 performance issues, and/or cause linking issues when linking in a
12689 nonstandard way, this option is provided to turn off millicode call
12693 @opindex mmixed-code
12694 Tweak register allocation to help 16-bit instruction generation.
12695 This generally has the effect of decreasing the average instruction size
12696 while increasing the instruction count.
12700 Enable 'q' instruction alternatives.
12701 This is the default for @option{-Os}.
12705 Enable Rcq constraint handling - most short code generation depends on this.
12706 This is the default.
12710 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12711 This is the default.
12713 @item -msize-level=@var{level}
12714 @opindex msize-level
12715 Fine-tune size optimization with regards to instruction lengths and alignment.
12716 The recognized values for @var{level} are:
12719 No size optimization. This level is deprecated and treated like @samp{1}.
12722 Short instructions are used opportunistically.
12725 In addition, alignment of loops and of code after barriers are dropped.
12728 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12732 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12733 the behavior when this is not set is equivalent to level @samp{1}.
12735 @item -mtune=@var{cpu}
12737 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12738 by @option{-mcpu=}.
12740 Supported values for @var{cpu} are
12744 Tune for ARC600 cpu.
12747 Tune for ARC601 cpu.
12750 Tune for ARC700 cpu with standard multiplier block.
12753 Tune for ARC700 cpu with XMAC block.
12756 Tune for ARC725D cpu.
12759 Tune for ARC750D cpu.
12763 @item -mmultcost=@var{num}
12765 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12766 normal instruction.
12768 @item -munalign-prob-threshold=@var{probability}
12769 @opindex munalign-prob-threshold
12770 Set probability threshold for unaligning branches.
12771 When tuning for @samp{ARC700} and optimizing for speed, branches without
12772 filled delay slot are preferably emitted unaligned and long, unless
12773 profiling indicates that the probability for the branch to be taken
12774 is below @var{probability}. @xref{Cross-profiling}.
12775 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
12779 The following options are maintained for backward compatibility, but
12780 are now deprecated and will be removed in a future release:
12782 @c Deprecated options
12790 @opindex mbig-endian
12793 Compile code for big endian targets. Use of these options is now
12794 deprecated. Users wanting big-endian code, should use the
12795 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
12796 building the tool chain, for which big-endian is the default.
12798 @item -mlittle-endian
12799 @opindex mlittle-endian
12802 Compile code for little endian targets. Use of these options is now
12803 deprecated. Users wanting little-endian code should use the
12804 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
12805 building the tool chain, for which little-endian is the default.
12807 @item -mbarrel_shifter
12808 @opindex mbarrel_shifter
12809 Replaced by @option{-mbarrel-shifter}.
12811 @item -mdpfp_compact
12812 @opindex mdpfp_compact
12813 Replaced by @option{-mdpfp-compact}.
12816 @opindex mdpfp_fast
12817 Replaced by @option{-mdpfp-fast}.
12820 @opindex mdsp_packa
12821 Replaced by @option{-mdsp-packa}.
12825 Replaced by @option{-mea}.
12829 Replaced by @option{-mmac-24}.
12833 Replaced by @option{-mmac-d16}.
12835 @item -mspfp_compact
12836 @opindex mspfp_compact
12837 Replaced by @option{-mspfp-compact}.
12840 @opindex mspfp_fast
12841 Replaced by @option{-mspfp-fast}.
12843 @item -mtune=@var{cpu}
12845 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
12846 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
12847 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
12849 @item -multcost=@var{num}
12851 Replaced by @option{-mmultcost}.
12856 @subsection ARM Options
12857 @cindex ARM options
12859 These @samp{-m} options are defined for the ARM port:
12862 @item -mabi=@var{name}
12864 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
12865 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
12868 @opindex mapcs-frame
12869 Generate a stack frame that is compliant with the ARM Procedure Call
12870 Standard for all functions, even if this is not strictly necessary for
12871 correct execution of the code. Specifying @option{-fomit-frame-pointer}
12872 with this option causes the stack frames not to be generated for
12873 leaf functions. The default is @option{-mno-apcs-frame}.
12874 This option is deprecated.
12878 This is a synonym for @option{-mapcs-frame} and is deprecated.
12881 @c not currently implemented
12882 @item -mapcs-stack-check
12883 @opindex mapcs-stack-check
12884 Generate code to check the amount of stack space available upon entry to
12885 every function (that actually uses some stack space). If there is
12886 insufficient space available then either the function
12887 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
12888 called, depending upon the amount of stack space required. The runtime
12889 system is required to provide these functions. The default is
12890 @option{-mno-apcs-stack-check}, since this produces smaller code.
12892 @c not currently implemented
12894 @opindex mapcs-float
12895 Pass floating-point arguments using the floating-point registers. This is
12896 one of the variants of the APCS@. This option is recommended if the
12897 target hardware has a floating-point unit or if a lot of floating-point
12898 arithmetic is going to be performed by the code. The default is
12899 @option{-mno-apcs-float}, since the size of integer-only code is
12900 slightly increased if @option{-mapcs-float} is used.
12902 @c not currently implemented
12903 @item -mapcs-reentrant
12904 @opindex mapcs-reentrant
12905 Generate reentrant, position-independent code. The default is
12906 @option{-mno-apcs-reentrant}.
12909 @item -mthumb-interwork
12910 @opindex mthumb-interwork
12911 Generate code that supports calling between the ARM and Thumb
12912 instruction sets. Without this option, on pre-v5 architectures, the
12913 two instruction sets cannot be reliably used inside one program. The
12914 default is @option{-mno-thumb-interwork}, since slightly larger code
12915 is generated when @option{-mthumb-interwork} is specified. In AAPCS
12916 configurations this option is meaningless.
12918 @item -mno-sched-prolog
12919 @opindex mno-sched-prolog
12920 Prevent the reordering of instructions in the function prologue, or the
12921 merging of those instruction with the instructions in the function's
12922 body. This means that all functions start with a recognizable set
12923 of instructions (or in fact one of a choice from a small set of
12924 different function prologues), and this information can be used to
12925 locate the start of functions inside an executable piece of code. The
12926 default is @option{-msched-prolog}.
12928 @item -mfloat-abi=@var{name}
12929 @opindex mfloat-abi
12930 Specifies which floating-point ABI to use. Permissible values
12931 are: @samp{soft}, @samp{softfp} and @samp{hard}.
12933 Specifying @samp{soft} causes GCC to generate output containing
12934 library calls for floating-point operations.
12935 @samp{softfp} allows the generation of code using hardware floating-point
12936 instructions, but still uses the soft-float calling conventions.
12937 @samp{hard} allows generation of floating-point instructions
12938 and uses FPU-specific calling conventions.
12940 The default depends on the specific target configuration. Note that
12941 the hard-float and soft-float ABIs are not link-compatible; you must
12942 compile your entire program with the same ABI, and link with a
12943 compatible set of libraries.
12945 @item -mlittle-endian
12946 @opindex mlittle-endian
12947 Generate code for a processor running in little-endian mode. This is
12948 the default for all standard configurations.
12951 @opindex mbig-endian
12952 Generate code for a processor running in big-endian mode; the default is
12953 to compile code for a little-endian processor.
12955 @item -march=@var{name}
12957 This specifies the name of the target ARM architecture. GCC uses this
12958 name to determine what kind of instructions it can emit when generating
12959 assembly code. This option can be used in conjunction with or instead
12960 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
12961 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
12962 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
12963 @samp{armv6}, @samp{armv6j},
12964 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
12965 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
12966 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
12967 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
12969 @option{-march=armv7ve} is the armv7-a architecture with virtualization
12972 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
12973 architecture together with the optional CRC32 extensions.
12975 @option{-march=native} causes the compiler to auto-detect the architecture
12976 of the build computer. At present, this feature is only supported on
12977 GNU/Linux, and not all architectures are recognized. If the auto-detect
12978 is unsuccessful the option has no effect.
12980 @item -mtune=@var{name}
12982 This option specifies the name of the target ARM processor for
12983 which GCC should tune the performance of the code.
12984 For some ARM implementations better performance can be obtained by using
12986 Permissible names are: @samp{arm2}, @samp{arm250},
12987 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
12988 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
12989 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
12990 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
12992 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
12993 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
12994 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
12995 @samp{strongarm1110},
12996 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
12997 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
12998 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
12999 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13000 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13001 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13002 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13003 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
13004 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53},
13005 @samp{cortex-a57}, @samp{cortex-a72},
13007 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13012 @samp{cortex-m0plus},
13013 @samp{cortex-m1.small-multiply},
13014 @samp{cortex-m0.small-multiply},
13015 @samp{cortex-m0plus.small-multiply},
13016 @samp{marvell-pj4},
13017 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13018 @samp{fa526}, @samp{fa626},
13019 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13022 Additionally, this option can specify that GCC should tune the performance
13023 of the code for a big.LITTLE system. Permissible names are:
13024 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53},
13025 @samp{cortex-a72.cortex-a53}.
13027 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13028 performance for a blend of processors within architecture @var{arch}.
13029 The aim is to generate code that run well on the current most popular
13030 processors, balancing between optimizations that benefit some CPUs in the
13031 range, and avoiding performance pitfalls of other CPUs. The effects of
13032 this option may change in future GCC versions as CPU models come and go.
13034 @option{-mtune=native} causes the compiler to auto-detect the CPU
13035 of the build computer. At present, this feature is only supported on
13036 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13037 unsuccessful the option has no effect.
13039 @item -mcpu=@var{name}
13041 This specifies the name of the target ARM processor. GCC uses this name
13042 to derive the name of the target ARM architecture (as if specified
13043 by @option{-march}) and the ARM processor type for which to tune for
13044 performance (as if specified by @option{-mtune}). Where this option
13045 is used in conjunction with @option{-march} or @option{-mtune},
13046 those options take precedence over the appropriate part of this option.
13048 Permissible names for this option are the same as those for
13051 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13052 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13053 See @option{-mtune} for more information.
13055 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13056 of the build computer. At present, this feature is only supported on
13057 GNU/Linux, and not all architectures are recognized. If the auto-detect
13058 is unsuccessful the option has no effect.
13060 @item -mfpu=@var{name}
13062 This specifies what floating-point hardware (or hardware emulation) is
13063 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13064 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13065 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13066 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13067 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13068 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13070 If @option{-msoft-float} is specified this specifies the format of
13071 floating-point values.
13073 If the selected floating-point hardware includes the NEON extension
13074 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13075 operations are not generated by GCC's auto-vectorization pass unless
13076 @option{-funsafe-math-optimizations} is also specified. This is
13077 because NEON hardware does not fully implement the IEEE 754 standard for
13078 floating-point arithmetic (in particular denormal values are treated as
13079 zero), so the use of NEON instructions may lead to a loss of precision.
13081 @item -mfp16-format=@var{name}
13082 @opindex mfp16-format
13083 Specify the format of the @code{__fp16} half-precision floating-point type.
13084 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13085 the default is @samp{none}, in which case the @code{__fp16} type is not
13086 defined. @xref{Half-Precision}, for more information.
13088 @item -mstructure-size-boundary=@var{n}
13089 @opindex mstructure-size-boundary
13090 The sizes of all structures and unions are rounded up to a multiple
13091 of the number of bits set by this option. Permissible values are 8, 32
13092 and 64. The default value varies for different toolchains. For the COFF
13093 targeted toolchain the default value is 8. A value of 64 is only allowed
13094 if the underlying ABI supports it.
13096 Specifying a larger number can produce faster, more efficient code, but
13097 can also increase the size of the program. Different values are potentially
13098 incompatible. Code compiled with one value cannot necessarily expect to
13099 work with code or libraries compiled with another value, if they exchange
13100 information using structures or unions.
13102 @item -mabort-on-noreturn
13103 @opindex mabort-on-noreturn
13104 Generate a call to the function @code{abort} at the end of a
13105 @code{noreturn} function. It is executed if the function tries to
13109 @itemx -mno-long-calls
13110 @opindex mlong-calls
13111 @opindex mno-long-calls
13112 Tells the compiler to perform function calls by first loading the
13113 address of the function into a register and then performing a subroutine
13114 call on this register. This switch is needed if the target function
13115 lies outside of the 64-megabyte addressing range of the offset-based
13116 version of subroutine call instruction.
13118 Even if this switch is enabled, not all function calls are turned
13119 into long calls. The heuristic is that static functions, functions
13120 that have the @code{short_call} attribute, functions that are inside
13121 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13122 definitions have already been compiled within the current compilation
13123 unit are not turned into long calls. The exceptions to this rule are
13124 that weak function definitions, functions with the @code{long_call}
13125 attribute or the @code{section} attribute, and functions that are within
13126 the scope of a @code{#pragma long_calls} directive are always
13127 turned into long calls.
13129 This feature is not enabled by default. Specifying
13130 @option{-mno-long-calls} restores the default behavior, as does
13131 placing the function calls within the scope of a @code{#pragma
13132 long_calls_off} directive. Note these switches have no effect on how
13133 the compiler generates code to handle function calls via function
13136 @item -msingle-pic-base
13137 @opindex msingle-pic-base
13138 Treat the register used for PIC addressing as read-only, rather than
13139 loading it in the prologue for each function. The runtime system is
13140 responsible for initializing this register with an appropriate value
13141 before execution begins.
13143 @item -mpic-register=@var{reg}
13144 @opindex mpic-register
13145 Specify the register to be used for PIC addressing.
13146 For standard PIC base case, the default is any suitable register
13147 determined by compiler. For single PIC base case, the default is
13148 @samp{R9} if target is EABI based or stack-checking is enabled,
13149 otherwise the default is @samp{R10}.
13151 @item -mpic-data-is-text-relative
13152 @opindex mpic-data-is-text-relative
13153 Assume that each data segments are relative to text segment at load time.
13154 Therefore, it permits addressing data using PC-relative operations.
13155 This option is on by default for targets other than VxWorks RTP.
13157 @item -mpoke-function-name
13158 @opindex mpoke-function-name
13159 Write the name of each function into the text section, directly
13160 preceding the function prologue. The generated code is similar to this:
13164 .ascii "arm_poke_function_name", 0
13167 .word 0xff000000 + (t1 - t0)
13168 arm_poke_function_name
13170 stmfd sp!, @{fp, ip, lr, pc@}
13174 When performing a stack backtrace, code can inspect the value of
13175 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13176 location @code{pc - 12} and the top 8 bits are set, then we know that
13177 there is a function name embedded immediately preceding this location
13178 and has length @code{((pc[-3]) & 0xff000000)}.
13185 Select between generating code that executes in ARM and Thumb
13186 states. The default for most configurations is to generate code
13187 that executes in ARM state, but the default can be changed by
13188 configuring GCC with the @option{--with-mode=}@var{state}
13192 @opindex mtpcs-frame
13193 Generate a stack frame that is compliant with the Thumb Procedure Call
13194 Standard for all non-leaf functions. (A leaf function is one that does
13195 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13197 @item -mtpcs-leaf-frame
13198 @opindex mtpcs-leaf-frame
13199 Generate a stack frame that is compliant with the Thumb Procedure Call
13200 Standard for all leaf functions. (A leaf function is one that does
13201 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13203 @item -mcallee-super-interworking
13204 @opindex mcallee-super-interworking
13205 Gives all externally visible functions in the file being compiled an ARM
13206 instruction set header which switches to Thumb mode before executing the
13207 rest of the function. This allows these functions to be called from
13208 non-interworking code. This option is not valid in AAPCS configurations
13209 because interworking is enabled by default.
13211 @item -mcaller-super-interworking
13212 @opindex mcaller-super-interworking
13213 Allows calls via function pointers (including virtual functions) to
13214 execute correctly regardless of whether the target code has been
13215 compiled for interworking or not. There is a small overhead in the cost
13216 of executing a function pointer if this option is enabled. This option
13217 is not valid in AAPCS configurations because interworking is enabled
13220 @item -mtp=@var{name}
13222 Specify the access model for the thread local storage pointer. The valid
13223 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13224 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13225 (supported in the arm6k architecture), and @samp{auto}, which uses the
13226 best available method for the selected processor. The default setting is
13229 @item -mtls-dialect=@var{dialect}
13230 @opindex mtls-dialect
13231 Specify the dialect to use for accessing thread local storage. Two
13232 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13233 @samp{gnu} dialect selects the original GNU scheme for supporting
13234 local and global dynamic TLS models. The @samp{gnu2} dialect
13235 selects the GNU descriptor scheme, which provides better performance
13236 for shared libraries. The GNU descriptor scheme is compatible with
13237 the original scheme, but does require new assembler, linker and
13238 library support. Initial and local exec TLS models are unaffected by
13239 this option and always use the original scheme.
13241 @item -mword-relocations
13242 @opindex mword-relocations
13243 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13244 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13245 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13248 @item -mfix-cortex-m3-ldrd
13249 @opindex mfix-cortex-m3-ldrd
13250 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13251 with overlapping destination and base registers are used. This option avoids
13252 generating these instructions. This option is enabled by default when
13253 @option{-mcpu=cortex-m3} is specified.
13255 @item -munaligned-access
13256 @itemx -mno-unaligned-access
13257 @opindex munaligned-access
13258 @opindex mno-unaligned-access
13259 Enables (or disables) reading and writing of 16- and 32- bit values
13260 from addresses that are not 16- or 32- bit aligned. By default
13261 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13262 architectures, and enabled for all other architectures. If unaligned
13263 access is not enabled then words in packed data structures are
13264 accessed a byte at a time.
13266 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13267 generated object file to either true or false, depending upon the
13268 setting of this option. If unaligned access is enabled then the
13269 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13272 @item -mneon-for-64bits
13273 @opindex mneon-for-64bits
13274 Enables using Neon to handle scalar 64-bits operations. This is
13275 disabled by default since the cost of moving data from core registers
13278 @item -mslow-flash-data
13279 @opindex mslow-flash-data
13280 Assume loading data from flash is slower than fetching instruction.
13281 Therefore literal load is minimized for better performance.
13282 This option is only supported when compiling for ARMv7 M-profile and
13285 @item -masm-syntax-unified
13286 @opindex masm-syntax-unified
13287 Assume inline assembler is using unified asm syntax. The default is
13288 currently off which implies divided syntax. Currently this option is
13289 available only for Thumb1 and has no effect on ARM state and Thumb2.
13290 However, this may change in future releases of GCC. Divided syntax
13291 should be considered deprecated.
13293 @item -mrestrict-it
13294 @opindex mrestrict-it
13295 Restricts generation of IT blocks to conform to the rules of ARMv8.
13296 IT blocks can only contain a single 16-bit instruction from a select
13297 set of instructions. This option is on by default for ARMv8 Thumb mode.
13301 @subsection AVR Options
13302 @cindex AVR Options
13304 These options are defined for AVR implementations:
13307 @item -mmcu=@var{mcu}
13309 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13311 The default for this option is@tie{}@samp{avr2}.
13313 GCC supports the following AVR devices and ISAs:
13315 @include avr-mmcu.texi
13317 @item -maccumulate-args
13318 @opindex maccumulate-args
13319 Accumulate outgoing function arguments and acquire/release the needed
13320 stack space for outgoing function arguments once in function
13321 prologue/epilogue. Without this option, outgoing arguments are pushed
13322 before calling a function and popped afterwards.
13324 Popping the arguments after the function call can be expensive on
13325 AVR so that accumulating the stack space might lead to smaller
13326 executables because arguments need not to be removed from the
13327 stack after such a function call.
13329 This option can lead to reduced code size for functions that perform
13330 several calls to functions that get their arguments on the stack like
13331 calls to printf-like functions.
13333 @item -mbranch-cost=@var{cost}
13334 @opindex mbranch-cost
13335 Set the branch costs for conditional branch instructions to
13336 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13337 integers. The default branch cost is 0.
13339 @item -mcall-prologues
13340 @opindex mcall-prologues
13341 Functions prologues/epilogues are expanded as calls to appropriate
13342 subroutines. Code size is smaller.
13346 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13347 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13348 and @code{long long} is 4 bytes. Please note that this option does not
13349 conform to the C standards, but it results in smaller code
13352 @item -mno-interrupts
13353 @opindex mno-interrupts
13354 Generated code is not compatible with hardware interrupts.
13355 Code size is smaller.
13359 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13360 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13361 Setting @option{-mrelax} just adds the @option{--relax} option to the
13362 linker command line when the linker is called.
13364 Jump relaxing is performed by the linker because jump offsets are not
13365 known before code is located. Therefore, the assembler code generated by the
13366 compiler is the same, but the instructions in the executable may
13367 differ from instructions in the assembler code.
13369 Relaxing must be turned on if linker stubs are needed, see the
13370 section on @code{EIND} and linker stubs below.
13374 Treat the stack pointer register as an 8-bit register,
13375 i.e.@: assume the high byte of the stack pointer is zero.
13376 In general, you don't need to set this option by hand.
13378 This option is used internally by the compiler to select and
13379 build multilibs for architectures @code{avr2} and @code{avr25}.
13380 These architectures mix devices with and without @code{SPH}.
13381 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13382 the compiler driver adds or removes this option from the compiler
13383 proper's command line, because the compiler then knows if the device
13384 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13389 Use address register @code{X} in a way proposed by the hardware. This means
13390 that @code{X} is only used in indirect, post-increment or
13391 pre-decrement addressing.
13393 Without this option, the @code{X} register may be used in the same way
13394 as @code{Y} or @code{Z} which then is emulated by additional
13396 For example, loading a value with @code{X+const} addressing with a
13397 small non-negative @code{const < 64} to a register @var{Rn} is
13401 adiw r26, const ; X += const
13402 ld @var{Rn}, X ; @var{Rn} = *X
13403 sbiw r26, const ; X -= const
13407 @opindex mtiny-stack
13408 Only change the lower 8@tie{}bits of the stack pointer.
13410 @item -Waddr-space-convert
13411 @opindex Waddr-space-convert
13412 Warn about conversions between address spaces in the case where the
13413 resulting address space is not contained in the incoming address space.
13416 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
13417 @cindex @code{EIND}
13418 Pointers in the implementation are 16@tie{}bits wide.
13419 The address of a function or label is represented as word address so
13420 that indirect jumps and calls can target any code address in the
13421 range of 64@tie{}Ki words.
13423 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13424 bytes of program memory space, there is a special function register called
13425 @code{EIND} that serves as most significant part of the target address
13426 when @code{EICALL} or @code{EIJMP} instructions are used.
13428 Indirect jumps and calls on these devices are handled as follows by
13429 the compiler and are subject to some limitations:
13434 The compiler never sets @code{EIND}.
13437 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13438 instructions or might read @code{EIND} directly in order to emulate an
13439 indirect call/jump by means of a @code{RET} instruction.
13442 The compiler assumes that @code{EIND} never changes during the startup
13443 code or during the application. In particular, @code{EIND} is not
13444 saved/restored in function or interrupt service routine
13448 For indirect calls to functions and computed goto, the linker
13449 generates @emph{stubs}. Stubs are jump pads sometimes also called
13450 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13451 The stub contains a direct jump to the desired address.
13454 Linker relaxation must be turned on so that the linker generates
13455 the stubs correctly in all situations. See the compiler option
13456 @option{-mrelax} and the linker option @option{--relax}.
13457 There are corner cases where the linker is supposed to generate stubs
13458 but aborts without relaxation and without a helpful error message.
13461 The default linker script is arranged for code with @code{EIND = 0}.
13462 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13463 linker script has to be used in order to place the sections whose
13464 name start with @code{.trampolines} into the segment where @code{EIND}
13468 The startup code from libgcc never sets @code{EIND}.
13469 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13470 For the impact of AVR-LibC on @code{EIND}, see the
13471 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13474 It is legitimate for user-specific startup code to set up @code{EIND}
13475 early, for example by means of initialization code located in
13476 section @code{.init3}. Such code runs prior to general startup code
13477 that initializes RAM and calls constructors, but after the bit
13478 of startup code from AVR-LibC that sets @code{EIND} to the segment
13479 where the vector table is located.
13481 #include <avr/io.h>
13484 __attribute__((section(".init3"),naked,used,no_instrument_function))
13485 init3_set_eind (void)
13487 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13488 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13493 The @code{__trampolines_start} symbol is defined in the linker script.
13496 Stubs are generated automatically by the linker if
13497 the following two conditions are met:
13500 @item The address of a label is taken by means of the @code{gs} modifier
13501 (short for @emph{generate stubs}) like so:
13503 LDI r24, lo8(gs(@var{func}))
13504 LDI r25, hi8(gs(@var{func}))
13506 @item The final location of that label is in a code segment
13507 @emph{outside} the segment where the stubs are located.
13511 The compiler emits such @code{gs} modifiers for code labels in the
13512 following situations:
13514 @item Taking address of a function or code label.
13515 @item Computed goto.
13516 @item If prologue-save function is used, see @option{-mcall-prologues}
13517 command-line option.
13518 @item Switch/case dispatch tables. If you do not want such dispatch
13519 tables you can specify the @option{-fno-jump-tables} command-line option.
13520 @item C and C++ constructors/destructors called during startup/shutdown.
13521 @item If the tools hit a @code{gs()} modifier explained above.
13525 Jumping to non-symbolic addresses like so is @emph{not} supported:
13530 /* Call function at word address 0x2 */
13531 return ((int(*)(void)) 0x2)();
13535 Instead, a stub has to be set up, i.e.@: the function has to be called
13536 through a symbol (@code{func_4} in the example):
13541 extern int func_4 (void);
13543 /* Call function at byte address 0x4 */
13548 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13549 Alternatively, @code{func_4} can be defined in the linker script.
13552 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13553 @cindex @code{RAMPD}
13554 @cindex @code{RAMPX}
13555 @cindex @code{RAMPY}
13556 @cindex @code{RAMPZ}
13557 Some AVR devices support memories larger than the 64@tie{}KiB range
13558 that can be accessed with 16-bit pointers. To access memory locations
13559 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13560 register is used as high part of the address:
13561 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13562 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13563 register, respectively, to get a wide address. Similarly,
13564 @code{RAMPD} is used together with direct addressing.
13568 The startup code initializes the @code{RAMP} special function
13569 registers with zero.
13572 If a @ref{AVR Named Address Spaces,named address space} other than
13573 generic or @code{__flash} is used, then @code{RAMPZ} is set
13574 as needed before the operation.
13577 If the device supports RAM larger than 64@tie{}KiB and the compiler
13578 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13579 is reset to zero after the operation.
13582 If the device comes with a specific @code{RAMP} register, the ISR
13583 prologue/epilogue saves/restores that SFR and initializes it with
13584 zero in case the ISR code might (implicitly) use it.
13587 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13588 If you use inline assembler to read from locations outside the
13589 16-bit address range and change one of the @code{RAMP} registers,
13590 you must reset it to zero after the access.
13594 @subsubsection AVR Built-in Macros
13596 GCC defines several built-in macros so that the user code can test
13597 for the presence or absence of features. Almost any of the following
13598 built-in macros are deduced from device capabilities and thus
13599 triggered by the @option{-mmcu=} command-line option.
13601 For even more AVR-specific built-in macros see
13602 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13607 Build-in macro that resolves to a decimal number that identifies the
13608 architecture and depends on the @option{-mmcu=@var{mcu}} option.
13609 Possible values are:
13611 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13612 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
13613 @code{105}, @code{106}, @code{107}
13615 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
13616 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
13617 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
13618 @code{avrxmega6}, @code{avrxmega7}, respectively.
13619 If @var{mcu} specifies a device, this built-in macro is set
13620 accordingly. For example, with @option{-mmcu=atmega8} the macro is
13621 defined to @code{4}.
13623 @item __AVR_@var{Device}__
13624 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
13625 the device's name. For example, @option{-mmcu=atmega8} defines the
13626 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
13627 @code{__AVR_ATtiny261A__}, etc.
13629 The built-in macros' names follow
13630 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13631 the device name as from the AVR user manual. The difference between
13632 @var{Device} in the built-in macro and @var{device} in
13633 @option{-mmcu=@var{device}} is that the latter is always lowercase.
13635 If @var{device} is not a device but only a core architecture like
13636 @samp{avr51}, this macro is not defined.
13638 @item __AVR_DEVICE_NAME__
13639 Setting @option{-mmcu=@var{device}} defines this built-in macro to
13640 the device's name. For example, with @option{-mmcu=atmega8} the macro
13641 is defined to @code{atmega8}.
13643 If @var{device} is not a device but only a core architecture like
13644 @samp{avr51}, this macro is not defined.
13646 @item __AVR_XMEGA__
13647 The device / architecture belongs to the XMEGA family of devices.
13649 @item __AVR_HAVE_ELPM__
13650 The device has the the @code{ELPM} instruction.
13652 @item __AVR_HAVE_ELPMX__
13653 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13654 R@var{n},Z+} instructions.
13656 @item __AVR_HAVE_MOVW__
13657 The device has the @code{MOVW} instruction to perform 16-bit
13658 register-register moves.
13660 @item __AVR_HAVE_LPMX__
13661 The device has the @code{LPM R@var{n},Z} and
13662 @code{LPM R@var{n},Z+} instructions.
13664 @item __AVR_HAVE_MUL__
13665 The device has a hardware multiplier.
13667 @item __AVR_HAVE_JMP_CALL__
13668 The device has the @code{JMP} and @code{CALL} instructions.
13669 This is the case for devices with at least 16@tie{}KiB of program
13672 @item __AVR_HAVE_EIJMP_EICALL__
13673 @itemx __AVR_3_BYTE_PC__
13674 The device has the @code{EIJMP} and @code{EICALL} instructions.
13675 This is the case for devices with more than 128@tie{}KiB of program memory.
13676 This also means that the program counter
13677 (PC) is 3@tie{}bytes wide.
13679 @item __AVR_2_BYTE_PC__
13680 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13681 with up to 128@tie{}KiB of program memory.
13683 @item __AVR_HAVE_8BIT_SP__
13684 @itemx __AVR_HAVE_16BIT_SP__
13685 The stack pointer (SP) register is treated as 8-bit respectively
13686 16-bit register by the compiler.
13687 The definition of these macros is affected by @option{-mtiny-stack}.
13689 @item __AVR_HAVE_SPH__
13691 The device has the SPH (high part of stack pointer) special function
13692 register or has an 8-bit stack pointer, respectively.
13693 The definition of these macros is affected by @option{-mmcu=} and
13694 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
13697 @item __AVR_HAVE_RAMPD__
13698 @itemx __AVR_HAVE_RAMPX__
13699 @itemx __AVR_HAVE_RAMPY__
13700 @itemx __AVR_HAVE_RAMPZ__
13701 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13702 @code{RAMPZ} special function register, respectively.
13704 @item __NO_INTERRUPTS__
13705 This macro reflects the @option{-mno-interrupts} command line option.
13707 @item __AVR_ERRATA_SKIP__
13708 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13709 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13710 instructions because of a hardware erratum. Skip instructions are
13711 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13712 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13715 @item __AVR_ISA_RMW__
13716 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13718 @item __AVR_SFR_OFFSET__=@var{offset}
13719 Instructions that can address I/O special function registers directly
13720 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13721 address as if addressed by an instruction to access RAM like @code{LD}
13722 or @code{STS}. This offset depends on the device architecture and has
13723 to be subtracted from the RAM address in order to get the
13724 respective I/O@tie{}address.
13726 @item __WITH_AVRLIBC__
13727 The compiler is configured to be used together with AVR-Libc.
13728 See the @option{--with-avrlibc} configure option.
13732 @node Blackfin Options
13733 @subsection Blackfin Options
13734 @cindex Blackfin Options
13737 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13739 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13740 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13741 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13742 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
13743 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
13744 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
13745 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
13746 @samp{bf561}, @samp{bf592}.
13748 The optional @var{sirevision} specifies the silicon revision of the target
13749 Blackfin processor. Any workarounds available for the targeted silicon revision
13750 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
13751 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
13752 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
13753 hexadecimal digits representing the major and minor numbers in the silicon
13754 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
13755 is not defined. If @var{sirevision} is @samp{any}, the
13756 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
13757 If this optional @var{sirevision} is not used, GCC assumes the latest known
13758 silicon revision of the targeted Blackfin processor.
13760 GCC defines a preprocessor macro for the specified @var{cpu}.
13761 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
13762 provided by libgloss to be linked in if @option{-msim} is not given.
13764 Without this option, @samp{bf532} is used as the processor by default.
13766 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
13767 only the preprocessor macro is defined.
13771 Specifies that the program will be run on the simulator. This causes
13772 the simulator BSP provided by libgloss to be linked in. This option
13773 has effect only for @samp{bfin-elf} toolchain.
13774 Certain other options, such as @option{-mid-shared-library} and
13775 @option{-mfdpic}, imply @option{-msim}.
13777 @item -momit-leaf-frame-pointer
13778 @opindex momit-leaf-frame-pointer
13779 Don't keep the frame pointer in a register for leaf functions. This
13780 avoids the instructions to save, set up and restore frame pointers and
13781 makes an extra register available in leaf functions. The option
13782 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13783 which might make debugging harder.
13785 @item -mspecld-anomaly
13786 @opindex mspecld-anomaly
13787 When enabled, the compiler ensures that the generated code does not
13788 contain speculative loads after jump instructions. If this option is used,
13789 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
13791 @item -mno-specld-anomaly
13792 @opindex mno-specld-anomaly
13793 Don't generate extra code to prevent speculative loads from occurring.
13795 @item -mcsync-anomaly
13796 @opindex mcsync-anomaly
13797 When enabled, the compiler ensures that the generated code does not
13798 contain CSYNC or SSYNC instructions too soon after conditional branches.
13799 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
13801 @item -mno-csync-anomaly
13802 @opindex mno-csync-anomaly
13803 Don't generate extra code to prevent CSYNC or SSYNC instructions from
13804 occurring too soon after a conditional branch.
13808 When enabled, the compiler is free to take advantage of the knowledge that
13809 the entire program fits into the low 64k of memory.
13812 @opindex mno-low-64k
13813 Assume that the program is arbitrarily large. This is the default.
13815 @item -mstack-check-l1
13816 @opindex mstack-check-l1
13817 Do stack checking using information placed into L1 scratchpad memory by the
13820 @item -mid-shared-library
13821 @opindex mid-shared-library
13822 Generate code that supports shared libraries via the library ID method.
13823 This allows for execute in place and shared libraries in an environment
13824 without virtual memory management. This option implies @option{-fPIC}.
13825 With a @samp{bfin-elf} target, this option implies @option{-msim}.
13827 @item -mno-id-shared-library
13828 @opindex mno-id-shared-library
13829 Generate code that doesn't assume ID-based shared libraries are being used.
13830 This is the default.
13832 @item -mleaf-id-shared-library
13833 @opindex mleaf-id-shared-library
13834 Generate code that supports shared libraries via the library ID method,
13835 but assumes that this library or executable won't link against any other
13836 ID shared libraries. That allows the compiler to use faster code for jumps
13839 @item -mno-leaf-id-shared-library
13840 @opindex mno-leaf-id-shared-library
13841 Do not assume that the code being compiled won't link against any ID shared
13842 libraries. Slower code is generated for jump and call insns.
13844 @item -mshared-library-id=n
13845 @opindex mshared-library-id
13846 Specifies the identification number of the ID-based shared library being
13847 compiled. Specifying a value of 0 generates more compact code; specifying
13848 other values forces the allocation of that number to the current
13849 library but is no more space- or time-efficient than omitting this option.
13853 Generate code that allows the data segment to be located in a different
13854 area of memory from the text segment. This allows for execute in place in
13855 an environment without virtual memory management by eliminating relocations
13856 against the text section.
13858 @item -mno-sep-data
13859 @opindex mno-sep-data
13860 Generate code that assumes that the data segment follows the text segment.
13861 This is the default.
13864 @itemx -mno-long-calls
13865 @opindex mlong-calls
13866 @opindex mno-long-calls
13867 Tells the compiler to perform function calls by first loading the
13868 address of the function into a register and then performing a subroutine
13869 call on this register. This switch is needed if the target function
13870 lies outside of the 24-bit addressing range of the offset-based
13871 version of subroutine call instruction.
13873 This feature is not enabled by default. Specifying
13874 @option{-mno-long-calls} restores the default behavior. Note these
13875 switches have no effect on how the compiler generates code to handle
13876 function calls via function pointers.
13880 Link with the fast floating-point library. This library relaxes some of
13881 the IEEE floating-point standard's rules for checking inputs against
13882 Not-a-Number (NAN), in the interest of performance.
13885 @opindex minline-plt
13886 Enable inlining of PLT entries in function calls to functions that are
13887 not known to bind locally. It has no effect without @option{-mfdpic}.
13890 @opindex mmulticore
13891 Build a standalone application for multicore Blackfin processors.
13892 This option causes proper start files and link scripts supporting
13893 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
13894 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
13896 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
13897 selects the one-application-per-core programming model. Without
13898 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
13899 programming model is used. In this model, the main function of Core B
13900 should be named as @code{coreb_main}.
13902 If this option is not used, the single-core application programming
13907 Build a standalone application for Core A of BF561 when using
13908 the one-application-per-core programming model. Proper start files
13909 and link scripts are used to support Core A, and the macro
13910 @code{__BFIN_COREA} is defined.
13911 This option can only be used in conjunction with @option{-mmulticore}.
13915 Build a standalone application for Core B of BF561 when using
13916 the one-application-per-core programming model. Proper start files
13917 and link scripts are used to support Core B, and the macro
13918 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
13919 should be used instead of @code{main}.
13920 This option can only be used in conjunction with @option{-mmulticore}.
13924 Build a standalone application for SDRAM. Proper start files and
13925 link scripts are used to put the application into SDRAM, and the macro
13926 @code{__BFIN_SDRAM} is defined.
13927 The loader should initialize SDRAM before loading the application.
13931 Assume that ICPLBs are enabled at run time. This has an effect on certain
13932 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
13933 are enabled; for standalone applications the default is off.
13937 @subsection C6X Options
13938 @cindex C6X Options
13941 @item -march=@var{name}
13943 This specifies the name of the target architecture. GCC uses this
13944 name to determine what kind of instructions it can emit when generating
13945 assembly code. Permissible names are: @samp{c62x},
13946 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
13949 @opindex mbig-endian
13950 Generate code for a big-endian target.
13952 @item -mlittle-endian
13953 @opindex mlittle-endian
13954 Generate code for a little-endian target. This is the default.
13958 Choose startup files and linker script suitable for the simulator.
13960 @item -msdata=default
13961 @opindex msdata=default
13962 Put small global and static data in the @code{.neardata} section,
13963 which is pointed to by register @code{B14}. Put small uninitialized
13964 global and static data in the @code{.bss} section, which is adjacent
13965 to the @code{.neardata} section. Put small read-only data into the
13966 @code{.rodata} section. The corresponding sections used for large
13967 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
13970 @opindex msdata=all
13971 Put all data, not just small objects, into the sections reserved for
13972 small data, and use addressing relative to the @code{B14} register to
13976 @opindex msdata=none
13977 Make no use of the sections reserved for small data, and use absolute
13978 addresses to access all data. Put all initialized global and static
13979 data in the @code{.fardata} section, and all uninitialized data in the
13980 @code{.far} section. Put all constant data into the @code{.const}
13985 @subsection CRIS Options
13986 @cindex CRIS Options
13988 These options are defined specifically for the CRIS ports.
13991 @item -march=@var{architecture-type}
13992 @itemx -mcpu=@var{architecture-type}
13995 Generate code for the specified architecture. The choices for
13996 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
13997 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
13998 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14001 @item -mtune=@var{architecture-type}
14003 Tune to @var{architecture-type} everything applicable about the generated
14004 code, except for the ABI and the set of available instructions. The
14005 choices for @var{architecture-type} are the same as for
14006 @option{-march=@var{architecture-type}}.
14008 @item -mmax-stack-frame=@var{n}
14009 @opindex mmax-stack-frame
14010 Warn when the stack frame of a function exceeds @var{n} bytes.
14016 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14017 @option{-march=v3} and @option{-march=v8} respectively.
14019 @item -mmul-bug-workaround
14020 @itemx -mno-mul-bug-workaround
14021 @opindex mmul-bug-workaround
14022 @opindex mno-mul-bug-workaround
14023 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14024 models where it applies. This option is active by default.
14028 Enable CRIS-specific verbose debug-related information in the assembly
14029 code. This option also has the effect of turning off the @samp{#NO_APP}
14030 formatted-code indicator to the assembler at the beginning of the
14035 Do not use condition-code results from previous instruction; always emit
14036 compare and test instructions before use of condition codes.
14038 @item -mno-side-effects
14039 @opindex mno-side-effects
14040 Do not emit instructions with side effects in addressing modes other than
14043 @item -mstack-align
14044 @itemx -mno-stack-align
14045 @itemx -mdata-align
14046 @itemx -mno-data-align
14047 @itemx -mconst-align
14048 @itemx -mno-const-align
14049 @opindex mstack-align
14050 @opindex mno-stack-align
14051 @opindex mdata-align
14052 @opindex mno-data-align
14053 @opindex mconst-align
14054 @opindex mno-const-align
14055 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14056 stack frame, individual data and constants to be aligned for the maximum
14057 single data access size for the chosen CPU model. The default is to
14058 arrange for 32-bit alignment. ABI details such as structure layout are
14059 not affected by these options.
14067 Similar to the stack- data- and const-align options above, these options
14068 arrange for stack frame, writable data and constants to all be 32-bit,
14069 16-bit or 8-bit aligned. The default is 32-bit alignment.
14071 @item -mno-prologue-epilogue
14072 @itemx -mprologue-epilogue
14073 @opindex mno-prologue-epilogue
14074 @opindex mprologue-epilogue
14075 With @option{-mno-prologue-epilogue}, the normal function prologue and
14076 epilogue which set up the stack frame are omitted and no return
14077 instructions or return sequences are generated in the code. Use this
14078 option only together with visual inspection of the compiled code: no
14079 warnings or errors are generated when call-saved registers must be saved,
14080 or storage for local variables needs to be allocated.
14084 @opindex mno-gotplt
14086 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14087 instruction sequences that load addresses for functions from the PLT part
14088 of the GOT rather than (traditional on other architectures) calls to the
14089 PLT@. The default is @option{-mgotplt}.
14093 Legacy no-op option only recognized with the cris-axis-elf and
14094 cris-axis-linux-gnu targets.
14098 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14102 This option, recognized for the cris-axis-elf, arranges
14103 to link with input-output functions from a simulator library. Code,
14104 initialized data and zero-initialized data are allocated consecutively.
14108 Like @option{-sim}, but pass linker options to locate initialized data at
14109 0x40000000 and zero-initialized data at 0x80000000.
14113 @subsection CR16 Options
14114 @cindex CR16 Options
14116 These options are defined specifically for the CR16 ports.
14122 Enable the use of multiply-accumulate instructions. Disabled by default.
14126 @opindex mcr16cplus
14128 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14133 Links the library libsim.a which is in compatible with simulator. Applicable
14134 to ELF compiler only.
14138 Choose integer type as 32-bit wide.
14142 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14144 @item -mdata-model=@var{model}
14145 @opindex mdata-model
14146 Choose a data model. The choices for @var{model} are @samp{near},
14147 @samp{far} or @samp{medium}. @samp{medium} is default.
14148 However, @samp{far} is not valid with @option{-mcr16c}, as the
14149 CR16C architecture does not support the far data model.
14152 @node Darwin Options
14153 @subsection Darwin Options
14154 @cindex Darwin options
14156 These options are defined for all architectures running the Darwin operating
14159 FSF GCC on Darwin does not create ``fat'' object files; it creates
14160 an object file for the single architecture that GCC was built to
14161 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14162 @option{-arch} options are used; it does so by running the compiler or
14163 linker multiple times and joining the results together with
14166 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14167 @samp{i686}) is determined by the flags that specify the ISA
14168 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14169 @option{-force_cpusubtype_ALL} option can be used to override this.
14171 The Darwin tools vary in their behavior when presented with an ISA
14172 mismatch. The assembler, @file{as}, only permits instructions to
14173 be used that are valid for the subtype of the file it is generating,
14174 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14175 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14176 and prints an error if asked to create a shared library with a less
14177 restrictive subtype than its input files (for instance, trying to put
14178 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14179 for executables, @command{ld}, quietly gives the executable the most
14180 restrictive subtype of any of its input files.
14185 Add the framework directory @var{dir} to the head of the list of
14186 directories to be searched for header files. These directories are
14187 interleaved with those specified by @option{-I} options and are
14188 scanned in a left-to-right order.
14190 A framework directory is a directory with frameworks in it. A
14191 framework is a directory with a @file{Headers} and/or
14192 @file{PrivateHeaders} directory contained directly in it that ends
14193 in @file{.framework}. The name of a framework is the name of this
14194 directory excluding the @file{.framework}. Headers associated with
14195 the framework are found in one of those two directories, with
14196 @file{Headers} being searched first. A subframework is a framework
14197 directory that is in a framework's @file{Frameworks} directory.
14198 Includes of subframework headers can only appear in a header of a
14199 framework that contains the subframework, or in a sibling subframework
14200 header. Two subframeworks are siblings if they occur in the same
14201 framework. A subframework should not have the same name as a
14202 framework; a warning is issued if this is violated. Currently a
14203 subframework cannot have subframeworks; in the future, the mechanism
14204 may be extended to support this. The standard frameworks can be found
14205 in @file{/System/Library/Frameworks} and
14206 @file{/Library/Frameworks}. An example include looks like
14207 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14208 the name of the framework and @file{header.h} is found in the
14209 @file{PrivateHeaders} or @file{Headers} directory.
14211 @item -iframework@var{dir}
14212 @opindex iframework
14213 Like @option{-F} except the directory is a treated as a system
14214 directory. The main difference between this @option{-iframework} and
14215 @option{-F} is that with @option{-iframework} the compiler does not
14216 warn about constructs contained within header files found via
14217 @var{dir}. This option is valid only for the C family of languages.
14221 Emit debugging information for symbols that are used. For stabs
14222 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14223 This is by default ON@.
14227 Emit debugging information for all symbols and types.
14229 @item -mmacosx-version-min=@var{version}
14230 The earliest version of MacOS X that this executable will run on
14231 is @var{version}. Typical values of @var{version} include @code{10.1},
14232 @code{10.2}, and @code{10.3.9}.
14234 If the compiler was built to use the system's headers by default,
14235 then the default for this option is the system version on which the
14236 compiler is running, otherwise the default is to make choices that
14237 are compatible with as many systems and code bases as possible.
14241 Enable kernel development mode. The @option{-mkernel} option sets
14242 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14243 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14244 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14245 applicable. This mode also sets @option{-mno-altivec},
14246 @option{-msoft-float}, @option{-fno-builtin} and
14247 @option{-mlong-branch} for PowerPC targets.
14249 @item -mone-byte-bool
14250 @opindex mone-byte-bool
14251 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14252 By default @code{sizeof(bool)} is @code{4} when compiling for
14253 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14254 option has no effect on x86.
14256 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14257 to generate code that is not binary compatible with code generated
14258 without that switch. Using this switch may require recompiling all
14259 other modules in a program, including system libraries. Use this
14260 switch to conform to a non-default data model.
14262 @item -mfix-and-continue
14263 @itemx -ffix-and-continue
14264 @itemx -findirect-data
14265 @opindex mfix-and-continue
14266 @opindex ffix-and-continue
14267 @opindex findirect-data
14268 Generate code suitable for fast turnaround development, such as to
14269 allow GDB to dynamically load @file{.o} files into already-running
14270 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14271 are provided for backwards compatibility.
14275 Loads all members of static archive libraries.
14276 See man ld(1) for more information.
14278 @item -arch_errors_fatal
14279 @opindex arch_errors_fatal
14280 Cause the errors having to do with files that have the wrong architecture
14283 @item -bind_at_load
14284 @opindex bind_at_load
14285 Causes the output file to be marked such that the dynamic linker will
14286 bind all undefined references when the file is loaded or launched.
14290 Produce a Mach-o bundle format file.
14291 See man ld(1) for more information.
14293 @item -bundle_loader @var{executable}
14294 @opindex bundle_loader
14295 This option specifies the @var{executable} that will load the build
14296 output file being linked. See man ld(1) for more information.
14299 @opindex dynamiclib
14300 When passed this option, GCC produces a dynamic library instead of
14301 an executable when linking, using the Darwin @file{libtool} command.
14303 @item -force_cpusubtype_ALL
14304 @opindex force_cpusubtype_ALL
14305 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14306 one controlled by the @option{-mcpu} or @option{-march} option.
14308 @item -allowable_client @var{client_name}
14309 @itemx -client_name
14310 @itemx -compatibility_version
14311 @itemx -current_version
14313 @itemx -dependency-file
14315 @itemx -dylinker_install_name
14317 @itemx -exported_symbols_list
14320 @itemx -flat_namespace
14321 @itemx -force_flat_namespace
14322 @itemx -headerpad_max_install_names
14325 @itemx -install_name
14326 @itemx -keep_private_externs
14327 @itemx -multi_module
14328 @itemx -multiply_defined
14329 @itemx -multiply_defined_unused
14332 @itemx -no_dead_strip_inits_and_terms
14333 @itemx -nofixprebinding
14334 @itemx -nomultidefs
14336 @itemx -noseglinkedit
14337 @itemx -pagezero_size
14339 @itemx -prebind_all_twolevel_modules
14340 @itemx -private_bundle
14342 @itemx -read_only_relocs
14344 @itemx -sectobjectsymbols
14348 @itemx -sectobjectsymbols
14351 @itemx -segs_read_only_addr
14353 @itemx -segs_read_write_addr
14354 @itemx -seg_addr_table
14355 @itemx -seg_addr_table_filename
14356 @itemx -seglinkedit
14358 @itemx -segs_read_only_addr
14359 @itemx -segs_read_write_addr
14360 @itemx -single_module
14362 @itemx -sub_library
14364 @itemx -sub_umbrella
14365 @itemx -twolevel_namespace
14368 @itemx -unexported_symbols_list
14369 @itemx -weak_reference_mismatches
14370 @itemx -whatsloaded
14371 @opindex allowable_client
14372 @opindex client_name
14373 @opindex compatibility_version
14374 @opindex current_version
14375 @opindex dead_strip
14376 @opindex dependency-file
14377 @opindex dylib_file
14378 @opindex dylinker_install_name
14380 @opindex exported_symbols_list
14382 @opindex flat_namespace
14383 @opindex force_flat_namespace
14384 @opindex headerpad_max_install_names
14385 @opindex image_base
14387 @opindex install_name
14388 @opindex keep_private_externs
14389 @opindex multi_module
14390 @opindex multiply_defined
14391 @opindex multiply_defined_unused
14392 @opindex noall_load
14393 @opindex no_dead_strip_inits_and_terms
14394 @opindex nofixprebinding
14395 @opindex nomultidefs
14397 @opindex noseglinkedit
14398 @opindex pagezero_size
14400 @opindex prebind_all_twolevel_modules
14401 @opindex private_bundle
14402 @opindex read_only_relocs
14404 @opindex sectobjectsymbols
14407 @opindex sectcreate
14408 @opindex sectobjectsymbols
14411 @opindex segs_read_only_addr
14412 @opindex segs_read_write_addr
14413 @opindex seg_addr_table
14414 @opindex seg_addr_table_filename
14415 @opindex seglinkedit
14417 @opindex segs_read_only_addr
14418 @opindex segs_read_write_addr
14419 @opindex single_module
14421 @opindex sub_library
14422 @opindex sub_umbrella
14423 @opindex twolevel_namespace
14426 @opindex unexported_symbols_list
14427 @opindex weak_reference_mismatches
14428 @opindex whatsloaded
14429 These options are passed to the Darwin linker. The Darwin linker man page
14430 describes them in detail.
14433 @node DEC Alpha Options
14434 @subsection DEC Alpha Options
14436 These @samp{-m} options are defined for the DEC Alpha implementations:
14439 @item -mno-soft-float
14440 @itemx -msoft-float
14441 @opindex mno-soft-float
14442 @opindex msoft-float
14443 Use (do not use) the hardware floating-point instructions for
14444 floating-point operations. When @option{-msoft-float} is specified,
14445 functions in @file{libgcc.a} are used to perform floating-point
14446 operations. Unless they are replaced by routines that emulate the
14447 floating-point operations, or compiled in such a way as to call such
14448 emulations routines, these routines issue floating-point
14449 operations. If you are compiling for an Alpha without floating-point
14450 operations, you must ensure that the library is built so as not to call
14453 Note that Alpha implementations without floating-point operations are
14454 required to have floating-point registers.
14457 @itemx -mno-fp-regs
14459 @opindex mno-fp-regs
14460 Generate code that uses (does not use) the floating-point register set.
14461 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14462 register set is not used, floating-point operands are passed in integer
14463 registers as if they were integers and floating-point results are passed
14464 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14465 so any function with a floating-point argument or return value called by code
14466 compiled with @option{-mno-fp-regs} must also be compiled with that
14469 A typical use of this option is building a kernel that does not use,
14470 and hence need not save and restore, any floating-point registers.
14474 The Alpha architecture implements floating-point hardware optimized for
14475 maximum performance. It is mostly compliant with the IEEE floating-point
14476 standard. However, for full compliance, software assistance is
14477 required. This option generates code fully IEEE-compliant code
14478 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14479 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14480 defined during compilation. The resulting code is less efficient but is
14481 able to correctly support denormalized numbers and exceptional IEEE
14482 values such as not-a-number and plus/minus infinity. Other Alpha
14483 compilers call this option @option{-ieee_with_no_inexact}.
14485 @item -mieee-with-inexact
14486 @opindex mieee-with-inexact
14487 This is like @option{-mieee} except the generated code also maintains
14488 the IEEE @var{inexact-flag}. Turning on this option causes the
14489 generated code to implement fully-compliant IEEE math. In addition to
14490 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14491 macro. On some Alpha implementations the resulting code may execute
14492 significantly slower than the code generated by default. Since there is
14493 very little code that depends on the @var{inexact-flag}, you should
14494 normally not specify this option. Other Alpha compilers call this
14495 option @option{-ieee_with_inexact}.
14497 @item -mfp-trap-mode=@var{trap-mode}
14498 @opindex mfp-trap-mode
14499 This option controls what floating-point related traps are enabled.
14500 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14501 The trap mode can be set to one of four values:
14505 This is the default (normal) setting. The only traps that are enabled
14506 are the ones that cannot be disabled in software (e.g., division by zero
14510 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14514 Like @samp{u}, but the instructions are marked to be safe for software
14515 completion (see Alpha architecture manual for details).
14518 Like @samp{su}, but inexact traps are enabled as well.
14521 @item -mfp-rounding-mode=@var{rounding-mode}
14522 @opindex mfp-rounding-mode
14523 Selects the IEEE rounding mode. Other Alpha compilers call this option
14524 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14529 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14530 the nearest machine number or towards the even machine number in case
14534 Round towards minus infinity.
14537 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14540 Dynamic rounding mode. A field in the floating-point control register
14541 (@var{fpcr}, see Alpha architecture reference manual) controls the
14542 rounding mode in effect. The C library initializes this register for
14543 rounding towards plus infinity. Thus, unless your program modifies the
14544 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14547 @item -mtrap-precision=@var{trap-precision}
14548 @opindex mtrap-precision
14549 In the Alpha architecture, floating-point traps are imprecise. This
14550 means without software assistance it is impossible to recover from a
14551 floating trap and program execution normally needs to be terminated.
14552 GCC can generate code that can assist operating system trap handlers
14553 in determining the exact location that caused a floating-point trap.
14554 Depending on the requirements of an application, different levels of
14555 precisions can be selected:
14559 Program precision. This option is the default and means a trap handler
14560 can only identify which program caused a floating-point exception.
14563 Function precision. The trap handler can determine the function that
14564 caused a floating-point exception.
14567 Instruction precision. The trap handler can determine the exact
14568 instruction that caused a floating-point exception.
14571 Other Alpha compilers provide the equivalent options called
14572 @option{-scope_safe} and @option{-resumption_safe}.
14574 @item -mieee-conformant
14575 @opindex mieee-conformant
14576 This option marks the generated code as IEEE conformant. You must not
14577 use this option unless you also specify @option{-mtrap-precision=i} and either
14578 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14579 is to emit the line @samp{.eflag 48} in the function prologue of the
14580 generated assembly file.
14582 @item -mbuild-constants
14583 @opindex mbuild-constants
14584 Normally GCC examines a 32- or 64-bit integer constant to
14585 see if it can construct it from smaller constants in two or three
14586 instructions. If it cannot, it outputs the constant as a literal and
14587 generates code to load it from the data segment at run time.
14589 Use this option to require GCC to construct @emph{all} integer constants
14590 using code, even if it takes more instructions (the maximum is six).
14592 You typically use this option to build a shared library dynamic
14593 loader. Itself a shared library, it must relocate itself in memory
14594 before it can find the variables and constants in its own data segment.
14612 Indicate whether GCC should generate code to use the optional BWX,
14613 CIX, FIX and MAX instruction sets. The default is to use the instruction
14614 sets supported by the CPU type specified via @option{-mcpu=} option or that
14615 of the CPU on which GCC was built if none is specified.
14618 @itemx -mfloat-ieee
14619 @opindex mfloat-vax
14620 @opindex mfloat-ieee
14621 Generate code that uses (does not use) VAX F and G floating-point
14622 arithmetic instead of IEEE single and double precision.
14624 @item -mexplicit-relocs
14625 @itemx -mno-explicit-relocs
14626 @opindex mexplicit-relocs
14627 @opindex mno-explicit-relocs
14628 Older Alpha assemblers provided no way to generate symbol relocations
14629 except via assembler macros. Use of these macros does not allow
14630 optimal instruction scheduling. GNU binutils as of version 2.12
14631 supports a new syntax that allows the compiler to explicitly mark
14632 which relocations should apply to which instructions. This option
14633 is mostly useful for debugging, as GCC detects the capabilities of
14634 the assembler when it is built and sets the default accordingly.
14637 @itemx -mlarge-data
14638 @opindex msmall-data
14639 @opindex mlarge-data
14640 When @option{-mexplicit-relocs} is in effect, static data is
14641 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14642 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14643 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14644 16-bit relocations off of the @code{$gp} register. This limits the
14645 size of the small data area to 64KB, but allows the variables to be
14646 directly accessed via a single instruction.
14648 The default is @option{-mlarge-data}. With this option the data area
14649 is limited to just below 2GB@. Programs that require more than 2GB of
14650 data must use @code{malloc} or @code{mmap} to allocate the data in the
14651 heap instead of in the program's data segment.
14653 When generating code for shared libraries, @option{-fpic} implies
14654 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14657 @itemx -mlarge-text
14658 @opindex msmall-text
14659 @opindex mlarge-text
14660 When @option{-msmall-text} is used, the compiler assumes that the
14661 code of the entire program (or shared library) fits in 4MB, and is
14662 thus reachable with a branch instruction. When @option{-msmall-data}
14663 is used, the compiler can assume that all local symbols share the
14664 same @code{$gp} value, and thus reduce the number of instructions
14665 required for a function call from 4 to 1.
14667 The default is @option{-mlarge-text}.
14669 @item -mcpu=@var{cpu_type}
14671 Set the instruction set and instruction scheduling parameters for
14672 machine type @var{cpu_type}. You can specify either the @samp{EV}
14673 style name or the corresponding chip number. GCC supports scheduling
14674 parameters for the EV4, EV5 and EV6 family of processors and
14675 chooses the default values for the instruction set from the processor
14676 you specify. If you do not specify a processor type, GCC defaults
14677 to the processor on which the compiler was built.
14679 Supported values for @var{cpu_type} are
14685 Schedules as an EV4 and has no instruction set extensions.
14689 Schedules as an EV5 and has no instruction set extensions.
14693 Schedules as an EV5 and supports the BWX extension.
14698 Schedules as an EV5 and supports the BWX and MAX extensions.
14702 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14706 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14709 Native toolchains also support the value @samp{native},
14710 which selects the best architecture option for the host processor.
14711 @option{-mcpu=native} has no effect if GCC does not recognize
14714 @item -mtune=@var{cpu_type}
14716 Set only the instruction scheduling parameters for machine type
14717 @var{cpu_type}. The instruction set is not changed.
14719 Native toolchains also support the value @samp{native},
14720 which selects the best architecture option for the host processor.
14721 @option{-mtune=native} has no effect if GCC does not recognize
14724 @item -mmemory-latency=@var{time}
14725 @opindex mmemory-latency
14726 Sets the latency the scheduler should assume for typical memory
14727 references as seen by the application. This number is highly
14728 dependent on the memory access patterns used by the application
14729 and the size of the external cache on the machine.
14731 Valid options for @var{time} are
14735 A decimal number representing clock cycles.
14741 The compiler contains estimates of the number of clock cycles for
14742 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
14743 (also called Dcache, Scache, and Bcache), as well as to main memory.
14744 Note that L3 is only valid for EV5.
14750 @subsection FR30 Options
14751 @cindex FR30 Options
14753 These options are defined specifically for the FR30 port.
14757 @item -msmall-model
14758 @opindex msmall-model
14759 Use the small address space model. This can produce smaller code, but
14760 it does assume that all symbolic values and addresses fit into a
14765 Assume that runtime support has been provided and so there is no need
14766 to include the simulator library (@file{libsim.a}) on the linker
14772 @subsection FRV Options
14773 @cindex FRV Options
14779 Only use the first 32 general-purpose registers.
14784 Use all 64 general-purpose registers.
14789 Use only the first 32 floating-point registers.
14794 Use all 64 floating-point registers.
14797 @opindex mhard-float
14799 Use hardware instructions for floating-point operations.
14802 @opindex msoft-float
14804 Use library routines for floating-point operations.
14809 Dynamically allocate condition code registers.
14814 Do not try to dynamically allocate condition code registers, only
14815 use @code{icc0} and @code{fcc0}.
14820 Change ABI to use double word insns.
14825 Do not use double word instructions.
14830 Use floating-point double instructions.
14833 @opindex mno-double
14835 Do not use floating-point double instructions.
14840 Use media instructions.
14845 Do not use media instructions.
14850 Use multiply and add/subtract instructions.
14853 @opindex mno-muladd
14855 Do not use multiply and add/subtract instructions.
14860 Select the FDPIC ABI, which uses function descriptors to represent
14861 pointers to functions. Without any PIC/PIE-related options, it
14862 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
14863 assumes GOT entries and small data are within a 12-bit range from the
14864 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
14865 are computed with 32 bits.
14866 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14869 @opindex minline-plt
14871 Enable inlining of PLT entries in function calls to functions that are
14872 not known to bind locally. It has no effect without @option{-mfdpic}.
14873 It's enabled by default if optimizing for speed and compiling for
14874 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
14875 optimization option such as @option{-O3} or above is present in the
14881 Assume a large TLS segment when generating thread-local code.
14886 Do not assume a large TLS segment when generating thread-local code.
14891 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
14892 that is known to be in read-only sections. It's enabled by default,
14893 except for @option{-fpic} or @option{-fpie}: even though it may help
14894 make the global offset table smaller, it trades 1 instruction for 4.
14895 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
14896 one of which may be shared by multiple symbols, and it avoids the need
14897 for a GOT entry for the referenced symbol, so it's more likely to be a
14898 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
14900 @item -multilib-library-pic
14901 @opindex multilib-library-pic
14903 Link with the (library, not FD) pic libraries. It's implied by
14904 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
14905 @option{-fpic} without @option{-mfdpic}. You should never have to use
14909 @opindex mlinked-fp
14911 Follow the EABI requirement of always creating a frame pointer whenever
14912 a stack frame is allocated. This option is enabled by default and can
14913 be disabled with @option{-mno-linked-fp}.
14916 @opindex mlong-calls
14918 Use indirect addressing to call functions outside the current
14919 compilation unit. This allows the functions to be placed anywhere
14920 within the 32-bit address space.
14922 @item -malign-labels
14923 @opindex malign-labels
14925 Try to align labels to an 8-byte boundary by inserting NOPs into the
14926 previous packet. This option only has an effect when VLIW packing
14927 is enabled. It doesn't create new packets; it merely adds NOPs to
14930 @item -mlibrary-pic
14931 @opindex mlibrary-pic
14933 Generate position-independent EABI code.
14938 Use only the first four media accumulator registers.
14943 Use all eight media accumulator registers.
14948 Pack VLIW instructions.
14953 Do not pack VLIW instructions.
14956 @opindex mno-eflags
14958 Do not mark ABI switches in e_flags.
14961 @opindex mcond-move
14963 Enable the use of conditional-move instructions (default).
14965 This switch is mainly for debugging the compiler and will likely be removed
14966 in a future version.
14968 @item -mno-cond-move
14969 @opindex mno-cond-move
14971 Disable the use of conditional-move instructions.
14973 This switch is mainly for debugging the compiler and will likely be removed
14974 in a future version.
14979 Enable the use of conditional set instructions (default).
14981 This switch is mainly for debugging the compiler and will likely be removed
14982 in a future version.
14987 Disable the use of conditional set instructions.
14989 This switch is mainly for debugging the compiler and will likely be removed
14990 in a future version.
14993 @opindex mcond-exec
14995 Enable the use of conditional execution (default).
14997 This switch is mainly for debugging the compiler and will likely be removed
14998 in a future version.
15000 @item -mno-cond-exec
15001 @opindex mno-cond-exec
15003 Disable the use of conditional execution.
15005 This switch is mainly for debugging the compiler and will likely be removed
15006 in a future version.
15008 @item -mvliw-branch
15009 @opindex mvliw-branch
15011 Run a pass to pack branches into VLIW instructions (default).
15013 This switch is mainly for debugging the compiler and will likely be removed
15014 in a future version.
15016 @item -mno-vliw-branch
15017 @opindex mno-vliw-branch
15019 Do not run a pass to pack branches into VLIW instructions.
15021 This switch is mainly for debugging the compiler and will likely be removed
15022 in a future version.
15024 @item -mmulti-cond-exec
15025 @opindex mmulti-cond-exec
15027 Enable optimization of @code{&&} and @code{||} in conditional execution
15030 This switch is mainly for debugging the compiler and will likely be removed
15031 in a future version.
15033 @item -mno-multi-cond-exec
15034 @opindex mno-multi-cond-exec
15036 Disable optimization of @code{&&} and @code{||} in conditional execution.
15038 This switch is mainly for debugging the compiler and will likely be removed
15039 in a future version.
15041 @item -mnested-cond-exec
15042 @opindex mnested-cond-exec
15044 Enable nested conditional execution optimizations (default).
15046 This switch is mainly for debugging the compiler and will likely be removed
15047 in a future version.
15049 @item -mno-nested-cond-exec
15050 @opindex mno-nested-cond-exec
15052 Disable nested conditional execution optimizations.
15054 This switch is mainly for debugging the compiler and will likely be removed
15055 in a future version.
15057 @item -moptimize-membar
15058 @opindex moptimize-membar
15060 This switch removes redundant @code{membar} instructions from the
15061 compiler-generated code. It is enabled by default.
15063 @item -mno-optimize-membar
15064 @opindex mno-optimize-membar
15066 This switch disables the automatic removal of redundant @code{membar}
15067 instructions from the generated code.
15069 @item -mtomcat-stats
15070 @opindex mtomcat-stats
15072 Cause gas to print out tomcat statistics.
15074 @item -mcpu=@var{cpu}
15077 Select the processor type for which to generate code. Possible values are
15078 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15079 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15083 @node GNU/Linux Options
15084 @subsection GNU/Linux Options
15086 These @samp{-m} options are defined for GNU/Linux targets:
15091 Use the GNU C library. This is the default except
15092 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
15096 Use uClibc C library. This is the default on
15097 @samp{*-*-linux-*uclibc*} targets.
15101 Use Bionic C library. This is the default on
15102 @samp{*-*-linux-*android*} targets.
15106 Compile code compatible with Android platform. This is the default on
15107 @samp{*-*-linux-*android*} targets.
15109 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15110 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15111 this option makes the GCC driver pass Android-specific options to the linker.
15112 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15115 @item -tno-android-cc
15116 @opindex tno-android-cc
15117 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15118 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15119 @option{-fno-rtti} by default.
15121 @item -tno-android-ld
15122 @opindex tno-android-ld
15123 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15124 linking options to the linker.
15128 @node H8/300 Options
15129 @subsection H8/300 Options
15131 These @samp{-m} options are defined for the H8/300 implementations:
15136 Shorten some address references at link time, when possible; uses the
15137 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15138 ld, Using ld}, for a fuller description.
15142 Generate code for the H8/300H@.
15146 Generate code for the H8S@.
15150 Generate code for the H8S and H8/300H in the normal mode. This switch
15151 must be used either with @option{-mh} or @option{-ms}.
15155 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15159 Extended registers are stored on stack before execution of function
15160 with monitor attribute. Default option is @option{-mexr}.
15161 This option is valid only for H8S targets.
15165 Extended registers are not stored on stack before execution of function
15166 with monitor attribute. Default option is @option{-mno-exr}.
15167 This option is valid only for H8S targets.
15171 Make @code{int} data 32 bits by default.
15174 @opindex malign-300
15175 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15176 The default for the H8/300H and H8S is to align longs and floats on
15178 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15179 This option has no effect on the H8/300.
15183 @subsection HPPA Options
15184 @cindex HPPA Options
15186 These @samp{-m} options are defined for the HPPA family of computers:
15189 @item -march=@var{architecture-type}
15191 Generate code for the specified architecture. The choices for
15192 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15193 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15194 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15195 architecture option for your machine. Code compiled for lower numbered
15196 architectures runs on higher numbered architectures, but not the
15199 @item -mpa-risc-1-0
15200 @itemx -mpa-risc-1-1
15201 @itemx -mpa-risc-2-0
15202 @opindex mpa-risc-1-0
15203 @opindex mpa-risc-1-1
15204 @opindex mpa-risc-2-0
15205 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15207 @item -mjump-in-delay
15208 @opindex mjump-in-delay
15209 This option is ignored and provided for compatibility purposes only.
15211 @item -mdisable-fpregs
15212 @opindex mdisable-fpregs
15213 Prevent floating-point registers from being used in any manner. This is
15214 necessary for compiling kernels that perform lazy context switching of
15215 floating-point registers. If you use this option and attempt to perform
15216 floating-point operations, the compiler aborts.
15218 @item -mdisable-indexing
15219 @opindex mdisable-indexing
15220 Prevent the compiler from using indexing address modes. This avoids some
15221 rather obscure problems when compiling MIG generated code under MACH@.
15223 @item -mno-space-regs
15224 @opindex mno-space-regs
15225 Generate code that assumes the target has no space registers. This allows
15226 GCC to generate faster indirect calls and use unscaled index address modes.
15228 Such code is suitable for level 0 PA systems and kernels.
15230 @item -mfast-indirect-calls
15231 @opindex mfast-indirect-calls
15232 Generate code that assumes calls never cross space boundaries. This
15233 allows GCC to emit code that performs faster indirect calls.
15235 This option does not work in the presence of shared libraries or nested
15238 @item -mfixed-range=@var{register-range}
15239 @opindex mfixed-range
15240 Generate code treating the given register range as fixed registers.
15241 A fixed register is one that the register allocator cannot use. This is
15242 useful when compiling kernel code. A register range is specified as
15243 two registers separated by a dash. Multiple register ranges can be
15244 specified separated by a comma.
15246 @item -mlong-load-store
15247 @opindex mlong-load-store
15248 Generate 3-instruction load and store sequences as sometimes required by
15249 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15252 @item -mportable-runtime
15253 @opindex mportable-runtime
15254 Use the portable calling conventions proposed by HP for ELF systems.
15258 Enable the use of assembler directives only GAS understands.
15260 @item -mschedule=@var{cpu-type}
15262 Schedule code according to the constraints for the machine type
15263 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15264 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15265 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15266 proper scheduling option for your machine. The default scheduling is
15270 @opindex mlinker-opt
15271 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15272 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15273 linkers in which they give bogus error messages when linking some programs.
15276 @opindex msoft-float
15277 Generate output containing library calls for floating point.
15278 @strong{Warning:} the requisite libraries are not available for all HPPA
15279 targets. Normally the facilities of the machine's usual C compiler are
15280 used, but this cannot be done directly in cross-compilation. You must make
15281 your own arrangements to provide suitable library functions for
15284 @option{-msoft-float} changes the calling convention in the output file;
15285 therefore, it is only useful if you compile @emph{all} of a program with
15286 this option. In particular, you need to compile @file{libgcc.a}, the
15287 library that comes with GCC, with @option{-msoft-float} in order for
15292 Generate the predefine, @code{_SIO}, for server IO@. The default is
15293 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15294 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15295 options are available under HP-UX and HI-UX@.
15299 Use options specific to GNU @command{ld}.
15300 This passes @option{-shared} to @command{ld} when
15301 building a shared library. It is the default when GCC is configured,
15302 explicitly or implicitly, with the GNU linker. This option does not
15303 affect which @command{ld} is called; it only changes what parameters
15304 are passed to that @command{ld}.
15305 The @command{ld} that is called is determined by the
15306 @option{--with-ld} configure option, GCC's program search path, and
15307 finally by the user's @env{PATH}. The linker used by GCC can be printed
15308 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15309 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15313 Use options specific to HP @command{ld}.
15314 This passes @option{-b} to @command{ld} when building
15315 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15316 links. It is the default when GCC is configured, explicitly or
15317 implicitly, with the HP linker. This option does not affect
15318 which @command{ld} is called; it only changes what parameters are passed to that
15320 The @command{ld} that is called is determined by the @option{--with-ld}
15321 configure option, GCC's program search path, and finally by the user's
15322 @env{PATH}. The linker used by GCC can be printed using @samp{which
15323 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15324 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15327 @opindex mno-long-calls
15328 Generate code that uses long call sequences. This ensures that a call
15329 is always able to reach linker generated stubs. The default is to generate
15330 long calls only when the distance from the call site to the beginning
15331 of the function or translation unit, as the case may be, exceeds a
15332 predefined limit set by the branch type being used. The limits for
15333 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15334 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15337 Distances are measured from the beginning of functions when using the
15338 @option{-ffunction-sections} option, or when using the @option{-mgas}
15339 and @option{-mno-portable-runtime} options together under HP-UX with
15342 It is normally not desirable to use this option as it degrades
15343 performance. However, it may be useful in large applications,
15344 particularly when partial linking is used to build the application.
15346 The types of long calls used depends on the capabilities of the
15347 assembler and linker, and the type of code being generated. The
15348 impact on systems that support long absolute calls, and long pic
15349 symbol-difference or pc-relative calls should be relatively small.
15350 However, an indirect call is used on 32-bit ELF systems in pic code
15351 and it is quite long.
15353 @item -munix=@var{unix-std}
15355 Generate compiler predefines and select a startfile for the specified
15356 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15357 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15358 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15359 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15360 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15363 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15364 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15365 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15366 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15367 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15368 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15370 It is @emph{important} to note that this option changes the interfaces
15371 for various library routines. It also affects the operational behavior
15372 of the C library. Thus, @emph{extreme} care is needed in using this
15375 Library code that is intended to operate with more than one UNIX
15376 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15377 as appropriate. Most GNU software doesn't provide this capability.
15381 Suppress the generation of link options to search libdld.sl when the
15382 @option{-static} option is specified on HP-UX 10 and later.
15386 The HP-UX implementation of setlocale in libc has a dependency on
15387 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15388 when the @option{-static} option is specified, special link options
15389 are needed to resolve this dependency.
15391 On HP-UX 10 and later, the GCC driver adds the necessary options to
15392 link with libdld.sl when the @option{-static} option is specified.
15393 This causes the resulting binary to be dynamic. On the 64-bit port,
15394 the linkers generate dynamic binaries by default in any case. The
15395 @option{-nolibdld} option can be used to prevent the GCC driver from
15396 adding these link options.
15400 Add support for multithreading with the @dfn{dce thread} library
15401 under HP-UX@. This option sets flags for both the preprocessor and
15405 @node IA-64 Options
15406 @subsection IA-64 Options
15407 @cindex IA-64 Options
15409 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15413 @opindex mbig-endian
15414 Generate code for a big-endian target. This is the default for HP-UX@.
15416 @item -mlittle-endian
15417 @opindex mlittle-endian
15418 Generate code for a little-endian target. This is the default for AIX5
15424 @opindex mno-gnu-as
15425 Generate (or don't) code for the GNU assembler. This is the default.
15426 @c Also, this is the default if the configure option @option{--with-gnu-as}
15432 @opindex mno-gnu-ld
15433 Generate (or don't) code for the GNU linker. This is the default.
15434 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15439 Generate code that does not use a global pointer register. The result
15440 is not position independent code, and violates the IA-64 ABI@.
15442 @item -mvolatile-asm-stop
15443 @itemx -mno-volatile-asm-stop
15444 @opindex mvolatile-asm-stop
15445 @opindex mno-volatile-asm-stop
15446 Generate (or don't) a stop bit immediately before and after volatile asm
15449 @item -mregister-names
15450 @itemx -mno-register-names
15451 @opindex mregister-names
15452 @opindex mno-register-names
15453 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15454 the stacked registers. This may make assembler output more readable.
15460 Disable (or enable) optimizations that use the small data section. This may
15461 be useful for working around optimizer bugs.
15463 @item -mconstant-gp
15464 @opindex mconstant-gp
15465 Generate code that uses a single constant global pointer value. This is
15466 useful when compiling kernel code.
15470 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15471 This is useful when compiling firmware code.
15473 @item -minline-float-divide-min-latency
15474 @opindex minline-float-divide-min-latency
15475 Generate code for inline divides of floating-point values
15476 using the minimum latency algorithm.
15478 @item -minline-float-divide-max-throughput
15479 @opindex minline-float-divide-max-throughput
15480 Generate code for inline divides of floating-point values
15481 using the maximum throughput algorithm.
15483 @item -mno-inline-float-divide
15484 @opindex mno-inline-float-divide
15485 Do not generate inline code for divides of floating-point values.
15487 @item -minline-int-divide-min-latency
15488 @opindex minline-int-divide-min-latency
15489 Generate code for inline divides of integer values
15490 using the minimum latency algorithm.
15492 @item -minline-int-divide-max-throughput
15493 @opindex minline-int-divide-max-throughput
15494 Generate code for inline divides of integer values
15495 using the maximum throughput algorithm.
15497 @item -mno-inline-int-divide
15498 @opindex mno-inline-int-divide
15499 Do not generate inline code for divides of integer values.
15501 @item -minline-sqrt-min-latency
15502 @opindex minline-sqrt-min-latency
15503 Generate code for inline square roots
15504 using the minimum latency algorithm.
15506 @item -minline-sqrt-max-throughput
15507 @opindex minline-sqrt-max-throughput
15508 Generate code for inline square roots
15509 using the maximum throughput algorithm.
15511 @item -mno-inline-sqrt
15512 @opindex mno-inline-sqrt
15513 Do not generate inline code for @code{sqrt}.
15516 @itemx -mno-fused-madd
15517 @opindex mfused-madd
15518 @opindex mno-fused-madd
15519 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
15520 instructions. The default is to use these instructions.
15522 @item -mno-dwarf2-asm
15523 @itemx -mdwarf2-asm
15524 @opindex mno-dwarf2-asm
15525 @opindex mdwarf2-asm
15526 Don't (or do) generate assembler code for the DWARF 2 line number debugging
15527 info. This may be useful when not using the GNU assembler.
15529 @item -mearly-stop-bits
15530 @itemx -mno-early-stop-bits
15531 @opindex mearly-stop-bits
15532 @opindex mno-early-stop-bits
15533 Allow stop bits to be placed earlier than immediately preceding the
15534 instruction that triggered the stop bit. This can improve instruction
15535 scheduling, but does not always do so.
15537 @item -mfixed-range=@var{register-range}
15538 @opindex mfixed-range
15539 Generate code treating the given register range as fixed registers.
15540 A fixed register is one that the register allocator cannot use. This is
15541 useful when compiling kernel code. A register range is specified as
15542 two registers separated by a dash. Multiple register ranges can be
15543 specified separated by a comma.
15545 @item -mtls-size=@var{tls-size}
15547 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
15550 @item -mtune=@var{cpu-type}
15552 Tune the instruction scheduling for a particular CPU, Valid values are
15553 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
15554 and @samp{mckinley}.
15560 Generate code for a 32-bit or 64-bit environment.
15561 The 32-bit environment sets int, long and pointer to 32 bits.
15562 The 64-bit environment sets int to 32 bits and long and pointer
15563 to 64 bits. These are HP-UX specific flags.
15565 @item -mno-sched-br-data-spec
15566 @itemx -msched-br-data-spec
15567 @opindex mno-sched-br-data-spec
15568 @opindex msched-br-data-spec
15569 (Dis/En)able data speculative scheduling before reload.
15570 This results in generation of @code{ld.a} instructions and
15571 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15572 The default is 'disable'.
15574 @item -msched-ar-data-spec
15575 @itemx -mno-sched-ar-data-spec
15576 @opindex msched-ar-data-spec
15577 @opindex mno-sched-ar-data-spec
15578 (En/Dis)able data speculative scheduling after reload.
15579 This results in generation of @code{ld.a} instructions and
15580 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15581 The default is 'enable'.
15583 @item -mno-sched-control-spec
15584 @itemx -msched-control-spec
15585 @opindex mno-sched-control-spec
15586 @opindex msched-control-spec
15587 (Dis/En)able control speculative scheduling. This feature is
15588 available only during region scheduling (i.e.@: before reload).
15589 This results in generation of the @code{ld.s} instructions and
15590 the corresponding check instructions @code{chk.s}.
15591 The default is 'disable'.
15593 @item -msched-br-in-data-spec
15594 @itemx -mno-sched-br-in-data-spec
15595 @opindex msched-br-in-data-spec
15596 @opindex mno-sched-br-in-data-spec
15597 (En/Dis)able speculative scheduling of the instructions that
15598 are dependent on the data speculative loads before reload.
15599 This is effective only with @option{-msched-br-data-spec} enabled.
15600 The default is 'enable'.
15602 @item -msched-ar-in-data-spec
15603 @itemx -mno-sched-ar-in-data-spec
15604 @opindex msched-ar-in-data-spec
15605 @opindex mno-sched-ar-in-data-spec
15606 (En/Dis)able speculative scheduling of the instructions that
15607 are dependent on the data speculative loads after reload.
15608 This is effective only with @option{-msched-ar-data-spec} enabled.
15609 The default is 'enable'.
15611 @item -msched-in-control-spec
15612 @itemx -mno-sched-in-control-spec
15613 @opindex msched-in-control-spec
15614 @opindex mno-sched-in-control-spec
15615 (En/Dis)able speculative scheduling of the instructions that
15616 are dependent on the control speculative loads.
15617 This is effective only with @option{-msched-control-spec} enabled.
15618 The default is 'enable'.
15620 @item -mno-sched-prefer-non-data-spec-insns
15621 @itemx -msched-prefer-non-data-spec-insns
15622 @opindex mno-sched-prefer-non-data-spec-insns
15623 @opindex msched-prefer-non-data-spec-insns
15624 If enabled, data-speculative instructions are chosen for schedule
15625 only if there are no other choices at the moment. This makes
15626 the use of the data speculation much more conservative.
15627 The default is 'disable'.
15629 @item -mno-sched-prefer-non-control-spec-insns
15630 @itemx -msched-prefer-non-control-spec-insns
15631 @opindex mno-sched-prefer-non-control-spec-insns
15632 @opindex msched-prefer-non-control-spec-insns
15633 If enabled, control-speculative instructions are chosen for schedule
15634 only if there are no other choices at the moment. This makes
15635 the use of the control speculation much more conservative.
15636 The default is 'disable'.
15638 @item -mno-sched-count-spec-in-critical-path
15639 @itemx -msched-count-spec-in-critical-path
15640 @opindex mno-sched-count-spec-in-critical-path
15641 @opindex msched-count-spec-in-critical-path
15642 If enabled, speculative dependencies are considered during
15643 computation of the instructions priorities. This makes the use of the
15644 speculation a bit more conservative.
15645 The default is 'disable'.
15647 @item -msched-spec-ldc
15648 @opindex msched-spec-ldc
15649 Use a simple data speculation check. This option is on by default.
15651 @item -msched-control-spec-ldc
15652 @opindex msched-spec-ldc
15653 Use a simple check for control speculation. This option is on by default.
15655 @item -msched-stop-bits-after-every-cycle
15656 @opindex msched-stop-bits-after-every-cycle
15657 Place a stop bit after every cycle when scheduling. This option is on
15660 @item -msched-fp-mem-deps-zero-cost
15661 @opindex msched-fp-mem-deps-zero-cost
15662 Assume that floating-point stores and loads are not likely to cause a conflict
15663 when placed into the same instruction group. This option is disabled by
15666 @item -msel-sched-dont-check-control-spec
15667 @opindex msel-sched-dont-check-control-spec
15668 Generate checks for control speculation in selective scheduling.
15669 This flag is disabled by default.
15671 @item -msched-max-memory-insns=@var{max-insns}
15672 @opindex msched-max-memory-insns
15673 Limit on the number of memory insns per instruction group, giving lower
15674 priority to subsequent memory insns attempting to schedule in the same
15675 instruction group. Frequently useful to prevent cache bank conflicts.
15676 The default value is 1.
15678 @item -msched-max-memory-insns-hard-limit
15679 @opindex msched-max-memory-insns-hard-limit
15680 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
15681 disallowing more than that number in an instruction group.
15682 Otherwise, the limit is ``soft'', meaning that non-memory operations
15683 are preferred when the limit is reached, but memory operations may still
15689 @subsection LM32 Options
15690 @cindex LM32 options
15692 These @option{-m} options are defined for the LatticeMico32 architecture:
15695 @item -mbarrel-shift-enabled
15696 @opindex mbarrel-shift-enabled
15697 Enable barrel-shift instructions.
15699 @item -mdivide-enabled
15700 @opindex mdivide-enabled
15701 Enable divide and modulus instructions.
15703 @item -mmultiply-enabled
15704 @opindex multiply-enabled
15705 Enable multiply instructions.
15707 @item -msign-extend-enabled
15708 @opindex msign-extend-enabled
15709 Enable sign extend instructions.
15711 @item -muser-enabled
15712 @opindex muser-enabled
15713 Enable user-defined instructions.
15718 @subsection M32C Options
15719 @cindex M32C options
15722 @item -mcpu=@var{name}
15724 Select the CPU for which code is generated. @var{name} may be one of
15725 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
15726 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
15727 the M32C/80 series.
15731 Specifies that the program will be run on the simulator. This causes
15732 an alternate runtime library to be linked in which supports, for
15733 example, file I/O@. You must not use this option when generating
15734 programs that will run on real hardware; you must provide your own
15735 runtime library for whatever I/O functions are needed.
15737 @item -memregs=@var{number}
15739 Specifies the number of memory-based pseudo-registers GCC uses
15740 during code generation. These pseudo-registers are used like real
15741 registers, so there is a tradeoff between GCC's ability to fit the
15742 code into available registers, and the performance penalty of using
15743 memory instead of registers. Note that all modules in a program must
15744 be compiled with the same value for this option. Because of that, you
15745 must not use this option with GCC's default runtime libraries.
15749 @node M32R/D Options
15750 @subsection M32R/D Options
15751 @cindex M32R/D options
15753 These @option{-m} options are defined for Renesas M32R/D architectures:
15758 Generate code for the M32R/2@.
15762 Generate code for the M32R/X@.
15766 Generate code for the M32R@. This is the default.
15768 @item -mmodel=small
15769 @opindex mmodel=small
15770 Assume all objects live in the lower 16MB of memory (so that their addresses
15771 can be loaded with the @code{ld24} instruction), and assume all subroutines
15772 are reachable with the @code{bl} instruction.
15773 This is the default.
15775 The addressability of a particular object can be set with the
15776 @code{model} attribute.
15778 @item -mmodel=medium
15779 @opindex mmodel=medium
15780 Assume objects may be anywhere in the 32-bit address space (the compiler
15781 generates @code{seth/add3} instructions to load their addresses), and
15782 assume all subroutines are reachable with the @code{bl} instruction.
15784 @item -mmodel=large
15785 @opindex mmodel=large
15786 Assume objects may be anywhere in the 32-bit address space (the compiler
15787 generates @code{seth/add3} instructions to load their addresses), and
15788 assume subroutines may not be reachable with the @code{bl} instruction
15789 (the compiler generates the much slower @code{seth/add3/jl}
15790 instruction sequence).
15793 @opindex msdata=none
15794 Disable use of the small data area. Variables are put into
15795 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
15796 @code{section} attribute has been specified).
15797 This is the default.
15799 The small data area consists of sections @code{.sdata} and @code{.sbss}.
15800 Objects may be explicitly put in the small data area with the
15801 @code{section} attribute using one of these sections.
15803 @item -msdata=sdata
15804 @opindex msdata=sdata
15805 Put small global and static data in the small data area, but do not
15806 generate special code to reference them.
15809 @opindex msdata=use
15810 Put small global and static data in the small data area, and generate
15811 special instructions to reference them.
15815 @cindex smaller data references
15816 Put global and static objects less than or equal to @var{num} bytes
15817 into the small data or BSS sections instead of the normal data or BSS
15818 sections. The default value of @var{num} is 8.
15819 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
15820 for this option to have any effect.
15822 All modules should be compiled with the same @option{-G @var{num}} value.
15823 Compiling with different values of @var{num} may or may not work; if it
15824 doesn't the linker gives an error message---incorrect code is not
15829 Makes the M32R-specific code in the compiler display some statistics
15830 that might help in debugging programs.
15832 @item -malign-loops
15833 @opindex malign-loops
15834 Align all loops to a 32-byte boundary.
15836 @item -mno-align-loops
15837 @opindex mno-align-loops
15838 Do not enforce a 32-byte alignment for loops. This is the default.
15840 @item -missue-rate=@var{number}
15841 @opindex missue-rate=@var{number}
15842 Issue @var{number} instructions per cycle. @var{number} can only be 1
15845 @item -mbranch-cost=@var{number}
15846 @opindex mbranch-cost=@var{number}
15847 @var{number} can only be 1 or 2. If it is 1 then branches are
15848 preferred over conditional code, if it is 2, then the opposite applies.
15850 @item -mflush-trap=@var{number}
15851 @opindex mflush-trap=@var{number}
15852 Specifies the trap number to use to flush the cache. The default is
15853 12. Valid numbers are between 0 and 15 inclusive.
15855 @item -mno-flush-trap
15856 @opindex mno-flush-trap
15857 Specifies that the cache cannot be flushed by using a trap.
15859 @item -mflush-func=@var{name}
15860 @opindex mflush-func=@var{name}
15861 Specifies the name of the operating system function to call to flush
15862 the cache. The default is @samp{_flush_cache}, but a function call
15863 is only used if a trap is not available.
15865 @item -mno-flush-func
15866 @opindex mno-flush-func
15867 Indicates that there is no OS function for flushing the cache.
15871 @node M680x0 Options
15872 @subsection M680x0 Options
15873 @cindex M680x0 options
15875 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
15876 The default settings depend on which architecture was selected when
15877 the compiler was configured; the defaults for the most common choices
15881 @item -march=@var{arch}
15883 Generate code for a specific M680x0 or ColdFire instruction set
15884 architecture. Permissible values of @var{arch} for M680x0
15885 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
15886 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
15887 architectures are selected according to Freescale's ISA classification
15888 and the permissible values are: @samp{isaa}, @samp{isaaplus},
15889 @samp{isab} and @samp{isac}.
15891 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
15892 code for a ColdFire target. The @var{arch} in this macro is one of the
15893 @option{-march} arguments given above.
15895 When used together, @option{-march} and @option{-mtune} select code
15896 that runs on a family of similar processors but that is optimized
15897 for a particular microarchitecture.
15899 @item -mcpu=@var{cpu}
15901 Generate code for a specific M680x0 or ColdFire processor.
15902 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
15903 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
15904 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
15905 below, which also classifies the CPUs into families:
15907 @multitable @columnfractions 0.20 0.80
15908 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
15909 @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}
15910 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
15911 @item @samp{5206e} @tab @samp{5206e}
15912 @item @samp{5208} @tab @samp{5207} @samp{5208}
15913 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
15914 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
15915 @item @samp{5216} @tab @samp{5214} @samp{5216}
15916 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
15917 @item @samp{5225} @tab @samp{5224} @samp{5225}
15918 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
15919 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
15920 @item @samp{5249} @tab @samp{5249}
15921 @item @samp{5250} @tab @samp{5250}
15922 @item @samp{5271} @tab @samp{5270} @samp{5271}
15923 @item @samp{5272} @tab @samp{5272}
15924 @item @samp{5275} @tab @samp{5274} @samp{5275}
15925 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
15926 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
15927 @item @samp{5307} @tab @samp{5307}
15928 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
15929 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
15930 @item @samp{5407} @tab @samp{5407}
15931 @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}
15934 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
15935 @var{arch} is compatible with @var{cpu}. Other combinations of
15936 @option{-mcpu} and @option{-march} are rejected.
15938 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
15939 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
15940 where the value of @var{family} is given by the table above.
15942 @item -mtune=@var{tune}
15944 Tune the code for a particular microarchitecture within the
15945 constraints set by @option{-march} and @option{-mcpu}.
15946 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
15947 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
15948 and @samp{cpu32}. The ColdFire microarchitectures
15949 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
15951 You can also use @option{-mtune=68020-40} for code that needs
15952 to run relatively well on 68020, 68030 and 68040 targets.
15953 @option{-mtune=68020-60} is similar but includes 68060 targets
15954 as well. These two options select the same tuning decisions as
15955 @option{-m68020-40} and @option{-m68020-60} respectively.
15957 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
15958 when tuning for 680x0 architecture @var{arch}. It also defines
15959 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
15960 option is used. If GCC is tuning for a range of architectures,
15961 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
15962 it defines the macros for every architecture in the range.
15964 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
15965 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
15966 of the arguments given above.
15972 Generate output for a 68000. This is the default
15973 when the compiler is configured for 68000-based systems.
15974 It is equivalent to @option{-march=68000}.
15976 Use this option for microcontrollers with a 68000 or EC000 core,
15977 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
15981 Generate output for a 68010. This is the default
15982 when the compiler is configured for 68010-based systems.
15983 It is equivalent to @option{-march=68010}.
15989 Generate output for a 68020. This is the default
15990 when the compiler is configured for 68020-based systems.
15991 It is equivalent to @option{-march=68020}.
15995 Generate output for a 68030. This is the default when the compiler is
15996 configured for 68030-based systems. It is equivalent to
15997 @option{-march=68030}.
16001 Generate output for a 68040. This is the default when the compiler is
16002 configured for 68040-based systems. It is equivalent to
16003 @option{-march=68040}.
16005 This option inhibits the use of 68881/68882 instructions that have to be
16006 emulated by software on the 68040. Use this option if your 68040 does not
16007 have code to emulate those instructions.
16011 Generate output for a 68060. This is the default when the compiler is
16012 configured for 68060-based systems. It is equivalent to
16013 @option{-march=68060}.
16015 This option inhibits the use of 68020 and 68881/68882 instructions that
16016 have to be emulated by software on the 68060. Use this option if your 68060
16017 does not have code to emulate those instructions.
16021 Generate output for a CPU32. This is the default
16022 when the compiler is configured for CPU32-based systems.
16023 It is equivalent to @option{-march=cpu32}.
16025 Use this option for microcontrollers with a
16026 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16027 68336, 68340, 68341, 68349 and 68360.
16031 Generate output for a 520X ColdFire CPU@. This is the default
16032 when the compiler is configured for 520X-based systems.
16033 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16034 in favor of that option.
16036 Use this option for microcontroller with a 5200 core, including
16037 the MCF5202, MCF5203, MCF5204 and MCF5206.
16041 Generate output for a 5206e ColdFire CPU@. The option is now
16042 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16046 Generate output for a member of the ColdFire 528X family.
16047 The option is now deprecated in favor of the equivalent
16048 @option{-mcpu=528x}.
16052 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16053 in favor of the equivalent @option{-mcpu=5307}.
16057 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16058 in favor of the equivalent @option{-mcpu=5407}.
16062 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16063 This includes use of hardware floating-point instructions.
16064 The option is equivalent to @option{-mcpu=547x}, and is now
16065 deprecated in favor of that option.
16069 Generate output for a 68040, without using any of the new instructions.
16070 This results in code that can run relatively efficiently on either a
16071 68020/68881 or a 68030 or a 68040. The generated code does use the
16072 68881 instructions that are emulated on the 68040.
16074 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16078 Generate output for a 68060, without using any of the new instructions.
16079 This results in code that can run relatively efficiently on either a
16080 68020/68881 or a 68030 or a 68040. The generated code does use the
16081 68881 instructions that are emulated on the 68060.
16083 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16087 @opindex mhard-float
16089 Generate floating-point instructions. This is the default for 68020
16090 and above, and for ColdFire devices that have an FPU@. It defines the
16091 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16092 on ColdFire targets.
16095 @opindex msoft-float
16096 Do not generate floating-point instructions; use library calls instead.
16097 This is the default for 68000, 68010, and 68832 targets. It is also
16098 the default for ColdFire devices that have no FPU.
16104 Generate (do not generate) ColdFire hardware divide and remainder
16105 instructions. If @option{-march} is used without @option{-mcpu},
16106 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16107 architectures. Otherwise, the default is taken from the target CPU
16108 (either the default CPU, or the one specified by @option{-mcpu}). For
16109 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16110 @option{-mcpu=5206e}.
16112 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16116 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16117 Additionally, parameters passed on the stack are also aligned to a
16118 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16122 Do not consider type @code{int} to be 16 bits wide. This is the default.
16125 @itemx -mno-bitfield
16126 @opindex mnobitfield
16127 @opindex mno-bitfield
16128 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16129 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16133 Do use the bit-field instructions. The @option{-m68020} option implies
16134 @option{-mbitfield}. This is the default if you use a configuration
16135 designed for a 68020.
16139 Use a different function-calling convention, in which functions
16140 that take a fixed number of arguments return with the @code{rtd}
16141 instruction, which pops their arguments while returning. This
16142 saves one instruction in the caller since there is no need to pop
16143 the arguments there.
16145 This calling convention is incompatible with the one normally
16146 used on Unix, so you cannot use it if you need to call libraries
16147 compiled with the Unix compiler.
16149 Also, you must provide function prototypes for all functions that
16150 take variable numbers of arguments (including @code{printf});
16151 otherwise incorrect code is generated for calls to those
16154 In addition, seriously incorrect code results if you call a
16155 function with too many arguments. (Normally, extra arguments are
16156 harmlessly ignored.)
16158 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16159 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16163 Do not use the calling conventions selected by @option{-mrtd}.
16164 This is the default.
16167 @itemx -mno-align-int
16168 @opindex malign-int
16169 @opindex mno-align-int
16170 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16171 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16172 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16173 Aligning variables on 32-bit boundaries produces code that runs somewhat
16174 faster on processors with 32-bit busses at the expense of more memory.
16176 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16177 aligns structures containing the above types differently than
16178 most published application binary interface specifications for the m68k.
16182 Use the pc-relative addressing mode of the 68000 directly, instead of
16183 using a global offset table. At present, this option implies @option{-fpic},
16184 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16185 not presently supported with @option{-mpcrel}, though this could be supported for
16186 68020 and higher processors.
16188 @item -mno-strict-align
16189 @itemx -mstrict-align
16190 @opindex mno-strict-align
16191 @opindex mstrict-align
16192 Do not (do) assume that unaligned memory references are handled by
16196 Generate code that allows the data segment to be located in a different
16197 area of memory from the text segment. This allows for execute-in-place in
16198 an environment without virtual memory management. This option implies
16201 @item -mno-sep-data
16202 Generate code that assumes that the data segment follows the text segment.
16203 This is the default.
16205 @item -mid-shared-library
16206 Generate code that supports shared libraries via the library ID method.
16207 This allows for execute-in-place and shared libraries in an environment
16208 without virtual memory management. This option implies @option{-fPIC}.
16210 @item -mno-id-shared-library
16211 Generate code that doesn't assume ID-based shared libraries are being used.
16212 This is the default.
16214 @item -mshared-library-id=n
16215 Specifies the identification number of the ID-based shared library being
16216 compiled. Specifying a value of 0 generates more compact code; specifying
16217 other values forces the allocation of that number to the current
16218 library, but is no more space- or time-efficient than omitting this option.
16224 When generating position-independent code for ColdFire, generate code
16225 that works if the GOT has more than 8192 entries. This code is
16226 larger and slower than code generated without this option. On M680x0
16227 processors, this option is not needed; @option{-fPIC} suffices.
16229 GCC normally uses a single instruction to load values from the GOT@.
16230 While this is relatively efficient, it only works if the GOT
16231 is smaller than about 64k. Anything larger causes the linker
16232 to report an error such as:
16234 @cindex relocation truncated to fit (ColdFire)
16236 relocation truncated to fit: R_68K_GOT16O foobar
16239 If this happens, you should recompile your code with @option{-mxgot}.
16240 It should then work with very large GOTs. However, code generated with
16241 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16242 the value of a global symbol.
16244 Note that some linkers, including newer versions of the GNU linker,
16245 can create multiple GOTs and sort GOT entries. If you have such a linker,
16246 you should only need to use @option{-mxgot} when compiling a single
16247 object file that accesses more than 8192 GOT entries. Very few do.
16249 These options have no effect unless GCC is generating
16250 position-independent code.
16254 @node MCore Options
16255 @subsection MCore Options
16256 @cindex MCore options
16258 These are the @samp{-m} options defined for the Motorola M*Core
16264 @itemx -mno-hardlit
16266 @opindex mno-hardlit
16267 Inline constants into the code stream if it can be done in two
16268 instructions or less.
16274 Use the divide instruction. (Enabled by default).
16276 @item -mrelax-immediate
16277 @itemx -mno-relax-immediate
16278 @opindex mrelax-immediate
16279 @opindex mno-relax-immediate
16280 Allow arbitrary-sized immediates in bit operations.
16282 @item -mwide-bitfields
16283 @itemx -mno-wide-bitfields
16284 @opindex mwide-bitfields
16285 @opindex mno-wide-bitfields
16286 Always treat bit-fields as @code{int}-sized.
16288 @item -m4byte-functions
16289 @itemx -mno-4byte-functions
16290 @opindex m4byte-functions
16291 @opindex mno-4byte-functions
16292 Force all functions to be aligned to a 4-byte boundary.
16294 @item -mcallgraph-data
16295 @itemx -mno-callgraph-data
16296 @opindex mcallgraph-data
16297 @opindex mno-callgraph-data
16298 Emit callgraph information.
16301 @itemx -mno-slow-bytes
16302 @opindex mslow-bytes
16303 @opindex mno-slow-bytes
16304 Prefer word access when reading byte quantities.
16306 @item -mlittle-endian
16307 @itemx -mbig-endian
16308 @opindex mlittle-endian
16309 @opindex mbig-endian
16310 Generate code for a little-endian target.
16316 Generate code for the 210 processor.
16320 Assume that runtime support has been provided and so omit the
16321 simulator library (@file{libsim.a)} from the linker command line.
16323 @item -mstack-increment=@var{size}
16324 @opindex mstack-increment
16325 Set the maximum amount for a single stack increment operation. Large
16326 values can increase the speed of programs that contain functions
16327 that need a large amount of stack space, but they can also trigger a
16328 segmentation fault if the stack is extended too much. The default
16334 @subsection MeP Options
16335 @cindex MeP options
16341 Enables the @code{abs} instruction, which is the absolute difference
16342 between two registers.
16346 Enables all the optional instructions---average, multiply, divide, bit
16347 operations, leading zero, absolute difference, min/max, clip, and
16353 Enables the @code{ave} instruction, which computes the average of two
16356 @item -mbased=@var{n}
16358 Variables of size @var{n} bytes or smaller are placed in the
16359 @code{.based} section by default. Based variables use the @code{$tp}
16360 register as a base register, and there is a 128-byte limit to the
16361 @code{.based} section.
16365 Enables the bit operation instructions---bit test (@code{btstm}), set
16366 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16367 test-and-set (@code{tas}).
16369 @item -mc=@var{name}
16371 Selects which section constant data is placed in. @var{name} may
16372 be @samp{tiny}, @samp{near}, or @samp{far}.
16376 Enables the @code{clip} instruction. Note that @option{-mclip} is not
16377 useful unless you also provide @option{-mminmax}.
16379 @item -mconfig=@var{name}
16381 Selects one of the built-in core configurations. Each MeP chip has
16382 one or more modules in it; each module has a core CPU and a variety of
16383 coprocessors, optional instructions, and peripherals. The
16384 @code{MeP-Integrator} tool, not part of GCC, provides these
16385 configurations through this option; using this option is the same as
16386 using all the corresponding command-line options. The default
16387 configuration is @samp{default}.
16391 Enables the coprocessor instructions. By default, this is a 32-bit
16392 coprocessor. Note that the coprocessor is normally enabled via the
16393 @option{-mconfig=} option.
16397 Enables the 32-bit coprocessor's instructions.
16401 Enables the 64-bit coprocessor's instructions.
16405 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16409 Causes constant variables to be placed in the @code{.near} section.
16413 Enables the @code{div} and @code{divu} instructions.
16417 Generate big-endian code.
16421 Generate little-endian code.
16423 @item -mio-volatile
16424 @opindex mio-volatile
16425 Tells the compiler that any variable marked with the @code{io}
16426 attribute is to be considered volatile.
16430 Causes variables to be assigned to the @code{.far} section by default.
16434 Enables the @code{leadz} (leading zero) instruction.
16438 Causes variables to be assigned to the @code{.near} section by default.
16442 Enables the @code{min} and @code{max} instructions.
16446 Enables the multiplication and multiply-accumulate instructions.
16450 Disables all the optional instructions enabled by @option{-mall-opts}.
16454 Enables the @code{repeat} and @code{erepeat} instructions, used for
16455 low-overhead looping.
16459 Causes all variables to default to the @code{.tiny} section. Note
16460 that there is a 65536-byte limit to this section. Accesses to these
16461 variables use the @code{%gp} base register.
16465 Enables the saturation instructions. Note that the compiler does not
16466 currently generate these itself, but this option is included for
16467 compatibility with other tools, like @code{as}.
16471 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16475 Link the simulator run-time libraries.
16479 Link the simulator runtime libraries, excluding built-in support
16480 for reset and exception vectors and tables.
16484 Causes all functions to default to the @code{.far} section. Without
16485 this option, functions default to the @code{.near} section.
16487 @item -mtiny=@var{n}
16489 Variables that are @var{n} bytes or smaller are allocated to the
16490 @code{.tiny} section. These variables use the @code{$gp} base
16491 register. The default for this option is 4, but note that there's a
16492 65536-byte limit to the @code{.tiny} section.
16496 @node MicroBlaze Options
16497 @subsection MicroBlaze Options
16498 @cindex MicroBlaze Options
16503 @opindex msoft-float
16504 Use software emulation for floating point (default).
16507 @opindex mhard-float
16508 Use hardware floating-point instructions.
16512 Do not optimize block moves, use @code{memcpy}.
16514 @item -mno-clearbss
16515 @opindex mno-clearbss
16516 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
16518 @item -mcpu=@var{cpu-type}
16520 Use features of, and schedule code for, the given CPU.
16521 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
16522 where @var{X} is a major version, @var{YY} is the minor version, and
16523 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
16524 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
16526 @item -mxl-soft-mul
16527 @opindex mxl-soft-mul
16528 Use software multiply emulation (default).
16530 @item -mxl-soft-div
16531 @opindex mxl-soft-div
16532 Use software emulation for divides (default).
16534 @item -mxl-barrel-shift
16535 @opindex mxl-barrel-shift
16536 Use the hardware barrel shifter.
16538 @item -mxl-pattern-compare
16539 @opindex mxl-pattern-compare
16540 Use pattern compare instructions.
16542 @item -msmall-divides
16543 @opindex msmall-divides
16544 Use table lookup optimization for small signed integer divisions.
16546 @item -mxl-stack-check
16547 @opindex mxl-stack-check
16548 This option is deprecated. Use @option{-fstack-check} instead.
16551 @opindex mxl-gp-opt
16552 Use GP-relative @code{.sdata}/@code{.sbss} sections.
16554 @item -mxl-multiply-high
16555 @opindex mxl-multiply-high
16556 Use multiply high instructions for high part of 32x32 multiply.
16558 @item -mxl-float-convert
16559 @opindex mxl-float-convert
16560 Use hardware floating-point conversion instructions.
16562 @item -mxl-float-sqrt
16563 @opindex mxl-float-sqrt
16564 Use hardware floating-point square root instruction.
16567 @opindex mbig-endian
16568 Generate code for a big-endian target.
16570 @item -mlittle-endian
16571 @opindex mlittle-endian
16572 Generate code for a little-endian target.
16575 @opindex mxl-reorder
16576 Use reorder instructions (swap and byte reversed load/store).
16578 @item -mxl-mode-@var{app-model}
16579 Select application model @var{app-model}. Valid models are
16582 normal executable (default), uses startup code @file{crt0.o}.
16585 for use with Xilinx Microprocessor Debugger (XMD) based
16586 software intrusive debug agent called xmdstub. This uses startup file
16587 @file{crt1.o} and sets the start address of the program to 0x800.
16590 for applications that are loaded using a bootloader.
16591 This model uses startup file @file{crt2.o} which does not contain a processor
16592 reset vector handler. This is suitable for transferring control on a
16593 processor reset to the bootloader rather than the application.
16596 for applications that do not require any of the
16597 MicroBlaze vectors. This option may be useful for applications running
16598 within a monitoring application. This model uses @file{crt3.o} as a startup file.
16601 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
16602 @option{-mxl-mode-@var{app-model}}.
16607 @subsection MIPS Options
16608 @cindex MIPS options
16614 Generate big-endian code.
16618 Generate little-endian code. This is the default for @samp{mips*el-*-*}
16621 @item -march=@var{arch}
16623 Generate code that runs on @var{arch}, which can be the name of a
16624 generic MIPS ISA, or the name of a particular processor.
16626 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
16627 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
16628 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
16629 @samp{mips64r5} and @samp{mips64r6}.
16630 The processor names are:
16631 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
16632 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
16633 @samp{5kc}, @samp{5kf},
16635 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
16636 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
16637 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
16638 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
16639 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
16640 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
16642 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
16643 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
16646 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
16647 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
16648 @samp{rm7000}, @samp{rm9000},
16649 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
16652 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
16653 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
16654 @samp{xlr} and @samp{xlp}.
16655 The special value @samp{from-abi} selects the
16656 most compatible architecture for the selected ABI (that is,
16657 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
16659 The native Linux/GNU toolchain also supports the value @samp{native},
16660 which selects the best architecture option for the host processor.
16661 @option{-march=native} has no effect if GCC does not recognize
16664 In processor names, a final @samp{000} can be abbreviated as @samp{k}
16665 (for example, @option{-march=r2k}). Prefixes are optional, and
16666 @samp{vr} may be written @samp{r}.
16668 Names of the form @samp{@var{n}f2_1} refer to processors with
16669 FPUs clocked at half the rate of the core, names of the form
16670 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
16671 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
16672 processors with FPUs clocked a ratio of 3:2 with respect to the core.
16673 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
16674 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
16675 accepted as synonyms for @samp{@var{n}f1_1}.
16677 GCC defines two macros based on the value of this option. The first
16678 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
16679 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
16680 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
16681 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
16682 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
16684 Note that the @code{_MIPS_ARCH} macro uses the processor names given
16685 above. In other words, it has the full prefix and does not
16686 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
16687 the macro names the resolved architecture (either @code{"mips1"} or
16688 @code{"mips3"}). It names the default architecture when no
16689 @option{-march} option is given.
16691 @item -mtune=@var{arch}
16693 Optimize for @var{arch}. Among other things, this option controls
16694 the way instructions are scheduled, and the perceived cost of arithmetic
16695 operations. The list of @var{arch} values is the same as for
16698 When this option is not used, GCC optimizes for the processor
16699 specified by @option{-march}. By using @option{-march} and
16700 @option{-mtune} together, it is possible to generate code that
16701 runs on a family of processors, but optimize the code for one
16702 particular member of that family.
16704 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
16705 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
16706 @option{-march} ones described above.
16710 Equivalent to @option{-march=mips1}.
16714 Equivalent to @option{-march=mips2}.
16718 Equivalent to @option{-march=mips3}.
16722 Equivalent to @option{-march=mips4}.
16726 Equivalent to @option{-march=mips32}.
16730 Equivalent to @option{-march=mips32r3}.
16734 Equivalent to @option{-march=mips32r5}.
16738 Equivalent to @option{-march=mips32r6}.
16742 Equivalent to @option{-march=mips64}.
16746 Equivalent to @option{-march=mips64r2}.
16750 Equivalent to @option{-march=mips64r3}.
16754 Equivalent to @option{-march=mips64r5}.
16758 Equivalent to @option{-march=mips64r6}.
16763 @opindex mno-mips16
16764 Generate (do not generate) MIPS16 code. If GCC is targeting a
16765 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
16767 MIPS16 code generation can also be controlled on a per-function basis
16768 by means of @code{mips16} and @code{nomips16} attributes.
16769 @xref{Function Attributes}, for more information.
16771 @item -mflip-mips16
16772 @opindex mflip-mips16
16773 Generate MIPS16 code on alternating functions. This option is provided
16774 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
16775 not intended for ordinary use in compiling user code.
16777 @item -minterlink-compressed
16778 @item -mno-interlink-compressed
16779 @opindex minterlink-compressed
16780 @opindex mno-interlink-compressed
16781 Require (do not require) that code using the standard (uncompressed) MIPS ISA
16782 be link-compatible with MIPS16 and microMIPS code, and vice versa.
16784 For example, code using the standard ISA encoding cannot jump directly
16785 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
16786 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
16787 knows that the target of the jump is not compressed.
16789 @item -minterlink-mips16
16790 @itemx -mno-interlink-mips16
16791 @opindex minterlink-mips16
16792 @opindex mno-interlink-mips16
16793 Aliases of @option{-minterlink-compressed} and
16794 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
16795 and are retained for backwards compatibility.
16807 Generate code for the given ABI@.
16809 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
16810 generates 64-bit code when you select a 64-bit architecture, but you
16811 can use @option{-mgp32} to get 32-bit code instead.
16813 For information about the O64 ABI, see
16814 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
16816 GCC supports a variant of the o32 ABI in which floating-point registers
16817 are 64 rather than 32 bits wide. You can select this combination with
16818 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
16819 and @code{mfhc1} instructions and is therefore only supported for
16820 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
16822 The register assignments for arguments and return values remain the
16823 same, but each scalar value is passed in a single 64-bit register
16824 rather than a pair of 32-bit registers. For example, scalar
16825 floating-point values are returned in @samp{$f0} only, not a
16826 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
16827 remains the same in that the even-numbered double-precision registers
16830 Two additional variants of the o32 ABI are supported to enable
16831 a transition from 32-bit to 64-bit registers. These are FPXX
16832 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
16833 The FPXX extension mandates that all code must execute correctly
16834 when run using 32-bit or 64-bit registers. The code can be interlinked
16835 with either FP32 or FP64, but not both.
16836 The FP64A extension is similar to the FP64 extension but forbids the
16837 use of odd-numbered single-precision registers. This can be used
16838 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
16839 processors and allows both FP32 and FP64A code to interlink and
16840 run in the same process without changing FPU modes.
16843 @itemx -mno-abicalls
16845 @opindex mno-abicalls
16846 Generate (do not generate) code that is suitable for SVR4-style
16847 dynamic objects. @option{-mabicalls} is the default for SVR4-based
16852 Generate (do not generate) code that is fully position-independent,
16853 and that can therefore be linked into shared libraries. This option
16854 only affects @option{-mabicalls}.
16856 All @option{-mabicalls} code has traditionally been position-independent,
16857 regardless of options like @option{-fPIC} and @option{-fpic}. However,
16858 as an extension, the GNU toolchain allows executables to use absolute
16859 accesses for locally-binding symbols. It can also use shorter GP
16860 initialization sequences and generate direct calls to locally-defined
16861 functions. This mode is selected by @option{-mno-shared}.
16863 @option{-mno-shared} depends on binutils 2.16 or higher and generates
16864 objects that can only be linked by the GNU linker. However, the option
16865 does not affect the ABI of the final executable; it only affects the ABI
16866 of relocatable objects. Using @option{-mno-shared} generally makes
16867 executables both smaller and quicker.
16869 @option{-mshared} is the default.
16875 Assume (do not assume) that the static and dynamic linkers
16876 support PLTs and copy relocations. This option only affects
16877 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
16878 has no effect without @option{-msym32}.
16880 You can make @option{-mplt} the default by configuring
16881 GCC with @option{--with-mips-plt}. The default is
16882 @option{-mno-plt} otherwise.
16888 Lift (do not lift) the usual restrictions on the size of the global
16891 GCC normally uses a single instruction to load values from the GOT@.
16892 While this is relatively efficient, it only works if the GOT
16893 is smaller than about 64k. Anything larger causes the linker
16894 to report an error such as:
16896 @cindex relocation truncated to fit (MIPS)
16898 relocation truncated to fit: R_MIPS_GOT16 foobar
16901 If this happens, you should recompile your code with @option{-mxgot}.
16902 This works with very large GOTs, although the code is also
16903 less efficient, since it takes three instructions to fetch the
16904 value of a global symbol.
16906 Note that some linkers can create multiple GOTs. If you have such a
16907 linker, you should only need to use @option{-mxgot} when a single object
16908 file accesses more than 64k's worth of GOT entries. Very few do.
16910 These options have no effect unless GCC is generating position
16915 Assume that general-purpose registers are 32 bits wide.
16919 Assume that general-purpose registers are 64 bits wide.
16923 Assume that floating-point registers are 32 bits wide.
16927 Assume that floating-point registers are 64 bits wide.
16931 Do not assume the width of floating-point registers.
16934 @opindex mhard-float
16935 Use floating-point coprocessor instructions.
16938 @opindex msoft-float
16939 Do not use floating-point coprocessor instructions. Implement
16940 floating-point calculations using library calls instead.
16944 Equivalent to @option{-msoft-float}, but additionally asserts that the
16945 program being compiled does not perform any floating-point operations.
16946 This option is presently supported only by some bare-metal MIPS
16947 configurations, where it may select a special set of libraries
16948 that lack all floating-point support (including, for example, the
16949 floating-point @code{printf} formats).
16950 If code compiled with @option{-mno-float} accidentally contains
16951 floating-point operations, it is likely to suffer a link-time
16952 or run-time failure.
16954 @item -msingle-float
16955 @opindex msingle-float
16956 Assume that the floating-point coprocessor only supports single-precision
16959 @item -mdouble-float
16960 @opindex mdouble-float
16961 Assume that the floating-point coprocessor supports double-precision
16962 operations. This is the default.
16965 @itemx -mno-odd-spreg
16966 @opindex modd-spreg
16967 @opindex mno-odd-spreg
16968 Enable the use of odd-numbered single-precision floating-point registers
16969 for the o32 ABI. This is the default for processors that are known to
16970 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
16974 @itemx -mabs=legacy
16976 @opindex mabs=legacy
16977 These options control the treatment of the special not-a-number (NaN)
16978 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
16979 @code{neg.@i{fmt}} machine instructions.
16981 By default or when the @option{-mabs=legacy} is used the legacy
16982 treatment is selected. In this case these instructions are considered
16983 arithmetic and avoided where correct operation is required and the
16984 input operand might be a NaN. A longer sequence of instructions that
16985 manipulate the sign bit of floating-point datum manually is used
16986 instead unless the @option{-ffinite-math-only} option has also been
16989 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
16990 this case these instructions are considered non-arithmetic and therefore
16991 operating correctly in all cases, including in particular where the
16992 input operand is a NaN. These instructions are therefore always used
16993 for the respective operations.
16996 @itemx -mnan=legacy
16998 @opindex mnan=legacy
16999 These options control the encoding of the special not-a-number (NaN)
17000 IEEE 754 floating-point data.
17002 The @option{-mnan=legacy} option selects the legacy encoding. In this
17003 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17004 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17005 by the first bit of their trailing significand field being 1.
17007 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17008 this case qNaNs are denoted by the first bit of their trailing
17009 significand field being 1, whereas sNaNs are denoted by the first bit of
17010 their trailing significand field being 0.
17012 The default is @option{-mnan=legacy} unless GCC has been configured with
17013 @option{--with-nan=2008}.
17019 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17020 implement atomic memory built-in functions. When neither option is
17021 specified, GCC uses the instructions if the target architecture
17024 @option{-mllsc} is useful if the runtime environment can emulate the
17025 instructions and @option{-mno-llsc} can be useful when compiling for
17026 nonstandard ISAs. You can make either option the default by
17027 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17028 respectively. @option{--with-llsc} is the default for some
17029 configurations; see the installation documentation for details.
17035 Use (do not use) revision 1 of the MIPS DSP ASE@.
17036 @xref{MIPS DSP Built-in Functions}. This option defines the
17037 preprocessor macro @code{__mips_dsp}. It also defines
17038 @code{__mips_dsp_rev} to 1.
17044 Use (do not use) revision 2 of the MIPS DSP ASE@.
17045 @xref{MIPS DSP Built-in Functions}. This option defines the
17046 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17047 It also defines @code{__mips_dsp_rev} to 2.
17050 @itemx -mno-smartmips
17051 @opindex msmartmips
17052 @opindex mno-smartmips
17053 Use (do not use) the MIPS SmartMIPS ASE.
17055 @item -mpaired-single
17056 @itemx -mno-paired-single
17057 @opindex mpaired-single
17058 @opindex mno-paired-single
17059 Use (do not use) paired-single floating-point instructions.
17060 @xref{MIPS Paired-Single Support}. This option requires
17061 hardware floating-point support to be enabled.
17067 Use (do not use) MIPS Digital Media Extension instructions.
17068 This option can only be used when generating 64-bit code and requires
17069 hardware floating-point support to be enabled.
17074 @opindex mno-mips3d
17075 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17076 The option @option{-mips3d} implies @option{-mpaired-single}.
17079 @itemx -mno-micromips
17080 @opindex mmicromips
17081 @opindex mno-mmicromips
17082 Generate (do not generate) microMIPS code.
17084 MicroMIPS code generation can also be controlled on a per-function basis
17085 by means of @code{micromips} and @code{nomicromips} attributes.
17086 @xref{Function Attributes}, for more information.
17092 Use (do not use) MT Multithreading instructions.
17098 Use (do not use) the MIPS MCU ASE instructions.
17104 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17110 Use (do not use) the MIPS Virtualization Application Specific instructions.
17116 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17120 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17121 an explanation of the default and the way that the pointer size is
17126 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17128 The default size of @code{int}s, @code{long}s and pointers depends on
17129 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17130 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17131 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17132 or the same size as integer registers, whichever is smaller.
17138 Assume (do not assume) that all symbols have 32-bit values, regardless
17139 of the selected ABI@. This option is useful in combination with
17140 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17141 to generate shorter and faster references to symbolic addresses.
17145 Put definitions of externally-visible data in a small data section
17146 if that data is no bigger than @var{num} bytes. GCC can then generate
17147 more efficient accesses to the data; see @option{-mgpopt} for details.
17149 The default @option{-G} option depends on the configuration.
17151 @item -mlocal-sdata
17152 @itemx -mno-local-sdata
17153 @opindex mlocal-sdata
17154 @opindex mno-local-sdata
17155 Extend (do not extend) the @option{-G} behavior to local data too,
17156 such as to static variables in C@. @option{-mlocal-sdata} is the
17157 default for all configurations.
17159 If the linker complains that an application is using too much small data,
17160 you might want to try rebuilding the less performance-critical parts with
17161 @option{-mno-local-sdata}. You might also want to build large
17162 libraries with @option{-mno-local-sdata}, so that the libraries leave
17163 more room for the main program.
17165 @item -mextern-sdata
17166 @itemx -mno-extern-sdata
17167 @opindex mextern-sdata
17168 @opindex mno-extern-sdata
17169 Assume (do not assume) that externally-defined data is in
17170 a small data section if the size of that data is within the @option{-G} limit.
17171 @option{-mextern-sdata} is the default for all configurations.
17173 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17174 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17175 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17176 is placed in a small data section. If @var{Var} is defined by another
17177 module, you must either compile that module with a high-enough
17178 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17179 definition. If @var{Var} is common, you must link the application
17180 with a high-enough @option{-G} setting.
17182 The easiest way of satisfying these restrictions is to compile
17183 and link every module with the same @option{-G} option. However,
17184 you may wish to build a library that supports several different
17185 small data limits. You can do this by compiling the library with
17186 the highest supported @option{-G} setting and additionally using
17187 @option{-mno-extern-sdata} to stop the library from making assumptions
17188 about externally-defined data.
17194 Use (do not use) GP-relative accesses for symbols that are known to be
17195 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17196 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17199 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17200 might not hold the value of @code{_gp}. For example, if the code is
17201 part of a library that might be used in a boot monitor, programs that
17202 call boot monitor routines pass an unknown value in @code{$gp}.
17203 (In such situations, the boot monitor itself is usually compiled
17204 with @option{-G0}.)
17206 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17207 @option{-mno-extern-sdata}.
17209 @item -membedded-data
17210 @itemx -mno-embedded-data
17211 @opindex membedded-data
17212 @opindex mno-embedded-data
17213 Allocate variables to the read-only data section first if possible, then
17214 next in the small data section if possible, otherwise in data. This gives
17215 slightly slower code than the default, but reduces the amount of RAM required
17216 when executing, and thus may be preferred for some embedded systems.
17218 @item -muninit-const-in-rodata
17219 @itemx -mno-uninit-const-in-rodata
17220 @opindex muninit-const-in-rodata
17221 @opindex mno-uninit-const-in-rodata
17222 Put uninitialized @code{const} variables in the read-only data section.
17223 This option is only meaningful in conjunction with @option{-membedded-data}.
17225 @item -mcode-readable=@var{setting}
17226 @opindex mcode-readable
17227 Specify whether GCC may generate code that reads from executable sections.
17228 There are three possible settings:
17231 @item -mcode-readable=yes
17232 Instructions may freely access executable sections. This is the
17235 @item -mcode-readable=pcrel
17236 MIPS16 PC-relative load instructions can access executable sections,
17237 but other instructions must not do so. This option is useful on 4KSc
17238 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17239 It is also useful on processors that can be configured to have a dual
17240 instruction/data SRAM interface and that, like the M4K, automatically
17241 redirect PC-relative loads to the instruction RAM.
17243 @item -mcode-readable=no
17244 Instructions must not access executable sections. This option can be
17245 useful on targets that are configured to have a dual instruction/data
17246 SRAM interface but that (unlike the M4K) do not automatically redirect
17247 PC-relative loads to the instruction RAM.
17250 @item -msplit-addresses
17251 @itemx -mno-split-addresses
17252 @opindex msplit-addresses
17253 @opindex mno-split-addresses
17254 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17255 relocation operators. This option has been superseded by
17256 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17258 @item -mexplicit-relocs
17259 @itemx -mno-explicit-relocs
17260 @opindex mexplicit-relocs
17261 @opindex mno-explicit-relocs
17262 Use (do not use) assembler relocation operators when dealing with symbolic
17263 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17264 is to use assembler macros instead.
17266 @option{-mexplicit-relocs} is the default if GCC was configured
17267 to use an assembler that supports relocation operators.
17269 @item -mcheck-zero-division
17270 @itemx -mno-check-zero-division
17271 @opindex mcheck-zero-division
17272 @opindex mno-check-zero-division
17273 Trap (do not trap) on integer division by zero.
17275 The default is @option{-mcheck-zero-division}.
17277 @item -mdivide-traps
17278 @itemx -mdivide-breaks
17279 @opindex mdivide-traps
17280 @opindex mdivide-breaks
17281 MIPS systems check for division by zero by generating either a
17282 conditional trap or a break instruction. Using traps results in
17283 smaller code, but is only supported on MIPS II and later. Also, some
17284 versions of the Linux kernel have a bug that prevents trap from
17285 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17286 allow conditional traps on architectures that support them and
17287 @option{-mdivide-breaks} to force the use of breaks.
17289 The default is usually @option{-mdivide-traps}, but this can be
17290 overridden at configure time using @option{--with-divide=breaks}.
17291 Divide-by-zero checks can be completely disabled using
17292 @option{-mno-check-zero-division}.
17297 @opindex mno-memcpy
17298 Force (do not force) the use of @code{memcpy} for non-trivial block
17299 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17300 most constant-sized copies.
17303 @itemx -mno-long-calls
17304 @opindex mlong-calls
17305 @opindex mno-long-calls
17306 Disable (do not disable) use of the @code{jal} instruction. Calling
17307 functions using @code{jal} is more efficient but requires the caller
17308 and callee to be in the same 256 megabyte segment.
17310 This option has no effect on abicalls code. The default is
17311 @option{-mno-long-calls}.
17317 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17318 instructions, as provided by the R4650 ISA@.
17324 Enable (disable) use of the @code{madd} and @code{msub} integer
17325 instructions. The default is @option{-mimadd} on architectures
17326 that support @code{madd} and @code{msub} except for the 74k
17327 architecture where it was found to generate slower code.
17330 @itemx -mno-fused-madd
17331 @opindex mfused-madd
17332 @opindex mno-fused-madd
17333 Enable (disable) use of the floating-point multiply-accumulate
17334 instructions, when they are available. The default is
17335 @option{-mfused-madd}.
17337 On the R8000 CPU when multiply-accumulate instructions are used,
17338 the intermediate product is calculated to infinite precision
17339 and is not subject to the FCSR Flush to Zero bit. This may be
17340 undesirable in some circumstances. On other processors the result
17341 is numerically identical to the equivalent computation using
17342 separate multiply, add, subtract and negate instructions.
17346 Tell the MIPS assembler to not run its preprocessor over user
17347 assembler files (with a @samp{.s} suffix) when assembling them.
17352 @opindex mno-fix-24k
17353 Work around the 24K E48 (lost data on stores during refill) errata.
17354 The workarounds are implemented by the assembler rather than by GCC@.
17357 @itemx -mno-fix-r4000
17358 @opindex mfix-r4000
17359 @opindex mno-fix-r4000
17360 Work around certain R4000 CPU errata:
17363 A double-word or a variable shift may give an incorrect result if executed
17364 immediately after starting an integer division.
17366 A double-word or a variable shift may give an incorrect result if executed
17367 while an integer multiplication is in progress.
17369 An integer division may give an incorrect result if started in a delay slot
17370 of a taken branch or a jump.
17374 @itemx -mno-fix-r4400
17375 @opindex mfix-r4400
17376 @opindex mno-fix-r4400
17377 Work around certain R4400 CPU errata:
17380 A double-word or a variable shift may give an incorrect result if executed
17381 immediately after starting an integer division.
17385 @itemx -mno-fix-r10000
17386 @opindex mfix-r10000
17387 @opindex mno-fix-r10000
17388 Work around certain R10000 errata:
17391 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17392 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17395 This option can only be used if the target architecture supports
17396 branch-likely instructions. @option{-mfix-r10000} is the default when
17397 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17401 @itemx -mno-fix-rm7000
17402 @opindex mfix-rm7000
17403 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17404 workarounds are implemented by the assembler rather than by GCC@.
17407 @itemx -mno-fix-vr4120
17408 @opindex mfix-vr4120
17409 Work around certain VR4120 errata:
17412 @code{dmultu} does not always produce the correct result.
17414 @code{div} and @code{ddiv} do not always produce the correct result if one
17415 of the operands is negative.
17417 The workarounds for the division errata rely on special functions in
17418 @file{libgcc.a}. At present, these functions are only provided by
17419 the @code{mips64vr*-elf} configurations.
17421 Other VR4120 errata require a NOP to be inserted between certain pairs of
17422 instructions. These errata are handled by the assembler, not by GCC itself.
17425 @opindex mfix-vr4130
17426 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17427 workarounds are implemented by the assembler rather than by GCC,
17428 although GCC avoids using @code{mflo} and @code{mfhi} if the
17429 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17430 instructions are available instead.
17433 @itemx -mno-fix-sb1
17435 Work around certain SB-1 CPU core errata.
17436 (This flag currently works around the SB-1 revision 2
17437 ``F1'' and ``F2'' floating-point errata.)
17439 @item -mr10k-cache-barrier=@var{setting}
17440 @opindex mr10k-cache-barrier
17441 Specify whether GCC should insert cache barriers to avoid the
17442 side-effects of speculation on R10K processors.
17444 In common with many processors, the R10K tries to predict the outcome
17445 of a conditional branch and speculatively executes instructions from
17446 the ``taken'' branch. It later aborts these instructions if the
17447 predicted outcome is wrong. However, on the R10K, even aborted
17448 instructions can have side effects.
17450 This problem only affects kernel stores and, depending on the system,
17451 kernel loads. As an example, a speculatively-executed store may load
17452 the target memory into cache and mark the cache line as dirty, even if
17453 the store itself is later aborted. If a DMA operation writes to the
17454 same area of memory before the ``dirty'' line is flushed, the cached
17455 data overwrites the DMA-ed data. See the R10K processor manual
17456 for a full description, including other potential problems.
17458 One workaround is to insert cache barrier instructions before every memory
17459 access that might be speculatively executed and that might have side
17460 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17461 controls GCC's implementation of this workaround. It assumes that
17462 aborted accesses to any byte in the following regions does not have
17467 the memory occupied by the current function's stack frame;
17470 the memory occupied by an incoming stack argument;
17473 the memory occupied by an object with a link-time-constant address.
17476 It is the kernel's responsibility to ensure that speculative
17477 accesses to these regions are indeed safe.
17479 If the input program contains a function declaration such as:
17485 then the implementation of @code{foo} must allow @code{j foo} and
17486 @code{jal foo} to be executed speculatively. GCC honors this
17487 restriction for functions it compiles itself. It expects non-GCC
17488 functions (such as hand-written assembly code) to do the same.
17490 The option has three forms:
17493 @item -mr10k-cache-barrier=load-store
17494 Insert a cache barrier before a load or store that might be
17495 speculatively executed and that might have side effects even
17498 @item -mr10k-cache-barrier=store
17499 Insert a cache barrier before a store that might be speculatively
17500 executed and that might have side effects even if aborted.
17502 @item -mr10k-cache-barrier=none
17503 Disable the insertion of cache barriers. This is the default setting.
17506 @item -mflush-func=@var{func}
17507 @itemx -mno-flush-func
17508 @opindex mflush-func
17509 Specifies the function to call to flush the I and D caches, or to not
17510 call any such function. If called, the function must take the same
17511 arguments as the common @code{_flush_func}, that is, the address of the
17512 memory range for which the cache is being flushed, the size of the
17513 memory range, and the number 3 (to flush both caches). The default
17514 depends on the target GCC was configured for, but commonly is either
17515 @code{_flush_func} or @code{__cpu_flush}.
17517 @item mbranch-cost=@var{num}
17518 @opindex mbranch-cost
17519 Set the cost of branches to roughly @var{num} ``simple'' instructions.
17520 This cost is only a heuristic and is not guaranteed to produce
17521 consistent results across releases. A zero cost redundantly selects
17522 the default, which is based on the @option{-mtune} setting.
17524 @item -mbranch-likely
17525 @itemx -mno-branch-likely
17526 @opindex mbranch-likely
17527 @opindex mno-branch-likely
17528 Enable or disable use of Branch Likely instructions, regardless of the
17529 default for the selected architecture. By default, Branch Likely
17530 instructions may be generated if they are supported by the selected
17531 architecture. An exception is for the MIPS32 and MIPS64 architectures
17532 and processors that implement those architectures; for those, Branch
17533 Likely instructions are not be generated by default because the MIPS32
17534 and MIPS64 architectures specifically deprecate their use.
17536 @item -mfp-exceptions
17537 @itemx -mno-fp-exceptions
17538 @opindex mfp-exceptions
17539 Specifies whether FP exceptions are enabled. This affects how
17540 FP instructions are scheduled for some processors.
17541 The default is that FP exceptions are
17544 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
17545 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
17548 @item -mvr4130-align
17549 @itemx -mno-vr4130-align
17550 @opindex mvr4130-align
17551 The VR4130 pipeline is two-way superscalar, but can only issue two
17552 instructions together if the first one is 8-byte aligned. When this
17553 option is enabled, GCC aligns pairs of instructions that it
17554 thinks should execute in parallel.
17556 This option only has an effect when optimizing for the VR4130.
17557 It normally makes code faster, but at the expense of making it bigger.
17558 It is enabled by default at optimization level @option{-O3}.
17563 Enable (disable) generation of @code{synci} instructions on
17564 architectures that support it. The @code{synci} instructions (if
17565 enabled) are generated when @code{__builtin___clear_cache} is
17568 This option defaults to @option{-mno-synci}, but the default can be
17569 overridden by configuring GCC with @option{--with-synci}.
17571 When compiling code for single processor systems, it is generally safe
17572 to use @code{synci}. However, on many multi-core (SMP) systems, it
17573 does not invalidate the instruction caches on all cores and may lead
17574 to undefined behavior.
17576 @item -mrelax-pic-calls
17577 @itemx -mno-relax-pic-calls
17578 @opindex mrelax-pic-calls
17579 Try to turn PIC calls that are normally dispatched via register
17580 @code{$25} into direct calls. This is only possible if the linker can
17581 resolve the destination at link-time and if the destination is within
17582 range for a direct call.
17584 @option{-mrelax-pic-calls} is the default if GCC was configured to use
17585 an assembler and a linker that support the @code{.reloc} assembly
17586 directive and @option{-mexplicit-relocs} is in effect. With
17587 @option{-mno-explicit-relocs}, this optimization can be performed by the
17588 assembler and the linker alone without help from the compiler.
17590 @item -mmcount-ra-address
17591 @itemx -mno-mcount-ra-address
17592 @opindex mmcount-ra-address
17593 @opindex mno-mcount-ra-address
17594 Emit (do not emit) code that allows @code{_mcount} to modify the
17595 calling function's return address. When enabled, this option extends
17596 the usual @code{_mcount} interface with a new @var{ra-address}
17597 parameter, which has type @code{intptr_t *} and is passed in register
17598 @code{$12}. @code{_mcount} can then modify the return address by
17599 doing both of the following:
17602 Returning the new address in register @code{$31}.
17604 Storing the new address in @code{*@var{ra-address}},
17605 if @var{ra-address} is nonnull.
17608 The default is @option{-mno-mcount-ra-address}.
17613 @subsection MMIX Options
17614 @cindex MMIX Options
17616 These options are defined for the MMIX:
17620 @itemx -mno-libfuncs
17622 @opindex mno-libfuncs
17623 Specify that intrinsic library functions are being compiled, passing all
17624 values in registers, no matter the size.
17627 @itemx -mno-epsilon
17629 @opindex mno-epsilon
17630 Generate floating-point comparison instructions that compare with respect
17631 to the @code{rE} epsilon register.
17633 @item -mabi=mmixware
17635 @opindex mabi=mmixware
17637 Generate code that passes function parameters and return values that (in
17638 the called function) are seen as registers @code{$0} and up, as opposed to
17639 the GNU ABI which uses global registers @code{$231} and up.
17641 @item -mzero-extend
17642 @itemx -mno-zero-extend
17643 @opindex mzero-extend
17644 @opindex mno-zero-extend
17645 When reading data from memory in sizes shorter than 64 bits, use (do not
17646 use) zero-extending load instructions by default, rather than
17647 sign-extending ones.
17650 @itemx -mno-knuthdiv
17652 @opindex mno-knuthdiv
17653 Make the result of a division yielding a remainder have the same sign as
17654 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
17655 remainder follows the sign of the dividend. Both methods are
17656 arithmetically valid, the latter being almost exclusively used.
17658 @item -mtoplevel-symbols
17659 @itemx -mno-toplevel-symbols
17660 @opindex mtoplevel-symbols
17661 @opindex mno-toplevel-symbols
17662 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
17663 code can be used with the @code{PREFIX} assembly directive.
17667 Generate an executable in the ELF format, rather than the default
17668 @samp{mmo} format used by the @command{mmix} simulator.
17670 @item -mbranch-predict
17671 @itemx -mno-branch-predict
17672 @opindex mbranch-predict
17673 @opindex mno-branch-predict
17674 Use (do not use) the probable-branch instructions, when static branch
17675 prediction indicates a probable branch.
17677 @item -mbase-addresses
17678 @itemx -mno-base-addresses
17679 @opindex mbase-addresses
17680 @opindex mno-base-addresses
17681 Generate (do not generate) code that uses @emph{base addresses}. Using a
17682 base address automatically generates a request (handled by the assembler
17683 and the linker) for a constant to be set up in a global register. The
17684 register is used for one or more base address requests within the range 0
17685 to 255 from the value held in the register. The generally leads to short
17686 and fast code, but the number of different data items that can be
17687 addressed is limited. This means that a program that uses lots of static
17688 data may require @option{-mno-base-addresses}.
17690 @item -msingle-exit
17691 @itemx -mno-single-exit
17692 @opindex msingle-exit
17693 @opindex mno-single-exit
17694 Force (do not force) generated code to have a single exit point in each
17698 @node MN10300 Options
17699 @subsection MN10300 Options
17700 @cindex MN10300 options
17702 These @option{-m} options are defined for Matsushita MN10300 architectures:
17707 Generate code to avoid bugs in the multiply instructions for the MN10300
17708 processors. This is the default.
17710 @item -mno-mult-bug
17711 @opindex mno-mult-bug
17712 Do not generate code to avoid bugs in the multiply instructions for the
17713 MN10300 processors.
17717 Generate code using features specific to the AM33 processor.
17721 Do not generate code using features specific to the AM33 processor. This
17726 Generate code using features specific to the AM33/2.0 processor.
17730 Generate code using features specific to the AM34 processor.
17732 @item -mtune=@var{cpu-type}
17734 Use the timing characteristics of the indicated CPU type when
17735 scheduling instructions. This does not change the targeted processor
17736 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
17737 @samp{am33-2} or @samp{am34}.
17739 @item -mreturn-pointer-on-d0
17740 @opindex mreturn-pointer-on-d0
17741 When generating a function that returns a pointer, return the pointer
17742 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
17743 only in @code{a0}, and attempts to call such functions without a prototype
17744 result in errors. Note that this option is on by default; use
17745 @option{-mno-return-pointer-on-d0} to disable it.
17749 Do not link in the C run-time initialization object file.
17753 Indicate to the linker that it should perform a relaxation optimization pass
17754 to shorten branches, calls and absolute memory addresses. This option only
17755 has an effect when used on the command line for the final link step.
17757 This option makes symbolic debugging impossible.
17761 Allow the compiler to generate @emph{Long Instruction Word}
17762 instructions if the target is the @samp{AM33} or later. This is the
17763 default. This option defines the preprocessor macro @code{__LIW__}.
17767 Do not allow the compiler to generate @emph{Long Instruction Word}
17768 instructions. This option defines the preprocessor macro
17773 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
17774 instructions if the target is the @samp{AM33} or later. This is the
17775 default. This option defines the preprocessor macro @code{__SETLB__}.
17779 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
17780 instructions. This option defines the preprocessor macro
17781 @code{__NO_SETLB__}.
17785 @node Moxie Options
17786 @subsection Moxie Options
17787 @cindex Moxie Options
17793 Generate big-endian code. This is the default for @samp{moxie-*-*}
17798 Generate little-endian code.
17802 Generate mul.x and umul.x instructions. This is the default for
17803 @samp{moxiebox-*-*} configurations.
17807 Do not link in the C run-time initialization object file.
17811 @node MSP430 Options
17812 @subsection MSP430 Options
17813 @cindex MSP430 Options
17815 These options are defined for the MSP430:
17821 Force assembly output to always use hex constants. Normally such
17822 constants are signed decimals, but this option is available for
17823 testsuite and/or aesthetic purposes.
17827 Select the MCU to target. This is used to create a C preprocessor
17828 symbol based upon the MCU name, converted to upper case and pre- and
17829 post-fixed with @samp{__}. This in turn is used by the
17830 @file{msp430.h} header file to select an MCU-specific supplementary
17833 The option also sets the ISA to use. If the MCU name is one that is
17834 known to only support the 430 ISA then that is selected, otherwise the
17835 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
17836 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
17837 name selects the 430X ISA.
17839 In addition an MCU-specific linker script is added to the linker
17840 command line. The script's name is the name of the MCU with
17841 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
17842 command line defines the C preprocessor symbol @code{__XXX__} and
17843 cause the linker to search for a script called @file{xxx.ld}.
17845 This option is also passed on to the assembler.
17849 Specifies the ISA to use. Accepted values are @samp{msp430},
17850 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
17851 @option{-mmcu=} option should be used to select the ISA.
17855 Link to the simulator runtime libraries and linker script. Overrides
17856 any scripts that would be selected by the @option{-mmcu=} option.
17860 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
17864 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
17868 This option is passed to the assembler and linker, and allows the
17869 linker to perform certain optimizations that cannot be done until
17874 Describes the type of hardware multiply supported by the target.
17875 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
17876 for the original 16-bit-only multiply supported by early MCUs.
17877 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
17878 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
17879 A value of @samp{auto} can also be given. This tells GCC to deduce
17880 the hardware multiply support based upon the MCU name provided by the
17881 @option{-mmcu} option. If no @option{-mmcu} option is specified then
17882 @samp{32bit} hardware multiply support is assumed. @samp{auto} is the
17885 Hardware multiplies are normally performed by calling a library
17886 routine. This saves space in the generated code. When compiling at
17887 @option{-O3} or higher however the hardware multiplier is invoked
17888 inline. This makes for bigger, but faster code.
17890 The hardware multiply routines disable interrupts whilst running and
17891 restore the previous interrupt state when they finish. This makes
17892 them safe to use inside interrupt handlers as well as in normal code.
17896 Enable the use of a minimum runtime environment - no static
17897 initializers or constructors. This is intended for memory-constrained
17898 devices. The compiler includes special symbols in some objects
17899 that tell the linker and runtime which code fragments are required.
17903 @node NDS32 Options
17904 @subsection NDS32 Options
17905 @cindex NDS32 Options
17907 These options are defined for NDS32 implementations:
17912 @opindex mbig-endian
17913 Generate code in big-endian mode.
17915 @item -mlittle-endian
17916 @opindex mlittle-endian
17917 Generate code in little-endian mode.
17919 @item -mreduced-regs
17920 @opindex mreduced-regs
17921 Use reduced-set registers for register allocation.
17924 @opindex mfull-regs
17925 Use full-set registers for register allocation.
17929 Generate conditional move instructions.
17933 Do not generate conditional move instructions.
17937 Generate performance extension instructions.
17939 @item -mno-perf-ext
17940 @opindex mno-perf-ext
17941 Do not generate performance extension instructions.
17945 Generate v3 push25/pop25 instructions.
17948 @opindex mno-v3push
17949 Do not generate v3 push25/pop25 instructions.
17953 Generate 16-bit instructions.
17956 @opindex mno-16-bit
17957 Do not generate 16-bit instructions.
17959 @item -misr-vector-size=@var{num}
17960 @opindex misr-vector-size
17961 Specify the size of each interrupt vector, which must be 4 or 16.
17963 @item -mcache-block-size=@var{num}
17964 @opindex mcache-block-size
17965 Specify the size of each cache block,
17966 which must be a power of 2 between 4 and 512.
17968 @item -march=@var{arch}
17970 Specify the name of the target architecture.
17972 @item -mcmodel=@var{code-model}
17974 Set the code model to one of
17977 All the data and read-only data segments must be within 512KB addressing space.
17978 The text segment must be within 16MB addressing space.
17979 @item @samp{medium}
17980 The data segment must be within 512KB while the read-only data segment can be
17981 within 4GB addressing space. The text segment should be still within 16MB
17984 All the text and data segments can be within 4GB addressing space.
17988 @opindex mctor-dtor
17989 Enable constructor/destructor feature.
17993 Guide linker to relax instructions.
17997 @node Nios II Options
17998 @subsection Nios II Options
17999 @cindex Nios II options
18000 @cindex Altera Nios II options
18002 These are the options defined for the Altera Nios II processor.
18008 @cindex smaller data references
18009 Put global and static objects less than or equal to @var{num} bytes
18010 into the small data or BSS sections instead of the normal data or BSS
18011 sections. The default value of @var{num} is 8.
18013 @item -mgpopt=@var{option}
18018 Generate (do not generate) GP-relative accesses. The following
18019 @var{option} names are recognized:
18024 Do not generate GP-relative accesses.
18027 Generate GP-relative accesses for small data objects that are not
18028 external or weak. Also use GP-relative addressing for objects that
18029 have been explicitly placed in a small data section via a @code{section}
18033 As for @samp{local}, but also generate GP-relative accesses for
18034 small data objects that are external or weak. If you use this option,
18035 you must ensure that all parts of your program (including libraries) are
18036 compiled with the same @option{-G} setting.
18039 Generate GP-relative accesses for all data objects in the program. If you
18040 use this option, the entire data and BSS segments
18041 of your program must fit in 64K of memory and you must use an appropriate
18042 linker script to allocate them within the addressible range of the
18046 Generate GP-relative addresses for function pointers as well as data
18047 pointers. If you use this option, the entire text, data, and BSS segments
18048 of your program must fit in 64K of memory and you must use an appropriate
18049 linker script to allocate them within the addressible range of the
18054 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18055 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18057 The default is @option{-mgpopt} except when @option{-fpic} or
18058 @option{-fPIC} is specified to generate position-independent code.
18059 Note that the Nios II ABI does not permit GP-relative accesses from
18062 You may need to specify @option{-mno-gpopt} explicitly when building
18063 programs that include large amounts of small data, including large
18064 GOT data sections. In this case, the 16-bit offset for GP-relative
18065 addressing may not be large enough to allow access to the entire
18066 small data section.
18072 Generate little-endian (default) or big-endian (experimental) code,
18075 @item -mbypass-cache
18076 @itemx -mno-bypass-cache
18077 @opindex mno-bypass-cache
18078 @opindex mbypass-cache
18079 Force all load and store instructions to always bypass cache by
18080 using I/O variants of the instructions. The default is not to
18083 @item -mno-cache-volatile
18084 @itemx -mcache-volatile
18085 @opindex mcache-volatile
18086 @opindex mno-cache-volatile
18087 Volatile memory access bypass the cache using the I/O variants of
18088 the load and store instructions. The default is not to bypass the cache.
18090 @item -mno-fast-sw-div
18091 @itemx -mfast-sw-div
18092 @opindex mno-fast-sw-div
18093 @opindex mfast-sw-div
18094 Do not use table-based fast divide for small numbers. The default
18095 is to use the fast divide at @option{-O3} and above.
18099 @itemx -mno-hw-mulx
18103 @opindex mno-hw-mul
18105 @opindex mno-hw-mulx
18107 @opindex mno-hw-div
18109 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18110 instructions by the compiler. The default is to emit @code{mul}
18111 and not emit @code{div} and @code{mulx}.
18113 @item -mcustom-@var{insn}=@var{N}
18114 @itemx -mno-custom-@var{insn}
18115 @opindex mcustom-@var{insn}
18116 @opindex mno-custom-@var{insn}
18117 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18118 custom instruction with encoding @var{N} when generating code that uses
18119 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18120 instruction 253 for single-precision floating-point add operations instead
18121 of the default behavior of using a library call.
18123 The following values of @var{insn} are supported. Except as otherwise
18124 noted, floating-point operations are expected to be implemented with
18125 normal IEEE 754 semantics and correspond directly to the C operators or the
18126 equivalent GCC built-in functions (@pxref{Other Builtins}).
18128 Single-precision floating point:
18131 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18132 Binary arithmetic operations.
18138 Unary absolute value.
18140 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18141 Comparison operations.
18143 @item @samp{fmins}, @samp{fmaxs}
18144 Floating-point minimum and maximum. These instructions are only
18145 generated if @option{-ffinite-math-only} is specified.
18147 @item @samp{fsqrts}
18148 Unary square root operation.
18150 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18151 Floating-point trigonometric and exponential functions. These instructions
18152 are only generated if @option{-funsafe-math-optimizations} is also specified.
18156 Double-precision floating point:
18159 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18160 Binary arithmetic operations.
18166 Unary absolute value.
18168 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18169 Comparison operations.
18171 @item @samp{fmind}, @samp{fmaxd}
18172 Double-precision minimum and maximum. These instructions are only
18173 generated if @option{-ffinite-math-only} is specified.
18175 @item @samp{fsqrtd}
18176 Unary square root operation.
18178 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18179 Double-precision trigonometric and exponential functions. These instructions
18180 are only generated if @option{-funsafe-math-optimizations} is also specified.
18186 @item @samp{fextsd}
18187 Conversion from single precision to double precision.
18189 @item @samp{ftruncds}
18190 Conversion from double precision to single precision.
18192 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18193 Conversion from floating point to signed or unsigned integer types, with
18194 truncation towards zero.
18197 Conversion from single-precision floating point to signed integer,
18198 rounding to the nearest integer and ties away from zero.
18199 This corresponds to the @code{__builtin_lroundf} function when
18200 @option{-fno-math-errno} is used.
18202 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18203 Conversion from signed or unsigned integer types to floating-point types.
18207 In addition, all of the following transfer instructions for internal
18208 registers X and Y must be provided to use any of the double-precision
18209 floating-point instructions. Custom instructions taking two
18210 double-precision source operands expect the first operand in the
18211 64-bit register X. The other operand (or only operand of a unary
18212 operation) is given to the custom arithmetic instruction with the
18213 least significant half in source register @var{src1} and the most
18214 significant half in @var{src2}. A custom instruction that returns a
18215 double-precision result returns the most significant 32 bits in the
18216 destination register and the other half in 32-bit register Y.
18217 GCC automatically generates the necessary code sequences to write
18218 register X and/or read register Y when double-precision floating-point
18219 instructions are used.
18224 Write @var{src1} into the least significant half of X and @var{src2} into
18225 the most significant half of X.
18228 Write @var{src1} into Y.
18230 @item @samp{frdxhi}, @samp{frdxlo}
18231 Read the most or least (respectively) significant half of X and store it in
18235 Read the value of Y and store it into @var{dest}.
18238 Note that you can gain more local control over generation of Nios II custom
18239 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18240 and @code{target("no-custom-@var{insn}")} function attributes
18241 (@pxref{Function Attributes})
18242 or pragmas (@pxref{Function Specific Option Pragmas}).
18244 @item -mcustom-fpu-cfg=@var{name}
18245 @opindex mcustom-fpu-cfg
18247 This option enables a predefined, named set of custom instruction encodings
18248 (see @option{-mcustom-@var{insn}} above).
18249 Currently, the following sets are defined:
18251 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18252 @gccoptlist{-mcustom-fmuls=252 @gol
18253 -mcustom-fadds=253 @gol
18254 -mcustom-fsubs=254 @gol
18255 -fsingle-precision-constant}
18257 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18258 @gccoptlist{-mcustom-fmuls=252 @gol
18259 -mcustom-fadds=253 @gol
18260 -mcustom-fsubs=254 @gol
18261 -mcustom-fdivs=255 @gol
18262 -fsingle-precision-constant}
18264 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18265 @gccoptlist{-mcustom-floatus=243 @gol
18266 -mcustom-fixsi=244 @gol
18267 -mcustom-floatis=245 @gol
18268 -mcustom-fcmpgts=246 @gol
18269 -mcustom-fcmples=249 @gol
18270 -mcustom-fcmpeqs=250 @gol
18271 -mcustom-fcmpnes=251 @gol
18272 -mcustom-fmuls=252 @gol
18273 -mcustom-fadds=253 @gol
18274 -mcustom-fsubs=254 @gol
18275 -mcustom-fdivs=255 @gol
18276 -fsingle-precision-constant}
18278 Custom instruction assignments given by individual
18279 @option{-mcustom-@var{insn}=} options override those given by
18280 @option{-mcustom-fpu-cfg=}, regardless of the
18281 order of the options on the command line.
18283 Note that you can gain more local control over selection of a FPU
18284 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18285 function attribute (@pxref{Function Attributes})
18286 or pragma (@pxref{Function Specific Option Pragmas}).
18290 These additional @samp{-m} options are available for the Altera Nios II
18291 ELF (bare-metal) target:
18297 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18298 startup and termination code, and is typically used in conjunction with
18299 @option{-msys-crt0=} to specify the location of the alternate startup code
18300 provided by the HAL BSP.
18304 Link with a limited version of the C library, @option{-lsmallc}, rather than
18307 @item -msys-crt0=@var{startfile}
18309 @var{startfile} is the file name of the startfile (crt0) to use
18310 when linking. This option is only useful in conjunction with @option{-mhal}.
18312 @item -msys-lib=@var{systemlib}
18314 @var{systemlib} is the library name of the library that provides
18315 low-level system calls required by the C library,
18316 e.g. @code{read} and @code{write}.
18317 This option is typically used to link with a library provided by a HAL BSP.
18321 @node Nvidia PTX Options
18322 @subsection Nvidia PTX Options
18323 @cindex Nvidia PTX options
18324 @cindex nvptx options
18326 These options are defined for Nvidia PTX:
18334 Generate code for 32-bit or 64-bit ABI.
18337 @opindex mmainkernel
18338 Link in code for a __main kernel. This is for stand-alone instead of
18339 offloading execution.
18343 @node PDP-11 Options
18344 @subsection PDP-11 Options
18345 @cindex PDP-11 Options
18347 These options are defined for the PDP-11:
18352 Use hardware FPP floating point. This is the default. (FIS floating
18353 point on the PDP-11/40 is not supported.)
18356 @opindex msoft-float
18357 Do not use hardware floating point.
18361 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18365 Return floating-point results in memory. This is the default.
18369 Generate code for a PDP-11/40.
18373 Generate code for a PDP-11/45. This is the default.
18377 Generate code for a PDP-11/10.
18379 @item -mbcopy-builtin
18380 @opindex mbcopy-builtin
18381 Use inline @code{movmemhi} patterns for copying memory. This is the
18386 Do not use inline @code{movmemhi} patterns for copying memory.
18392 Use 16-bit @code{int}. This is the default.
18398 Use 32-bit @code{int}.
18401 @itemx -mno-float32
18403 @opindex mno-float32
18404 Use 64-bit @code{float}. This is the default.
18407 @itemx -mno-float64
18409 @opindex mno-float64
18410 Use 32-bit @code{float}.
18414 Use @code{abshi2} pattern. This is the default.
18418 Do not use @code{abshi2} pattern.
18420 @item -mbranch-expensive
18421 @opindex mbranch-expensive
18422 Pretend that branches are expensive. This is for experimenting with
18423 code generation only.
18425 @item -mbranch-cheap
18426 @opindex mbranch-cheap
18427 Do not pretend that branches are expensive. This is the default.
18431 Use Unix assembler syntax. This is the default when configured for
18432 @samp{pdp11-*-bsd}.
18436 Use DEC assembler syntax. This is the default when configured for any
18437 PDP-11 target other than @samp{pdp11-*-bsd}.
18440 @node picoChip Options
18441 @subsection picoChip Options
18442 @cindex picoChip options
18444 These @samp{-m} options are defined for picoChip implementations:
18448 @item -mae=@var{ae_type}
18450 Set the instruction set, register set, and instruction scheduling
18451 parameters for array element type @var{ae_type}. Supported values
18452 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18454 @option{-mae=ANY} selects a completely generic AE type. Code
18455 generated with this option runs on any of the other AE types. The
18456 code is not as efficient as it would be if compiled for a specific
18457 AE type, and some types of operation (e.g., multiplication) do not
18458 work properly on all types of AE.
18460 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
18461 for compiled code, and is the default.
18463 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
18464 option may suffer from poor performance of byte (char) manipulation,
18465 since the DSP AE does not provide hardware support for byte load/stores.
18467 @item -msymbol-as-address
18468 Enable the compiler to directly use a symbol name as an address in a
18469 load/store instruction, without first loading it into a
18470 register. Typically, the use of this option generates larger
18471 programs, which run faster than when the option isn't used. However, the
18472 results vary from program to program, so it is left as a user option,
18473 rather than being permanently enabled.
18475 @item -mno-inefficient-warnings
18476 Disables warnings about the generation of inefficient code. These
18477 warnings can be generated, for example, when compiling code that
18478 performs byte-level memory operations on the MAC AE type. The MAC AE has
18479 no hardware support for byte-level memory operations, so all byte
18480 load/stores must be synthesized from word load/store operations. This is
18481 inefficient and a warning is generated to indicate
18482 that you should rewrite the code to avoid byte operations, or to target
18483 an AE type that has the necessary hardware support. This option disables
18488 @node PowerPC Options
18489 @subsection PowerPC Options
18490 @cindex PowerPC options
18492 These are listed under @xref{RS/6000 and PowerPC Options}.
18495 @subsection RL78 Options
18496 @cindex RL78 Options
18502 Links in additional target libraries to support operation within a
18509 Specifies the type of hardware multiplication support to be used. The
18510 default is @samp{none}, which uses software multiplication functions.
18511 The @samp{g13} option is for the hardware multiply/divide peripheral
18512 only on the RL78/G13 targets. The @samp{rl78} option is for the
18513 standard hardware multiplication defined in the RL78 software manual.
18515 @item -m64bit-doubles
18516 @itemx -m32bit-doubles
18517 @opindex m64bit-doubles
18518 @opindex m32bit-doubles
18519 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
18520 or 32 bits (@option{-m32bit-doubles}) in size. The default is
18521 @option{-m32bit-doubles}.
18525 @node RS/6000 and PowerPC Options
18526 @subsection IBM RS/6000 and PowerPC Options
18527 @cindex RS/6000 and PowerPC Options
18528 @cindex IBM RS/6000 and PowerPC Options
18530 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
18532 @item -mpowerpc-gpopt
18533 @itemx -mno-powerpc-gpopt
18534 @itemx -mpowerpc-gfxopt
18535 @itemx -mno-powerpc-gfxopt
18538 @itemx -mno-powerpc64
18542 @itemx -mno-popcntb
18544 @itemx -mno-popcntd
18553 @itemx -mno-hard-dfp
18554 @opindex mpowerpc-gpopt
18555 @opindex mno-powerpc-gpopt
18556 @opindex mpowerpc-gfxopt
18557 @opindex mno-powerpc-gfxopt
18558 @opindex mpowerpc64
18559 @opindex mno-powerpc64
18563 @opindex mno-popcntb
18565 @opindex mno-popcntd
18571 @opindex mno-mfpgpr
18573 @opindex mno-hard-dfp
18574 You use these options to specify which instructions are available on the
18575 processor you are using. The default value of these options is
18576 determined when configuring GCC@. Specifying the
18577 @option{-mcpu=@var{cpu_type}} overrides the specification of these
18578 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
18579 rather than the options listed above.
18581 Specifying @option{-mpowerpc-gpopt} allows
18582 GCC to use the optional PowerPC architecture instructions in the
18583 General Purpose group, including floating-point square root. Specifying
18584 @option{-mpowerpc-gfxopt} allows GCC to
18585 use the optional PowerPC architecture instructions in the Graphics
18586 group, including floating-point select.
18588 The @option{-mmfcrf} option allows GCC to generate the move from
18589 condition register field instruction implemented on the POWER4
18590 processor and other processors that support the PowerPC V2.01
18592 The @option{-mpopcntb} option allows GCC to generate the popcount and
18593 double-precision FP reciprocal estimate instruction implemented on the
18594 POWER5 processor and other processors that support the PowerPC V2.02
18596 The @option{-mpopcntd} option allows GCC to generate the popcount
18597 instruction implemented on the POWER7 processor and other processors
18598 that support the PowerPC V2.06 architecture.
18599 The @option{-mfprnd} option allows GCC to generate the FP round to
18600 integer instructions implemented on the POWER5+ processor and other
18601 processors that support the PowerPC V2.03 architecture.
18602 The @option{-mcmpb} option allows GCC to generate the compare bytes
18603 instruction implemented on the POWER6 processor and other processors
18604 that support the PowerPC V2.05 architecture.
18605 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
18606 general-purpose register instructions implemented on the POWER6X
18607 processor and other processors that support the extended PowerPC V2.05
18609 The @option{-mhard-dfp} option allows GCC to generate the decimal
18610 floating-point instructions implemented on some POWER processors.
18612 The @option{-mpowerpc64} option allows GCC to generate the additional
18613 64-bit instructions that are found in the full PowerPC64 architecture
18614 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
18615 @option{-mno-powerpc64}.
18617 @item -mcpu=@var{cpu_type}
18619 Set architecture type, register usage, and
18620 instruction scheduling parameters for machine type @var{cpu_type}.
18621 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
18622 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
18623 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
18624 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
18625 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
18626 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
18627 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
18628 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
18629 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
18630 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
18631 @samp{powerpc64}, @samp{powerpc64le}, and @samp{rs64}.
18633 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
18634 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
18635 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
18636 architecture machine types, with an appropriate, generic processor
18637 model assumed for scheduling purposes.
18639 The other options specify a specific processor. Code generated under
18640 those options runs best on that processor, and may not run at all on
18643 The @option{-mcpu} options automatically enable or disable the
18646 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
18647 -mpopcntb -mpopcntd -mpowerpc64 @gol
18648 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
18649 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
18650 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
18651 -mquad-memory -mquad-memory-atomic}
18653 The particular options set for any particular CPU varies between
18654 compiler versions, depending on what setting seems to produce optimal
18655 code for that CPU; it doesn't necessarily reflect the actual hardware's
18656 capabilities. If you wish to set an individual option to a particular
18657 value, you may specify it after the @option{-mcpu} option, like
18658 @option{-mcpu=970 -mno-altivec}.
18660 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
18661 not enabled or disabled by the @option{-mcpu} option at present because
18662 AIX does not have full support for these options. You may still
18663 enable or disable them individually if you're sure it'll work in your
18666 @item -mtune=@var{cpu_type}
18668 Set the instruction scheduling parameters for machine type
18669 @var{cpu_type}, but do not set the architecture type or register usage,
18670 as @option{-mcpu=@var{cpu_type}} does. The same
18671 values for @var{cpu_type} are used for @option{-mtune} as for
18672 @option{-mcpu}. If both are specified, the code generated uses the
18673 architecture and registers set by @option{-mcpu}, but the
18674 scheduling parameters set by @option{-mtune}.
18676 @item -mcmodel=small
18677 @opindex mcmodel=small
18678 Generate PowerPC64 code for the small model: The TOC is limited to
18681 @item -mcmodel=medium
18682 @opindex mcmodel=medium
18683 Generate PowerPC64 code for the medium model: The TOC and other static
18684 data may be up to a total of 4G in size.
18686 @item -mcmodel=large
18687 @opindex mcmodel=large
18688 Generate PowerPC64 code for the large model: The TOC may be up to 4G
18689 in size. Other data and code is only limited by the 64-bit address
18693 @itemx -mno-altivec
18695 @opindex mno-altivec
18696 Generate code that uses (does not use) AltiVec instructions, and also
18697 enable the use of built-in functions that allow more direct access to
18698 the AltiVec instruction set. You may also need to set
18699 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
18702 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
18703 @option{-maltivec=be}, the element order for Altivec intrinsics such
18704 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
18705 match array element order corresponding to the endianness of the
18706 target. That is, element zero identifies the leftmost element in a
18707 vector register when targeting a big-endian platform, and identifies
18708 the rightmost element in a vector register when targeting a
18709 little-endian platform.
18712 @opindex maltivec=be
18713 Generate Altivec instructions using big-endian element order,
18714 regardless of whether the target is big- or little-endian. This is
18715 the default when targeting a big-endian platform.
18717 The element order is used to interpret element numbers in Altivec
18718 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18719 @code{vec_insert}. By default, these match array element order
18720 corresponding to the endianness for the target.
18723 @opindex maltivec=le
18724 Generate Altivec instructions using little-endian element order,
18725 regardless of whether the target is big- or little-endian. This is
18726 the default when targeting a little-endian platform. This option is
18727 currently ignored when targeting a big-endian platform.
18729 The element order is used to interpret element numbers in Altivec
18730 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
18731 @code{vec_insert}. By default, these match array element order
18732 corresponding to the endianness for the target.
18737 @opindex mno-vrsave
18738 Generate VRSAVE instructions when generating AltiVec code.
18740 @item -mgen-cell-microcode
18741 @opindex mgen-cell-microcode
18742 Generate Cell microcode instructions.
18744 @item -mwarn-cell-microcode
18745 @opindex mwarn-cell-microcode
18746 Warn when a Cell microcode instruction is emitted. An example
18747 of a Cell microcode instruction is a variable shift.
18750 @opindex msecure-plt
18751 Generate code that allows @command{ld} and @command{ld.so}
18752 to build executables and shared
18753 libraries with non-executable @code{.plt} and @code{.got} sections.
18755 32-bit SYSV ABI option.
18759 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
18761 requires @code{.plt} and @code{.got}
18762 sections that are both writable and executable.
18763 This is a PowerPC 32-bit SYSV ABI option.
18769 This switch enables or disables the generation of ISEL instructions.
18771 @item -misel=@var{yes/no}
18772 This switch has been deprecated. Use @option{-misel} and
18773 @option{-mno-isel} instead.
18779 This switch enables or disables the generation of SPE simd
18785 @opindex mno-paired
18786 This switch enables or disables the generation of PAIRED simd
18789 @item -mspe=@var{yes/no}
18790 This option has been deprecated. Use @option{-mspe} and
18791 @option{-mno-spe} instead.
18797 Generate code that uses (does not use) vector/scalar (VSX)
18798 instructions, and also enable the use of built-in functions that allow
18799 more direct access to the VSX instruction set.
18804 @opindex mno-crypto
18805 Enable the use (disable) of the built-in functions that allow direct
18806 access to the cryptographic instructions that were added in version
18807 2.07 of the PowerPC ISA.
18809 @item -mdirect-move
18810 @itemx -mno-direct-move
18811 @opindex mdirect-move
18812 @opindex mno-direct-move
18813 Generate code that uses (does not use) the instructions to move data
18814 between the general purpose registers and the vector/scalar (VSX)
18815 registers that were added in version 2.07 of the PowerPC ISA.
18817 @item -mpower8-fusion
18818 @itemx -mno-power8-fusion
18819 @opindex mpower8-fusion
18820 @opindex mno-power8-fusion
18821 Generate code that keeps (does not keeps) some integer operations
18822 adjacent so that the instructions can be fused together on power8 and
18825 @item -mpower8-vector
18826 @itemx -mno-power8-vector
18827 @opindex mpower8-vector
18828 @opindex mno-power8-vector
18829 Generate code that uses (does not use) the vector and scalar
18830 instructions that were added in version 2.07 of the PowerPC ISA. Also
18831 enable the use of built-in functions that allow more direct access to
18832 the vector instructions.
18834 @item -mquad-memory
18835 @itemx -mno-quad-memory
18836 @opindex mquad-memory
18837 @opindex mno-quad-memory
18838 Generate code that uses (does not use) the non-atomic quad word memory
18839 instructions. The @option{-mquad-memory} option requires use of
18842 @item -mquad-memory-atomic
18843 @itemx -mno-quad-memory-atomic
18844 @opindex mquad-memory-atomic
18845 @opindex mno-quad-memory-atomic
18846 Generate code that uses (does not use) the atomic quad word memory
18847 instructions. The @option{-mquad-memory-atomic} option requires use of
18850 @item -mupper-regs-df
18851 @itemx -mno-upper-regs-df
18852 @opindex mupper-regs-df
18853 @opindex mno-upper-regs-df
18854 Generate code that uses (does not use) the scalar double precision
18855 instructions that target all 64 registers in the vector/scalar
18856 floating point register set that were added in version 2.06 of the
18857 PowerPC ISA. The @option{-mupper-regs-df} turned on by default if you
18858 use either of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
18859 @option{-mvsx} options.
18861 @item -mupper-regs-sf
18862 @itemx -mno-upper-regs-sf
18863 @opindex mupper-regs-sf
18864 @opindex mno-upper-regs-sf
18865 Generate code that uses (does not use) the scalar single precision
18866 instructions that target all 64 registers in the vector/scalar
18867 floating point register set that were added in version 2.07 of the
18868 PowerPC ISA. The @option{-mupper-regs-sf} turned on by default if you
18869 use either of the @option{-mcpu=power8}, or @option{-mpower8-vector}
18873 @itemx -mno-upper-regs
18874 @opindex mupper-regs
18875 @opindex mno-upper-regs
18876 Generate code that uses (does not use) the scalar
18877 instructions that target all 64 registers in the vector/scalar
18878 floating point register set, depending on the model of the machine.
18880 If the @option{-mno-upper-regs} option is used, it turns off both
18881 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
18883 @item -mfloat-gprs=@var{yes/single/double/no}
18884 @itemx -mfloat-gprs
18885 @opindex mfloat-gprs
18886 This switch enables or disables the generation of floating-point
18887 operations on the general-purpose registers for architectures that
18890 The argument @samp{yes} or @samp{single} enables the use of
18891 single-precision floating-point operations.
18893 The argument @samp{double} enables the use of single and
18894 double-precision floating-point operations.
18896 The argument @samp{no} disables floating-point operations on the
18897 general-purpose registers.
18899 This option is currently only available on the MPC854x.
18905 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
18906 targets (including GNU/Linux). The 32-bit environment sets int, long
18907 and pointer to 32 bits and generates code that runs on any PowerPC
18908 variant. The 64-bit environment sets int to 32 bits and long and
18909 pointer to 64 bits, and generates code for PowerPC64, as for
18910 @option{-mpowerpc64}.
18913 @itemx -mno-fp-in-toc
18914 @itemx -mno-sum-in-toc
18915 @itemx -mminimal-toc
18917 @opindex mno-fp-in-toc
18918 @opindex mno-sum-in-toc
18919 @opindex mminimal-toc
18920 Modify generation of the TOC (Table Of Contents), which is created for
18921 every executable file. The @option{-mfull-toc} option is selected by
18922 default. In that case, GCC allocates at least one TOC entry for
18923 each unique non-automatic variable reference in your program. GCC
18924 also places floating-point constants in the TOC@. However, only
18925 16,384 entries are available in the TOC@.
18927 If you receive a linker error message that saying you have overflowed
18928 the available TOC space, you can reduce the amount of TOC space used
18929 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
18930 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
18931 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
18932 generate code to calculate the sum of an address and a constant at
18933 run time instead of putting that sum into the TOC@. You may specify one
18934 or both of these options. Each causes GCC to produce very slightly
18935 slower and larger code at the expense of conserving TOC space.
18937 If you still run out of space in the TOC even when you specify both of
18938 these options, specify @option{-mminimal-toc} instead. This option causes
18939 GCC to make only one TOC entry for every file. When you specify this
18940 option, GCC produces code that is slower and larger but which
18941 uses extremely little TOC space. You may wish to use this option
18942 only on files that contain less frequently-executed code.
18948 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
18949 @code{long} type, and the infrastructure needed to support them.
18950 Specifying @option{-maix64} implies @option{-mpowerpc64},
18951 while @option{-maix32} disables the 64-bit ABI and
18952 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
18955 @itemx -mno-xl-compat
18956 @opindex mxl-compat
18957 @opindex mno-xl-compat
18958 Produce code that conforms more closely to IBM XL compiler semantics
18959 when using AIX-compatible ABI@. Pass floating-point arguments to
18960 prototyped functions beyond the register save area (RSA) on the stack
18961 in addition to argument FPRs. Do not assume that most significant
18962 double in 128-bit long double value is properly rounded when comparing
18963 values and converting to double. Use XL symbol names for long double
18966 The AIX calling convention was extended but not initially documented to
18967 handle an obscure K&R C case of calling a function that takes the
18968 address of its arguments with fewer arguments than declared. IBM XL
18969 compilers access floating-point arguments that do not fit in the
18970 RSA from the stack when a subroutine is compiled without
18971 optimization. Because always storing floating-point arguments on the
18972 stack is inefficient and rarely needed, this option is not enabled by
18973 default and only is necessary when calling subroutines compiled by IBM
18974 XL compilers without optimization.
18978 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
18979 application written to use message passing with special startup code to
18980 enable the application to run. The system must have PE installed in the
18981 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
18982 must be overridden with the @option{-specs=} option to specify the
18983 appropriate directory location. The Parallel Environment does not
18984 support threads, so the @option{-mpe} option and the @option{-pthread}
18985 option are incompatible.
18987 @item -malign-natural
18988 @itemx -malign-power
18989 @opindex malign-natural
18990 @opindex malign-power
18991 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
18992 @option{-malign-natural} overrides the ABI-defined alignment of larger
18993 types, such as floating-point doubles, on their natural size-based boundary.
18994 The option @option{-malign-power} instructs GCC to follow the ABI-specified
18995 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
18997 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19001 @itemx -mhard-float
19002 @opindex msoft-float
19003 @opindex mhard-float
19004 Generate code that does not use (uses) the floating-point register set.
19005 Software floating-point emulation is provided if you use the
19006 @option{-msoft-float} option, and pass the option to GCC when linking.
19008 @item -msingle-float
19009 @itemx -mdouble-float
19010 @opindex msingle-float
19011 @opindex mdouble-float
19012 Generate code for single- or double-precision floating-point operations.
19013 @option{-mdouble-float} implies @option{-msingle-float}.
19016 @opindex msimple-fpu
19017 Do not generate @code{sqrt} and @code{div} instructions for hardware
19018 floating-point unit.
19020 @item -mfpu=@var{name}
19022 Specify type of floating-point unit. Valid values for @var{name} are
19023 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19024 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19025 @samp{sp_full} (equivalent to @option{-msingle-float}),
19026 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19029 @opindex mxilinx-fpu
19030 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19033 @itemx -mno-multiple
19035 @opindex mno-multiple
19036 Generate code that uses (does not use) the load multiple word
19037 instructions and the store multiple word instructions. These
19038 instructions are generated by default on POWER systems, and not
19039 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19040 PowerPC systems, since those instructions do not work when the
19041 processor is in little-endian mode. The exceptions are PPC740 and
19042 PPC750 which permit these instructions in little-endian mode.
19047 @opindex mno-string
19048 Generate code that uses (does not use) the load string instructions
19049 and the store string word instructions to save multiple registers and
19050 do small block moves. These instructions are generated by default on
19051 POWER systems, and not generated on PowerPC systems. Do not use
19052 @option{-mstring} on little-endian PowerPC systems, since those
19053 instructions do not work when the processor is in little-endian mode.
19054 The exceptions are PPC740 and PPC750 which permit these instructions
19055 in little-endian mode.
19060 @opindex mno-update
19061 Generate code that uses (does not use) the load or store instructions
19062 that update the base register to the address of the calculated memory
19063 location. These instructions are generated by default. If you use
19064 @option{-mno-update}, there is a small window between the time that the
19065 stack pointer is updated and the address of the previous frame is
19066 stored, which means code that walks the stack frame across interrupts or
19067 signals may get corrupted data.
19069 @item -mavoid-indexed-addresses
19070 @itemx -mno-avoid-indexed-addresses
19071 @opindex mavoid-indexed-addresses
19072 @opindex mno-avoid-indexed-addresses
19073 Generate code that tries to avoid (not avoid) the use of indexed load
19074 or store instructions. These instructions can incur a performance
19075 penalty on Power6 processors in certain situations, such as when
19076 stepping through large arrays that cross a 16M boundary. This option
19077 is enabled by default when targeting Power6 and disabled otherwise.
19080 @itemx -mno-fused-madd
19081 @opindex mfused-madd
19082 @opindex mno-fused-madd
19083 Generate code that uses (does not use) the floating-point multiply and
19084 accumulate instructions. These instructions are generated by default
19085 if hardware floating point is used. The machine-dependent
19086 @option{-mfused-madd} option is now mapped to the machine-independent
19087 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19088 mapped to @option{-ffp-contract=off}.
19094 Generate code that uses (does not use) the half-word multiply and
19095 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19096 These instructions are generated by default when targeting those
19103 Generate code that uses (does not use) the string-search @samp{dlmzb}
19104 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19105 generated by default when targeting those processors.
19107 @item -mno-bit-align
19109 @opindex mno-bit-align
19110 @opindex mbit-align
19111 On System V.4 and embedded PowerPC systems do not (do) force structures
19112 and unions that contain bit-fields to be aligned to the base type of the
19115 For example, by default a structure containing nothing but 8
19116 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19117 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19118 the structure is aligned to a 1-byte boundary and is 1 byte in
19121 @item -mno-strict-align
19122 @itemx -mstrict-align
19123 @opindex mno-strict-align
19124 @opindex mstrict-align
19125 On System V.4 and embedded PowerPC systems do not (do) assume that
19126 unaligned memory references are handled by the system.
19128 @item -mrelocatable
19129 @itemx -mno-relocatable
19130 @opindex mrelocatable
19131 @opindex mno-relocatable
19132 Generate code that allows (does not allow) a static executable to be
19133 relocated to a different address at run time. A simple embedded
19134 PowerPC system loader should relocate the entire contents of
19135 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19136 a table of 32-bit addresses generated by this option. For this to
19137 work, all objects linked together must be compiled with
19138 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19139 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19141 @item -mrelocatable-lib
19142 @itemx -mno-relocatable-lib
19143 @opindex mrelocatable-lib
19144 @opindex mno-relocatable-lib
19145 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19146 @code{.fixup} section to allow static executables to be relocated at
19147 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19148 alignment of @option{-mrelocatable}. Objects compiled with
19149 @option{-mrelocatable-lib} may be linked with objects compiled with
19150 any combination of the @option{-mrelocatable} options.
19156 On System V.4 and embedded PowerPC systems do not (do) assume that
19157 register 2 contains a pointer to a global area pointing to the addresses
19158 used in the program.
19161 @itemx -mlittle-endian
19163 @opindex mlittle-endian
19164 On System V.4 and embedded PowerPC systems compile code for the
19165 processor in little-endian mode. The @option{-mlittle-endian} option is
19166 the same as @option{-mlittle}.
19169 @itemx -mbig-endian
19171 @opindex mbig-endian
19172 On System V.4 and embedded PowerPC systems compile code for the
19173 processor in big-endian mode. The @option{-mbig-endian} option is
19174 the same as @option{-mbig}.
19176 @item -mdynamic-no-pic
19177 @opindex mdynamic-no-pic
19178 On Darwin and Mac OS X systems, compile code so that it is not
19179 relocatable, but that its external references are relocatable. The
19180 resulting code is suitable for applications, but not shared
19183 @item -msingle-pic-base
19184 @opindex msingle-pic-base
19185 Treat the register used for PIC addressing as read-only, rather than
19186 loading it in the prologue for each function. The runtime system is
19187 responsible for initializing this register with an appropriate value
19188 before execution begins.
19190 @item -mprioritize-restricted-insns=@var{priority}
19191 @opindex mprioritize-restricted-insns
19192 This option controls the priority that is assigned to
19193 dispatch-slot restricted instructions during the second scheduling
19194 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19195 or @samp{2} to assign no, highest, or second-highest (respectively)
19196 priority to dispatch-slot restricted
19199 @item -msched-costly-dep=@var{dependence_type}
19200 @opindex msched-costly-dep
19201 This option controls which dependences are considered costly
19202 by the target during instruction scheduling. The argument
19203 @var{dependence_type} takes one of the following values:
19207 No dependence is costly.
19210 All dependences are costly.
19212 @item @samp{true_store_to_load}
19213 A true dependence from store to load is costly.
19215 @item @samp{store_to_load}
19216 Any dependence from store to load is costly.
19219 Any dependence for which the latency is greater than or equal to
19220 @var{number} is costly.
19223 @item -minsert-sched-nops=@var{scheme}
19224 @opindex minsert-sched-nops
19225 This option controls which NOP insertion scheme is used during
19226 the second scheduling pass. The argument @var{scheme} takes one of the
19234 Pad with NOPs any dispatch group that has vacant issue slots,
19235 according to the scheduler's grouping.
19237 @item @samp{regroup_exact}
19238 Insert NOPs to force costly dependent insns into
19239 separate groups. Insert exactly as many NOPs as needed to force an insn
19240 to a new group, according to the estimated processor grouping.
19243 Insert NOPs to force costly dependent insns into
19244 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19248 @opindex mcall-sysv
19249 On System V.4 and embedded PowerPC systems compile code using calling
19250 conventions that adhere to the March 1995 draft of the System V
19251 Application Binary Interface, PowerPC processor supplement. This is the
19252 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19254 @item -mcall-sysv-eabi
19256 @opindex mcall-sysv-eabi
19257 @opindex mcall-eabi
19258 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19260 @item -mcall-sysv-noeabi
19261 @opindex mcall-sysv-noeabi
19262 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19264 @item -mcall-aixdesc
19266 On System V.4 and embedded PowerPC systems compile code for the AIX
19270 @opindex mcall-linux
19271 On System V.4 and embedded PowerPC systems compile code for the
19272 Linux-based GNU system.
19274 @item -mcall-freebsd
19275 @opindex mcall-freebsd
19276 On System V.4 and embedded PowerPC systems compile code for the
19277 FreeBSD operating system.
19279 @item -mcall-netbsd
19280 @opindex mcall-netbsd
19281 On System V.4 and embedded PowerPC systems compile code for the
19282 NetBSD operating system.
19284 @item -mcall-openbsd
19285 @opindex mcall-netbsd
19286 On System V.4 and embedded PowerPC systems compile code for the
19287 OpenBSD operating system.
19289 @item -maix-struct-return
19290 @opindex maix-struct-return
19291 Return all structures in memory (as specified by the AIX ABI)@.
19293 @item -msvr4-struct-return
19294 @opindex msvr4-struct-return
19295 Return structures smaller than 8 bytes in registers (as specified by the
19298 @item -mabi=@var{abi-type}
19300 Extend the current ABI with a particular extension, or remove such extension.
19301 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
19302 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
19303 @samp{elfv1}, @samp{elfv2}@.
19307 Extend the current ABI with SPE ABI extensions. This does not change
19308 the default ABI, instead it adds the SPE ABI extensions to the current
19312 @opindex mabi=no-spe
19313 Disable Book-E SPE ABI extensions for the current ABI@.
19315 @item -mabi=ibmlongdouble
19316 @opindex mabi=ibmlongdouble
19317 Change the current ABI to use IBM extended-precision long double.
19318 This is a PowerPC 32-bit SYSV ABI option.
19320 @item -mabi=ieeelongdouble
19321 @opindex mabi=ieeelongdouble
19322 Change the current ABI to use IEEE extended-precision long double.
19323 This is a PowerPC 32-bit Linux ABI option.
19326 @opindex mabi=elfv1
19327 Change the current ABI to use the ELFv1 ABI.
19328 This is the default ABI for big-endian PowerPC 64-bit Linux.
19329 Overriding the default ABI requires special system support and is
19330 likely to fail in spectacular ways.
19333 @opindex mabi=elfv2
19334 Change the current ABI to use the ELFv2 ABI.
19335 This is the default ABI for little-endian PowerPC 64-bit Linux.
19336 Overriding the default ABI requires special system support and is
19337 likely to fail in spectacular ways.
19340 @itemx -mno-prototype
19341 @opindex mprototype
19342 @opindex mno-prototype
19343 On System V.4 and embedded PowerPC systems assume that all calls to
19344 variable argument functions are properly prototyped. Otherwise, the
19345 compiler must insert an instruction before every non-prototyped call to
19346 set or clear bit 6 of the condition code register (@code{CR}) to
19347 indicate whether floating-point values are passed in the floating-point
19348 registers in case the function takes variable arguments. With
19349 @option{-mprototype}, only calls to prototyped variable argument functions
19350 set or clear the bit.
19354 On embedded PowerPC systems, assume that the startup module is called
19355 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19356 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19361 On embedded PowerPC systems, assume that the startup module is called
19362 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19367 On embedded PowerPC systems, assume that the startup module is called
19368 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19371 @item -myellowknife
19372 @opindex myellowknife
19373 On embedded PowerPC systems, assume that the startup module is called
19374 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19379 On System V.4 and embedded PowerPC systems, specify that you are
19380 compiling for a VxWorks system.
19384 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
19385 header to indicate that @samp{eabi} extended relocations are used.
19391 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19392 Embedded Applications Binary Interface (EABI), which is a set of
19393 modifications to the System V.4 specifications. Selecting @option{-meabi}
19394 means that the stack is aligned to an 8-byte boundary, a function
19395 @code{__eabi} is called from @code{main} to set up the EABI
19396 environment, and the @option{-msdata} option can use both @code{r2} and
19397 @code{r13} to point to two separate small data areas. Selecting
19398 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19399 no EABI initialization function is called from @code{main}, and the
19400 @option{-msdata} option only uses @code{r13} to point to a single
19401 small data area. The @option{-meabi} option is on by default if you
19402 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19405 @opindex msdata=eabi
19406 On System V.4 and embedded PowerPC systems, put small initialized
19407 @code{const} global and static data in the @code{.sdata2} section, which
19408 is pointed to by register @code{r2}. Put small initialized
19409 non-@code{const} global and static data in the @code{.sdata} section,
19410 which is pointed to by register @code{r13}. Put small uninitialized
19411 global and static data in the @code{.sbss} section, which is adjacent to
19412 the @code{.sdata} section. The @option{-msdata=eabi} option is
19413 incompatible with the @option{-mrelocatable} option. The
19414 @option{-msdata=eabi} option also sets the @option{-memb} option.
19417 @opindex msdata=sysv
19418 On System V.4 and embedded PowerPC systems, put small global and static
19419 data in the @code{.sdata} section, which is pointed to by register
19420 @code{r13}. Put small uninitialized global and static data in the
19421 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
19422 The @option{-msdata=sysv} option is incompatible with the
19423 @option{-mrelocatable} option.
19425 @item -msdata=default
19427 @opindex msdata=default
19429 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19430 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19431 same as @option{-msdata=sysv}.
19434 @opindex msdata=data
19435 On System V.4 and embedded PowerPC systems, put small global
19436 data in the @code{.sdata} section. Put small uninitialized global
19437 data in the @code{.sbss} section. Do not use register @code{r13}
19438 to address small data however. This is the default behavior unless
19439 other @option{-msdata} options are used.
19443 @opindex msdata=none
19445 On embedded PowerPC systems, put all initialized global and static data
19446 in the @code{.data} section, and all uninitialized data in the
19447 @code{.bss} section.
19449 @item -mblock-move-inline-limit=@var{num}
19450 @opindex mblock-move-inline-limit
19451 Inline all block moves (such as calls to @code{memcpy} or structure
19452 copies) less than or equal to @var{num} bytes. The minimum value for
19453 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19454 targets. The default value is target-specific.
19458 @cindex smaller data references (PowerPC)
19459 @cindex .sdata/.sdata2 references (PowerPC)
19460 On embedded PowerPC systems, put global and static items less than or
19461 equal to @var{num} bytes into the small data or BSS sections instead of
19462 the normal data or BSS section. By default, @var{num} is 8. The
19463 @option{-G @var{num}} switch is also passed to the linker.
19464 All modules should be compiled with the same @option{-G @var{num}} value.
19467 @itemx -mno-regnames
19469 @opindex mno-regnames
19470 On System V.4 and embedded PowerPC systems do (do not) emit register
19471 names in the assembly language output using symbolic forms.
19474 @itemx -mno-longcall
19476 @opindex mno-longcall
19477 By default assume that all calls are far away so that a longer and more
19478 expensive calling sequence is required. This is required for calls
19479 farther than 32 megabytes (33,554,432 bytes) from the current location.
19480 A short call is generated if the compiler knows
19481 the call cannot be that far away. This setting can be overridden by
19482 the @code{shortcall} function attribute, or by @code{#pragma
19485 Some linkers are capable of detecting out-of-range calls and generating
19486 glue code on the fly. On these systems, long calls are unnecessary and
19487 generate slower code. As of this writing, the AIX linker can do this,
19488 as can the GNU linker for PowerPC/64. It is planned to add this feature
19489 to the GNU linker for 32-bit PowerPC systems as well.
19491 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
19492 callee, L42}, plus a @dfn{branch island} (glue code). The two target
19493 addresses represent the callee and the branch island. The
19494 Darwin/PPC linker prefers the first address and generates a @code{bl
19495 callee} if the PPC @code{bl} instruction reaches the callee directly;
19496 otherwise, the linker generates @code{bl L42} to call the branch
19497 island. The branch island is appended to the body of the
19498 calling function; it computes the full 32-bit address of the callee
19501 On Mach-O (Darwin) systems, this option directs the compiler emit to
19502 the glue for every direct call, and the Darwin linker decides whether
19503 to use or discard it.
19505 In the future, GCC may ignore all longcall specifications
19506 when the linker is known to generate glue.
19508 @item -mtls-markers
19509 @itemx -mno-tls-markers
19510 @opindex mtls-markers
19511 @opindex mno-tls-markers
19512 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
19513 specifying the function argument. The relocation allows the linker to
19514 reliably associate function call with argument setup instructions for
19515 TLS optimization, which in turn allows GCC to better schedule the
19520 Adds support for multithreading with the @dfn{pthreads} library.
19521 This option sets flags for both the preprocessor and linker.
19526 This option enables use of the reciprocal estimate and
19527 reciprocal square root estimate instructions with additional
19528 Newton-Raphson steps to increase precision instead of doing a divide or
19529 square root and divide for floating-point arguments. You should use
19530 the @option{-ffast-math} option when using @option{-mrecip} (or at
19531 least @option{-funsafe-math-optimizations},
19532 @option{-finite-math-only}, @option{-freciprocal-math} and
19533 @option{-fno-trapping-math}). Note that while the throughput of the
19534 sequence is generally higher than the throughput of the non-reciprocal
19535 instruction, the precision of the sequence can be decreased by up to 2
19536 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
19539 @item -mrecip=@var{opt}
19540 @opindex mrecip=opt
19541 This option controls which reciprocal estimate instructions
19542 may be used. @var{opt} is a comma-separated list of options, which may
19543 be preceded by a @code{!} to invert the option:
19548 Enable all estimate instructions.
19551 Enable the default instructions, equivalent to @option{-mrecip}.
19554 Disable all estimate instructions, equivalent to @option{-mno-recip}.
19557 Enable the reciprocal approximation instructions for both
19558 single and double precision.
19561 Enable the single-precision reciprocal approximation instructions.
19564 Enable the double-precision reciprocal approximation instructions.
19567 Enable the reciprocal square root approximation instructions for both
19568 single and double precision.
19571 Enable the single-precision reciprocal square root approximation instructions.
19574 Enable the double-precision reciprocal square root approximation instructions.
19578 So, for example, @option{-mrecip=all,!rsqrtd} enables
19579 all of the reciprocal estimate instructions, except for the
19580 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
19581 which handle the double-precision reciprocal square root calculations.
19583 @item -mrecip-precision
19584 @itemx -mno-recip-precision
19585 @opindex mrecip-precision
19586 Assume (do not assume) that the reciprocal estimate instructions
19587 provide higher-precision estimates than is mandated by the PowerPC
19588 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
19589 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
19590 The double-precision square root estimate instructions are not generated by
19591 default on low-precision machines, since they do not provide an
19592 estimate that converges after three steps.
19594 @item -mveclibabi=@var{type}
19595 @opindex mveclibabi
19596 Specifies the ABI type to use for vectorizing intrinsics using an
19597 external library. The only type supported at present is @samp{mass},
19598 which specifies to use IBM's Mathematical Acceleration Subsystem
19599 (MASS) libraries for vectorizing intrinsics using external libraries.
19600 GCC currently emits calls to @code{acosd2}, @code{acosf4},
19601 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
19602 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
19603 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
19604 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
19605 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
19606 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
19607 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
19608 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
19609 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
19610 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
19611 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
19612 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
19613 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
19614 for power7. Both @option{-ftree-vectorize} and
19615 @option{-funsafe-math-optimizations} must also be enabled. The MASS
19616 libraries must be specified at link time.
19621 Generate (do not generate) the @code{friz} instruction when the
19622 @option{-funsafe-math-optimizations} option is used to optimize
19623 rounding of floating-point values to 64-bit integer and back to floating
19624 point. The @code{friz} instruction does not return the same value if
19625 the floating-point number is too large to fit in an integer.
19627 @item -mpointers-to-nested-functions
19628 @itemx -mno-pointers-to-nested-functions
19629 @opindex mpointers-to-nested-functions
19630 Generate (do not generate) code to load up the static chain register
19631 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
19632 systems where a function pointer points to a 3-word descriptor giving
19633 the function address, TOC value to be loaded in register @code{r2}, and
19634 static chain value to be loaded in register @code{r11}. The
19635 @option{-mpointers-to-nested-functions} is on by default. You cannot
19636 call through pointers to nested functions or pointers
19637 to functions compiled in other languages that use the static chain if
19638 you use the @option{-mno-pointers-to-nested-functions}.
19640 @item -msave-toc-indirect
19641 @itemx -mno-save-toc-indirect
19642 @opindex msave-toc-indirect
19643 Generate (do not generate) code to save the TOC value in the reserved
19644 stack location in the function prologue if the function calls through
19645 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
19646 saved in the prologue, it is saved just before the call through the
19647 pointer. The @option{-mno-save-toc-indirect} option is the default.
19649 @item -mcompat-align-parm
19650 @itemx -mno-compat-align-parm
19651 @opindex mcompat-align-parm
19652 Generate (do not generate) code to pass structure parameters with a
19653 maximum alignment of 64 bits, for compatibility with older versions
19656 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
19657 structure parameter on a 128-bit boundary when that structure contained
19658 a member requiring 128-bit alignment. This is corrected in more
19659 recent versions of GCC. This option may be used to generate code
19660 that is compatible with functions compiled with older versions of
19663 The @option{-mno-compat-align-parm} option is the default.
19667 @subsection RX Options
19670 These command-line options are defined for RX targets:
19673 @item -m64bit-doubles
19674 @itemx -m32bit-doubles
19675 @opindex m64bit-doubles
19676 @opindex m32bit-doubles
19677 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19678 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19679 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
19680 works on 32-bit values, which is why the default is
19681 @option{-m32bit-doubles}.
19687 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
19688 floating-point hardware. The default is enabled for the RX600
19689 series and disabled for the RX200 series.
19691 Floating-point instructions are only generated for 32-bit floating-point
19692 values, however, so the FPU hardware is not used for doubles if the
19693 @option{-m64bit-doubles} option is used.
19695 @emph{Note} If the @option{-fpu} option is enabled then
19696 @option{-funsafe-math-optimizations} is also enabled automatically.
19697 This is because the RX FPU instructions are themselves unsafe.
19699 @item -mcpu=@var{name}
19701 Selects the type of RX CPU to be targeted. Currently three types are
19702 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
19703 the specific @samp{RX610} CPU. The default is @samp{RX600}.
19705 The only difference between @samp{RX600} and @samp{RX610} is that the
19706 @samp{RX610} does not support the @code{MVTIPL} instruction.
19708 The @samp{RX200} series does not have a hardware floating-point unit
19709 and so @option{-nofpu} is enabled by default when this type is
19712 @item -mbig-endian-data
19713 @itemx -mlittle-endian-data
19714 @opindex mbig-endian-data
19715 @opindex mlittle-endian-data
19716 Store data (but not code) in the big-endian format. The default is
19717 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
19720 @item -msmall-data-limit=@var{N}
19721 @opindex msmall-data-limit
19722 Specifies the maximum size in bytes of global and static variables
19723 which can be placed into the small data area. Using the small data
19724 area can lead to smaller and faster code, but the size of area is
19725 limited and it is up to the programmer to ensure that the area does
19726 not overflow. Also when the small data area is used one of the RX's
19727 registers (usually @code{r13}) is reserved for use pointing to this
19728 area, so it is no longer available for use by the compiler. This
19729 could result in slower and/or larger code if variables are pushed onto
19730 the stack instead of being held in this register.
19732 Note, common variables (variables that have not been initialized) and
19733 constants are not placed into the small data area as they are assigned
19734 to other sections in the output executable.
19736 The default value is zero, which disables this feature. Note, this
19737 feature is not enabled by default with higher optimization levels
19738 (@option{-O2} etc) because of the potentially detrimental effects of
19739 reserving a register. It is up to the programmer to experiment and
19740 discover whether this feature is of benefit to their program. See the
19741 description of the @option{-mpid} option for a description of how the
19742 actual register to hold the small data area pointer is chosen.
19748 Use the simulator runtime. The default is to use the libgloss
19749 board-specific runtime.
19751 @item -mas100-syntax
19752 @itemx -mno-as100-syntax
19753 @opindex mas100-syntax
19754 @opindex mno-as100-syntax
19755 When generating assembler output use a syntax that is compatible with
19756 Renesas's AS100 assembler. This syntax can also be handled by the GAS
19757 assembler, but it has some restrictions so it is not generated by default.
19759 @item -mmax-constant-size=@var{N}
19760 @opindex mmax-constant-size
19761 Specifies the maximum size, in bytes, of a constant that can be used as
19762 an operand in a RX instruction. Although the RX instruction set does
19763 allow constants of up to 4 bytes in length to be used in instructions,
19764 a longer value equates to a longer instruction. Thus in some
19765 circumstances it can be beneficial to restrict the size of constants
19766 that are used in instructions. Constants that are too big are instead
19767 placed into a constant pool and referenced via register indirection.
19769 The value @var{N} can be between 0 and 4. A value of 0 (the default)
19770 or 4 means that constants of any size are allowed.
19774 Enable linker relaxation. Linker relaxation is a process whereby the
19775 linker attempts to reduce the size of a program by finding shorter
19776 versions of various instructions. Disabled by default.
19778 @item -mint-register=@var{N}
19779 @opindex mint-register
19780 Specify the number of registers to reserve for fast interrupt handler
19781 functions. The value @var{N} can be between 0 and 4. A value of 1
19782 means that register @code{r13} is reserved for the exclusive use
19783 of fast interrupt handlers. A value of 2 reserves @code{r13} and
19784 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
19785 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
19786 A value of 0, the default, does not reserve any registers.
19788 @item -msave-acc-in-interrupts
19789 @opindex msave-acc-in-interrupts
19790 Specifies that interrupt handler functions should preserve the
19791 accumulator register. This is only necessary if normal code might use
19792 the accumulator register, for example because it performs 64-bit
19793 multiplications. The default is to ignore the accumulator as this
19794 makes the interrupt handlers faster.
19800 Enables the generation of position independent data. When enabled any
19801 access to constant data is done via an offset from a base address
19802 held in a register. This allows the location of constant data to be
19803 determined at run time without requiring the executable to be
19804 relocated, which is a benefit to embedded applications with tight
19805 memory constraints. Data that can be modified is not affected by this
19808 Note, using this feature reserves a register, usually @code{r13}, for
19809 the constant data base address. This can result in slower and/or
19810 larger code, especially in complicated functions.
19812 The actual register chosen to hold the constant data base address
19813 depends upon whether the @option{-msmall-data-limit} and/or the
19814 @option{-mint-register} command-line options are enabled. Starting
19815 with register @code{r13} and proceeding downwards, registers are
19816 allocated first to satisfy the requirements of @option{-mint-register},
19817 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
19818 is possible for the small data area register to be @code{r8} if both
19819 @option{-mint-register=4} and @option{-mpid} are specified on the
19822 By default this feature is not enabled. The default can be restored
19823 via the @option{-mno-pid} command-line option.
19825 @item -mno-warn-multiple-fast-interrupts
19826 @itemx -mwarn-multiple-fast-interrupts
19827 @opindex mno-warn-multiple-fast-interrupts
19828 @opindex mwarn-multiple-fast-interrupts
19829 Prevents GCC from issuing a warning message if it finds more than one
19830 fast interrupt handler when it is compiling a file. The default is to
19831 issue a warning for each extra fast interrupt handler found, as the RX
19832 only supports one such interrupt.
19836 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
19837 has special significance to the RX port when used with the
19838 @code{interrupt} function attribute. This attribute indicates a
19839 function intended to process fast interrupts. GCC ensures
19840 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
19841 and/or @code{r13} and only provided that the normal use of the
19842 corresponding registers have been restricted via the
19843 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
19846 @node S/390 and zSeries Options
19847 @subsection S/390 and zSeries Options
19848 @cindex S/390 and zSeries Options
19850 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
19854 @itemx -msoft-float
19855 @opindex mhard-float
19856 @opindex msoft-float
19857 Use (do not use) the hardware floating-point instructions and registers
19858 for floating-point operations. When @option{-msoft-float} is specified,
19859 functions in @file{libgcc.a} are used to perform floating-point
19860 operations. When @option{-mhard-float} is specified, the compiler
19861 generates IEEE floating-point instructions. This is the default.
19864 @itemx -mno-hard-dfp
19866 @opindex mno-hard-dfp
19867 Use (do not use) the hardware decimal-floating-point instructions for
19868 decimal-floating-point operations. When @option{-mno-hard-dfp} is
19869 specified, functions in @file{libgcc.a} are used to perform
19870 decimal-floating-point operations. When @option{-mhard-dfp} is
19871 specified, the compiler generates decimal-floating-point hardware
19872 instructions. This is the default for @option{-march=z9-ec} or higher.
19874 @item -mlong-double-64
19875 @itemx -mlong-double-128
19876 @opindex mlong-double-64
19877 @opindex mlong-double-128
19878 These switches control the size of @code{long double} type. A size
19879 of 64 bits makes the @code{long double} type equivalent to the @code{double}
19880 type. This is the default.
19883 @itemx -mno-backchain
19884 @opindex mbackchain
19885 @opindex mno-backchain
19886 Store (do not store) the address of the caller's frame as backchain pointer
19887 into the callee's stack frame.
19888 A backchain may be needed to allow debugging using tools that do not understand
19889 DWARF 2 call frame information.
19890 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
19891 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
19892 the backchain is placed into the topmost word of the 96/160 byte register
19895 In general, code compiled with @option{-mbackchain} is call-compatible with
19896 code compiled with @option{-mmo-backchain}; however, use of the backchain
19897 for debugging purposes usually requires that the whole binary is built with
19898 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
19899 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
19900 to build a linux kernel use @option{-msoft-float}.
19902 The default is to not maintain the backchain.
19904 @item -mpacked-stack
19905 @itemx -mno-packed-stack
19906 @opindex mpacked-stack
19907 @opindex mno-packed-stack
19908 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
19909 specified, the compiler uses the all fields of the 96/160 byte register save
19910 area only for their default purpose; unused fields still take up stack space.
19911 When @option{-mpacked-stack} is specified, register save slots are densely
19912 packed at the top of the register save area; unused space is reused for other
19913 purposes, allowing for more efficient use of the available stack space.
19914 However, when @option{-mbackchain} is also in effect, the topmost word of
19915 the save area is always used to store the backchain, and the return address
19916 register is always saved two words below the backchain.
19918 As long as the stack frame backchain is not used, code generated with
19919 @option{-mpacked-stack} is call-compatible with code generated with
19920 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
19921 S/390 or zSeries generated code that uses the stack frame backchain at run
19922 time, not just for debugging purposes. Such code is not call-compatible
19923 with code compiled with @option{-mpacked-stack}. Also, note that the
19924 combination of @option{-mbackchain},
19925 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
19926 to build a linux kernel use @option{-msoft-float}.
19928 The default is to not use the packed stack layout.
19931 @itemx -mno-small-exec
19932 @opindex msmall-exec
19933 @opindex mno-small-exec
19934 Generate (or do not generate) code using the @code{bras} instruction
19935 to do subroutine calls.
19936 This only works reliably if the total executable size does not
19937 exceed 64k. The default is to use the @code{basr} instruction instead,
19938 which does not have this limitation.
19944 When @option{-m31} is specified, generate code compliant to the
19945 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
19946 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
19947 particular to generate 64-bit instructions. For the @samp{s390}
19948 targets, the default is @option{-m31}, while the @samp{s390x}
19949 targets default to @option{-m64}.
19955 When @option{-mzarch} is specified, generate code using the
19956 instructions available on z/Architecture.
19957 When @option{-mesa} is specified, generate code using the
19958 instructions available on ESA/390. Note that @option{-mesa} is
19959 not possible with @option{-m64}.
19960 When generating code compliant to the GNU/Linux for S/390 ABI,
19961 the default is @option{-mesa}. When generating code compliant
19962 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
19968 Generate (or do not generate) code using the @code{mvcle} instruction
19969 to perform block moves. When @option{-mno-mvcle} is specified,
19970 use a @code{mvc} loop instead. This is the default unless optimizing for
19977 Print (or do not print) additional debug information when compiling.
19978 The default is to not print debug information.
19980 @item -march=@var{cpu-type}
19982 Generate code that runs on @var{cpu-type}, which is the name of a system
19983 representing a certain processor type. Possible values for
19984 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
19985 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
19986 When generating code using the instructions available on z/Architecture,
19987 the default is @option{-march=z900}. Otherwise, the default is
19988 @option{-march=g5}.
19990 @item -mtune=@var{cpu-type}
19992 Tune to @var{cpu-type} everything applicable about the generated code,
19993 except for the ABI and the set of available instructions.
19994 The list of @var{cpu-type} values is the same as for @option{-march}.
19995 The default is the value used for @option{-march}.
19998 @itemx -mno-tpf-trace
19999 @opindex mtpf-trace
20000 @opindex mno-tpf-trace
20001 Generate code that adds (does not add) in TPF OS specific branches to trace
20002 routines in the operating system. This option is off by default, even
20003 when compiling for the TPF OS@.
20006 @itemx -mno-fused-madd
20007 @opindex mfused-madd
20008 @opindex mno-fused-madd
20009 Generate code that uses (does not use) the floating-point multiply and
20010 accumulate instructions. These instructions are generated by default if
20011 hardware floating point is used.
20013 @item -mwarn-framesize=@var{framesize}
20014 @opindex mwarn-framesize
20015 Emit a warning if the current function exceeds the given frame size. Because
20016 this is a compile-time check it doesn't need to be a real problem when the program
20017 runs. It is intended to identify functions that most probably cause
20018 a stack overflow. It is useful to be used in an environment with limited stack
20019 size e.g.@: the linux kernel.
20021 @item -mwarn-dynamicstack
20022 @opindex mwarn-dynamicstack
20023 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20024 arrays. This is generally a bad idea with a limited stack size.
20026 @item -mstack-guard=@var{stack-guard}
20027 @itemx -mstack-size=@var{stack-size}
20028 @opindex mstack-guard
20029 @opindex mstack-size
20030 If these options are provided the S/390 back end emits additional instructions in
20031 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20032 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20033 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20034 the frame size of the compiled function is chosen.
20035 These options are intended to be used to help debugging stack overflow problems.
20036 The additionally emitted code causes only little overhead and hence can also be
20037 used in production-like systems without greater performance degradation. The given
20038 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20039 @var{stack-guard} without exceeding 64k.
20040 In order to be efficient the extra code makes the assumption that the stack starts
20041 at an address aligned to the value given by @var{stack-size}.
20042 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20044 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20046 If the hotpatch option is enabled, a ``hot-patching'' function
20047 prologue is generated for all functions in the compilation unit.
20048 The funtion label is prepended with the given number of two-byte
20049 Nop instructions (@var{pre-halfwords}, maximum 1000000). After
20050 the label, 2 * @var{post-halfwords} bytes are appended, using the
20051 larges nop like instructions the architecture allows (maximum
20054 If both arguments are zero, hotpatching is disabled.
20056 This option can be overridden for individual functions with the
20057 @code{hotpatch} attribute.
20060 @node Score Options
20061 @subsection Score Options
20062 @cindex Score Options
20064 These options are defined for Score implementations:
20069 Compile code for big-endian mode. This is the default.
20073 Compile code for little-endian mode.
20077 Disable generation of @code{bcnz} instructions.
20081 Enable generation of unaligned load and store instructions.
20085 Enable the use of multiply-accumulate instructions. Disabled by default.
20089 Specify the SCORE5 as the target architecture.
20093 Specify the SCORE5U of the target architecture.
20097 Specify the SCORE7 as the target architecture. This is the default.
20101 Specify the SCORE7D as the target architecture.
20105 @subsection SH Options
20107 These @samp{-m} options are defined for the SH implementations:
20112 Generate code for the SH1.
20116 Generate code for the SH2.
20119 Generate code for the SH2e.
20123 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20124 that the floating-point unit is not used.
20126 @item -m2a-single-only
20127 @opindex m2a-single-only
20128 Generate code for the SH2a-FPU, in such a way that no double-precision
20129 floating-point operations are used.
20132 @opindex m2a-single
20133 Generate code for the SH2a-FPU assuming the floating-point unit is in
20134 single-precision mode by default.
20138 Generate code for the SH2a-FPU assuming the floating-point unit is in
20139 double-precision mode by default.
20143 Generate code for the SH3.
20147 Generate code for the SH3e.
20151 Generate code for the SH4 without a floating-point unit.
20153 @item -m4-single-only
20154 @opindex m4-single-only
20155 Generate code for the SH4 with a floating-point unit that only
20156 supports single-precision arithmetic.
20160 Generate code for the SH4 assuming the floating-point unit is in
20161 single-precision mode by default.
20165 Generate code for the SH4.
20169 Generate code for SH4-100.
20171 @item -m4-100-nofpu
20172 @opindex m4-100-nofpu
20173 Generate code for SH4-100 in such a way that the
20174 floating-point unit is not used.
20176 @item -m4-100-single
20177 @opindex m4-100-single
20178 Generate code for SH4-100 assuming the floating-point unit is in
20179 single-precision mode by default.
20181 @item -m4-100-single-only
20182 @opindex m4-100-single-only
20183 Generate code for SH4-100 in such a way that no double-precision
20184 floating-point operations are used.
20188 Generate code for SH4-200.
20190 @item -m4-200-nofpu
20191 @opindex m4-200-nofpu
20192 Generate code for SH4-200 without in such a way that the
20193 floating-point unit is not used.
20195 @item -m4-200-single
20196 @opindex m4-200-single
20197 Generate code for SH4-200 assuming the floating-point unit is in
20198 single-precision mode by default.
20200 @item -m4-200-single-only
20201 @opindex m4-200-single-only
20202 Generate code for SH4-200 in such a way that no double-precision
20203 floating-point operations are used.
20207 Generate code for SH4-300.
20209 @item -m4-300-nofpu
20210 @opindex m4-300-nofpu
20211 Generate code for SH4-300 without in such a way that the
20212 floating-point unit is not used.
20214 @item -m4-300-single
20215 @opindex m4-300-single
20216 Generate code for SH4-300 in such a way that no double-precision
20217 floating-point operations are used.
20219 @item -m4-300-single-only
20220 @opindex m4-300-single-only
20221 Generate code for SH4-300 in such a way that no double-precision
20222 floating-point operations are used.
20226 Generate code for SH4-340 (no MMU, no FPU).
20230 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
20235 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20236 floating-point unit is not used.
20238 @item -m4a-single-only
20239 @opindex m4a-single-only
20240 Generate code for the SH4a, in such a way that no double-precision
20241 floating-point operations are used.
20244 @opindex m4a-single
20245 Generate code for the SH4a assuming the floating-point unit is in
20246 single-precision mode by default.
20250 Generate code for the SH4a.
20254 Same as @option{-m4a-nofpu}, except that it implicitly passes
20255 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20256 instructions at the moment.
20259 @opindex m5-32media
20260 Generate 32-bit code for SHmedia.
20262 @item -m5-32media-nofpu
20263 @opindex m5-32media-nofpu
20264 Generate 32-bit code for SHmedia in such a way that the
20265 floating-point unit is not used.
20268 @opindex m5-64media
20269 Generate 64-bit code for SHmedia.
20271 @item -m5-64media-nofpu
20272 @opindex m5-64media-nofpu
20273 Generate 64-bit code for SHmedia in such a way that the
20274 floating-point unit is not used.
20277 @opindex m5-compact
20278 Generate code for SHcompact.
20280 @item -m5-compact-nofpu
20281 @opindex m5-compact-nofpu
20282 Generate code for SHcompact in such a way that the
20283 floating-point unit is not used.
20287 Compile code for the processor in big-endian mode.
20291 Compile code for the processor in little-endian mode.
20295 Align doubles at 64-bit boundaries. Note that this changes the calling
20296 conventions, and thus some functions from the standard C library do
20297 not work unless you recompile it first with @option{-mdalign}.
20301 Shorten some address references at link time, when possible; uses the
20302 linker option @option{-relax}.
20306 Use 32-bit offsets in @code{switch} tables. The default is to use
20311 Enable the use of bit manipulation instructions on SH2A.
20315 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20316 alignment constraints.
20320 Comply with the calling conventions defined by Renesas.
20323 @opindex mno-renesas
20324 Comply with the calling conventions defined for GCC before the Renesas
20325 conventions were available. This option is the default for all
20326 targets of the SH toolchain.
20329 @opindex mnomacsave
20330 Mark the @code{MAC} register as call-clobbered, even if
20331 @option{-mrenesas} is given.
20337 Control the IEEE compliance of floating-point comparisons, which affects the
20338 handling of cases where the result of a comparison is unordered. By default
20339 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20340 enabled @option{-mno-ieee} is implicitly set, which results in faster
20341 floating-point greater-equal and less-equal comparisons. The implcit settings
20342 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20344 @item -minline-ic_invalidate
20345 @opindex minline-ic_invalidate
20346 Inline code to invalidate instruction cache entries after setting up
20347 nested function trampolines.
20348 This option has no effect if @option{-musermode} is in effect and the selected
20349 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20351 If the selected code generation option does not allow the use of the @code{icbi}
20352 instruction, and @option{-musermode} is not in effect, the inlined code
20353 manipulates the instruction cache address array directly with an associative
20354 write. This not only requires privileged mode at run time, but it also
20355 fails if the cache line had been mapped via the TLB and has become unmapped.
20359 Dump instruction size and location in the assembly code.
20362 @opindex mpadstruct
20363 This option is deprecated. It pads structures to multiple of 4 bytes,
20364 which is incompatible with the SH ABI@.
20366 @item -matomic-model=@var{model}
20367 @opindex matomic-model=@var{model}
20368 Sets the model of atomic operations and additional parameters as a comma
20369 separated list. For details on the atomic built-in functions see
20370 @ref{__atomic Builtins}. The following models and parameters are supported:
20375 Disable compiler generated atomic sequences and emit library calls for atomic
20376 operations. This is the default if the target is not @code{sh*-*-linux*}.
20379 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20380 built-in functions. The generated atomic sequences require additional support
20381 from the interrupt/exception handling code of the system and are only suitable
20382 for SH3* and SH4* single-core systems. This option is enabled by default when
20383 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
20384 this option also partially utilizes the hardware atomic instructions
20385 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20386 @samp{strict} is specified.
20389 Generate software atomic sequences that use a variable in the thread control
20390 block. This is a variation of the gUSA sequences which can also be used on
20391 SH1* and SH2* targets. The generated atomic sequences require additional
20392 support from the interrupt/exception handling code of the system and are only
20393 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20394 parameter has to be specified as well.
20397 Generate software atomic sequences that temporarily disable interrupts by
20398 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20399 in privileged mode and is only suitable for single-core systems. Additional
20400 support from the interrupt/exception handling code of the system is not
20401 required. This model is enabled by default when the target is
20402 @code{sh*-*-linux*} and SH1* or SH2*.
20405 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20406 instructions only. This is only available on SH4A and is suitable for
20407 multi-core systems. Since the hardware instructions support only 32 bit atomic
20408 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20409 Code compiled with this option is also compatible with other software
20410 atomic model interrupt/exception handling systems if executed on an SH4A
20411 system. Additional support from the interrupt/exception handling code of the
20412 system is not required for this model.
20415 This parameter specifies the offset in bytes of the variable in the thread
20416 control block structure that should be used by the generated atomic sequences
20417 when the @samp{soft-tcb} model has been selected. For other models this
20418 parameter is ignored. The specified value must be an integer multiple of four
20419 and in the range 0-1020.
20422 This parameter prevents mixed usage of multiple atomic models, even if they
20423 are compatible, and makes the compiler generate atomic sequences of the
20424 specified model only.
20430 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20431 Notice that depending on the particular hardware and software configuration
20432 this can degrade overall performance due to the operand cache line flushes
20433 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20434 processors the @code{tas.b} instruction must be used with caution since it
20435 can result in data corruption for certain cache configurations.
20438 @opindex mprefergot
20439 When generating position-independent code, emit function calls using
20440 the Global Offset Table instead of the Procedure Linkage Table.
20443 @itemx -mno-usermode
20445 @opindex mno-usermode
20446 Don't allow (allow) the compiler generating privileged mode code. Specifying
20447 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
20448 inlined code would not work in user mode. @option{-musermode} is the default
20449 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
20450 @option{-musermode} has no effect, since there is no user mode.
20452 @item -multcost=@var{number}
20453 @opindex multcost=@var{number}
20454 Set the cost to assume for a multiply insn.
20456 @item -mdiv=@var{strategy}
20457 @opindex mdiv=@var{strategy}
20458 Set the division strategy to be used for integer division operations.
20459 For SHmedia @var{strategy} can be one of:
20464 Performs the operation in floating point. This has a very high latency,
20465 but needs only a few instructions, so it might be a good choice if
20466 your code has enough easily-exploitable ILP to allow the compiler to
20467 schedule the floating-point instructions together with other instructions.
20468 Division by zero causes a floating-point exception.
20471 Uses integer operations to calculate the inverse of the divisor,
20472 and then multiplies the dividend with the inverse. This strategy allows
20473 CSE and hoisting of the inverse calculation. Division by zero calculates
20474 an unspecified result, but does not trap.
20477 A variant of @samp{inv} where, if no CSE or hoisting opportunities
20478 have been found, or if the entire operation has been hoisted to the same
20479 place, the last stages of the inverse calculation are intertwined with the
20480 final multiply to reduce the overall latency, at the expense of using a few
20481 more instructions, and thus offering fewer scheduling opportunities with
20485 Calls a library function that usually implements the @samp{inv:minlat}
20487 This gives high code density for @code{m5-*media-nofpu} compilations.
20490 Uses a different entry point of the same library function, where it
20491 assumes that a pointer to a lookup table has already been set up, which
20492 exposes the pointer load to CSE and code hoisting optimizations.
20497 Use the @samp{inv} algorithm for initial
20498 code generation, but if the code stays unoptimized, revert to the @samp{call},
20499 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
20500 potentially-trapping side effect of division by zero is carried by a
20501 separate instruction, so it is possible that all the integer instructions
20502 are hoisted out, but the marker for the side effect stays where it is.
20503 A recombination to floating-point operations or a call is not possible
20508 Variants of the @samp{inv:minlat} strategy. In the case
20509 that the inverse calculation is not separated from the multiply, they speed
20510 up division where the dividend fits into 20 bits (plus sign where applicable)
20511 by inserting a test to skip a number of operations in this case; this test
20512 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
20513 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
20517 For targets other than SHmedia @var{strategy} can be one of:
20522 Calls a library function that uses the single-step division instruction
20523 @code{div1} to perform the operation. Division by zero calculates an
20524 unspecified result and does not trap. This is the default except for SH4,
20525 SH2A and SHcompact.
20528 Calls a library function that performs the operation in double precision
20529 floating point. Division by zero causes a floating-point exception. This is
20530 the default for SHcompact with FPU. Specifying this for targets that do not
20531 have a double precision FPU defaults to @code{call-div1}.
20534 Calls a library function that uses a lookup table for small divisors and
20535 the @code{div1} instruction with case distinction for larger divisors. Division
20536 by zero calculates an unspecified result and does not trap. This is the default
20537 for SH4. Specifying this for targets that do not have dynamic shift
20538 instructions defaults to @code{call-div1}.
20542 When a division strategy has not been specified the default strategy is
20543 selected based on the current target. For SH2A the default strategy is to
20544 use the @code{divs} and @code{divu} instructions instead of library function
20547 @item -maccumulate-outgoing-args
20548 @opindex maccumulate-outgoing-args
20549 Reserve space once for outgoing arguments in the function prologue rather
20550 than around each call. Generally beneficial for performance and size. Also
20551 needed for unwinding to avoid changing the stack frame around conditional code.
20553 @item -mdivsi3_libfunc=@var{name}
20554 @opindex mdivsi3_libfunc=@var{name}
20555 Set the name of the library function used for 32-bit signed division to
20557 This only affects the name used in the @samp{call} and @samp{inv:call}
20558 division strategies, and the compiler still expects the same
20559 sets of input/output/clobbered registers as if this option were not present.
20561 @item -mfixed-range=@var{register-range}
20562 @opindex mfixed-range
20563 Generate code treating the given register range as fixed registers.
20564 A fixed register is one that the register allocator can not use. This is
20565 useful when compiling kernel code. A register range is specified as
20566 two registers separated by a dash. Multiple register ranges can be
20567 specified separated by a comma.
20569 @item -mindexed-addressing
20570 @opindex mindexed-addressing
20571 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
20572 This is only safe if the hardware and/or OS implement 32-bit wrap-around
20573 semantics for the indexed addressing mode. The architecture allows the
20574 implementation of processors with 64-bit MMU, which the OS could use to
20575 get 32-bit addressing, but since no current hardware implementation supports
20576 this or any other way to make the indexed addressing mode safe to use in
20577 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
20579 @item -mgettrcost=@var{number}
20580 @opindex mgettrcost=@var{number}
20581 Set the cost assumed for the @code{gettr} instruction to @var{number}.
20582 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
20586 Assume @code{pt*} instructions won't trap. This generally generates
20587 better-scheduled code, but is unsafe on current hardware.
20588 The current architecture
20589 definition says that @code{ptabs} and @code{ptrel} trap when the target
20591 This has the unintentional effect of making it unsafe to schedule these
20592 instructions before a branch, or hoist them out of a loop. For example,
20593 @code{__do_global_ctors}, a part of @file{libgcc}
20594 that runs constructors at program
20595 startup, calls functions in a list which is delimited by @minus{}1. With the
20596 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
20597 That means that all the constructors run a bit more quickly, but when
20598 the loop comes to the end of the list, the program crashes because @code{ptabs}
20599 loads @minus{}1 into a target register.
20601 Since this option is unsafe for any
20602 hardware implementing the current architecture specification, the default
20603 is @option{-mno-pt-fixed}. Unless specified explicitly with
20604 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
20605 this deters register allocation from using target registers for storing
20608 @item -minvalid-symbols
20609 @opindex minvalid-symbols
20610 Assume symbols might be invalid. Ordinary function symbols generated by
20611 the compiler are always valid to load with
20612 @code{movi}/@code{shori}/@code{ptabs} or
20613 @code{movi}/@code{shori}/@code{ptrel},
20614 but with assembler and/or linker tricks it is possible
20615 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
20616 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
20617 It prevents cross-basic-block CSE, hoisting and most scheduling
20618 of symbol loads. The default is @option{-mno-invalid-symbols}.
20620 @item -mbranch-cost=@var{num}
20621 @opindex mbranch-cost=@var{num}
20622 Assume @var{num} to be the cost for a branch instruction. Higher numbers
20623 make the compiler try to generate more branch-free code if possible.
20624 If not specified the value is selected depending on the processor type that
20625 is being compiled for.
20628 @itemx -mno-zdcbranch
20629 @opindex mzdcbranch
20630 @opindex mno-zdcbranch
20631 Assume (do not assume) that zero displacement conditional branch instructions
20632 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
20633 compiler prefers zero displacement branch code sequences. This is
20634 enabled by default when generating code for SH4 and SH4A. It can be explicitly
20635 disabled by specifying @option{-mno-zdcbranch}.
20637 @item -mcbranch-force-delay-slot
20638 @opindex mcbranch-force-delay-slot
20639 Force the usage of delay slots for conditional branches, which stuffs the delay
20640 slot with a @code{nop} if a suitable instruction can't be found. By default
20641 this option is disabled. It can be enabled to work around hardware bugs as
20642 found in the original SH7055.
20645 @itemx -mno-fused-madd
20646 @opindex mfused-madd
20647 @opindex mno-fused-madd
20648 Generate code that uses (does not use) the floating-point multiply and
20649 accumulate instructions. These instructions are generated by default
20650 if hardware floating point is used. The machine-dependent
20651 @option{-mfused-madd} option is now mapped to the machine-independent
20652 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20653 mapped to @option{-ffp-contract=off}.
20659 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
20660 and cosine approximations. The option @option{-mfsca} must be used in
20661 combination with @option{-funsafe-math-optimizations}. It is enabled by default
20662 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
20663 approximations even if @option{-funsafe-math-optimizations} is in effect.
20669 Allow or disallow the compiler to emit the @code{fsrra} instruction for
20670 reciprocal square root approximations. The option @option{-mfsrra} must be used
20671 in combination with @option{-funsafe-math-optimizations} and
20672 @option{-ffinite-math-only}. It is enabled by default when generating code for
20673 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
20674 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
20677 @item -mpretend-cmove
20678 @opindex mpretend-cmove
20679 Prefer zero-displacement conditional branches for conditional move instruction
20680 patterns. This can result in faster code on the SH4 processor.
20684 @node Solaris 2 Options
20685 @subsection Solaris 2 Options
20686 @cindex Solaris 2 options
20688 These @samp{-m} options are supported on Solaris 2:
20691 @item -mclear-hwcap
20692 @opindex mclear-hwcap
20693 @option{-mclear-hwcap} tells the compiler to remove the hardware
20694 capabilities generated by the Solaris assembler. This is only necessary
20695 when object files use ISA extensions not supported by the current
20696 machine, but check at runtime whether or not to use them.
20698 @item -mimpure-text
20699 @opindex mimpure-text
20700 @option{-mimpure-text}, used in addition to @option{-shared}, tells
20701 the compiler to not pass @option{-z text} to the linker when linking a
20702 shared object. Using this option, you can link position-dependent
20703 code into a shared object.
20705 @option{-mimpure-text} suppresses the ``relocations remain against
20706 allocatable but non-writable sections'' linker error message.
20707 However, the necessary relocations trigger copy-on-write, and the
20708 shared object is not actually shared across processes. Instead of
20709 using @option{-mimpure-text}, you should compile all source code with
20710 @option{-fpic} or @option{-fPIC}.
20714 These switches are supported in addition to the above on Solaris 2:
20719 Add support for multithreading using the POSIX threads library. This
20720 option sets flags for both the preprocessor and linker. This option does
20721 not affect the thread safety of object code produced by the compiler or
20722 that of libraries supplied with it.
20726 This is a synonym for @option{-pthreads}.
20729 @node SPARC Options
20730 @subsection SPARC Options
20731 @cindex SPARC options
20733 These @samp{-m} options are supported on the SPARC:
20736 @item -mno-app-regs
20738 @opindex mno-app-regs
20740 Specify @option{-mapp-regs} to generate output using the global registers
20741 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
20742 global register 1, each global register 2 through 4 is then treated as an
20743 allocable register that is clobbered by function calls. This is the default.
20745 To be fully SVR4 ABI-compliant at the cost of some performance loss,
20746 specify @option{-mno-app-regs}. You should compile libraries and system
20747 software with this option.
20753 With @option{-mflat}, the compiler does not generate save/restore instructions
20754 and uses a ``flat'' or single register window model. This model is compatible
20755 with the regular register window model. The local registers and the input
20756 registers (0--5) are still treated as ``call-saved'' registers and are
20757 saved on the stack as needed.
20759 With @option{-mno-flat} (the default), the compiler generates save/restore
20760 instructions (except for leaf functions). This is the normal operating mode.
20763 @itemx -mhard-float
20765 @opindex mhard-float
20766 Generate output containing floating-point instructions. This is the
20770 @itemx -msoft-float
20772 @opindex msoft-float
20773 Generate output containing library calls for floating point.
20774 @strong{Warning:} the requisite libraries are not available for all SPARC
20775 targets. Normally the facilities of the machine's usual C compiler are
20776 used, but this cannot be done directly in cross-compilation. You must make
20777 your own arrangements to provide suitable library functions for
20778 cross-compilation. The embedded targets @samp{sparc-*-aout} and
20779 @samp{sparclite-*-*} do provide software floating-point support.
20781 @option{-msoft-float} changes the calling convention in the output file;
20782 therefore, it is only useful if you compile @emph{all} of a program with
20783 this option. In particular, you need to compile @file{libgcc.a}, the
20784 library that comes with GCC, with @option{-msoft-float} in order for
20787 @item -mhard-quad-float
20788 @opindex mhard-quad-float
20789 Generate output containing quad-word (long double) floating-point
20792 @item -msoft-quad-float
20793 @opindex msoft-quad-float
20794 Generate output containing library calls for quad-word (long double)
20795 floating-point instructions. The functions called are those specified
20796 in the SPARC ABI@. This is the default.
20798 As of this writing, there are no SPARC implementations that have hardware
20799 support for the quad-word floating-point instructions. They all invoke
20800 a trap handler for one of these instructions, and then the trap handler
20801 emulates the effect of the instruction. Because of the trap handler overhead,
20802 this is much slower than calling the ABI library routines. Thus the
20803 @option{-msoft-quad-float} option is the default.
20805 @item -mno-unaligned-doubles
20806 @itemx -munaligned-doubles
20807 @opindex mno-unaligned-doubles
20808 @opindex munaligned-doubles
20809 Assume that doubles have 8-byte alignment. This is the default.
20811 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
20812 alignment only if they are contained in another type, or if they have an
20813 absolute address. Otherwise, it assumes they have 4-byte alignment.
20814 Specifying this option avoids some rare compatibility problems with code
20815 generated by other compilers. It is not the default because it results
20816 in a performance loss, especially for floating-point code.
20819 @itemx -mno-user-mode
20820 @opindex muser-mode
20821 @opindex mno-user-mode
20822 Do not generate code that can only run in supervisor mode. This is relevant
20823 only for the @code{casa} instruction emitted for the LEON3 processor. The
20824 default is @option{-mno-user-mode}.
20826 @item -mno-faster-structs
20827 @itemx -mfaster-structs
20828 @opindex mno-faster-structs
20829 @opindex mfaster-structs
20830 With @option{-mfaster-structs}, the compiler assumes that structures
20831 should have 8-byte alignment. This enables the use of pairs of
20832 @code{ldd} and @code{std} instructions for copies in structure
20833 assignment, in place of twice as many @code{ld} and @code{st} pairs.
20834 However, the use of this changed alignment directly violates the SPARC
20835 ABI@. Thus, it's intended only for use on targets where the developer
20836 acknowledges that their resulting code is not directly in line with
20837 the rules of the ABI@.
20839 @item -mcpu=@var{cpu_type}
20841 Set the instruction set, register set, and instruction scheduling parameters
20842 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
20843 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
20844 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
20845 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
20846 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
20847 @samp{niagara3} and @samp{niagara4}.
20849 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
20850 which selects the best architecture option for the host processor.
20851 @option{-mcpu=native} has no effect if GCC does not recognize
20854 Default instruction scheduling parameters are used for values that select
20855 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
20856 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
20858 Here is a list of each supported architecture and their supported
20866 supersparc, hypersparc, leon, leon3
20869 f930, f934, sparclite86x
20875 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
20878 By default (unless configured otherwise), GCC generates code for the V7
20879 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
20880 additionally optimizes it for the Cypress CY7C602 chip, as used in the
20881 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
20882 SPARCStation 1, 2, IPX etc.
20884 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
20885 architecture. The only difference from V7 code is that the compiler emits
20886 the integer multiply and integer divide instructions which exist in SPARC-V8
20887 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
20888 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
20891 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
20892 the SPARC architecture. This adds the integer multiply, integer divide step
20893 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
20894 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
20895 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
20896 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
20897 MB86934 chip, which is the more recent SPARClite with FPU@.
20899 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
20900 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
20901 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
20902 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
20903 optimizes it for the TEMIC SPARClet chip.
20905 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
20906 architecture. This adds 64-bit integer and floating-point move instructions,
20907 3 additional floating-point condition code registers and conditional move
20908 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
20909 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
20910 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
20911 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
20912 @option{-mcpu=niagara}, the compiler additionally optimizes it for
20913 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
20914 additionally optimizes it for Sun UltraSPARC T2 chips. With
20915 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
20916 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
20917 additionally optimizes it for Sun UltraSPARC T4 chips.
20919 @item -mtune=@var{cpu_type}
20921 Set the instruction scheduling parameters for machine type
20922 @var{cpu_type}, but do not set the instruction set or register set that the
20923 option @option{-mcpu=@var{cpu_type}} does.
20925 The same values for @option{-mcpu=@var{cpu_type}} can be used for
20926 @option{-mtune=@var{cpu_type}}, but the only useful values are those
20927 that select a particular CPU implementation. Those are @samp{cypress},
20928 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
20929 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
20930 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
20931 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
20932 toolchains, @samp{native} can also be used.
20937 @opindex mno-v8plus
20938 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
20939 difference from the V8 ABI is that the global and out registers are
20940 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
20941 mode for all SPARC-V9 processors.
20947 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
20948 Visual Instruction Set extensions. The default is @option{-mno-vis}.
20954 With @option{-mvis2}, GCC generates code that takes advantage of
20955 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
20956 default is @option{-mvis2} when targeting a cpu that supports such
20957 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
20958 also sets @option{-mvis}.
20964 With @option{-mvis3}, GCC generates code that takes advantage of
20965 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
20966 default is @option{-mvis3} when targeting a cpu that supports such
20967 instructions, such as niagara-3 and later. Setting @option{-mvis3}
20968 also sets @option{-mvis2} and @option{-mvis}.
20973 @opindex mno-cbcond
20974 With @option{-mcbcond}, GCC generates code that takes advantage of
20975 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
20976 The default is @option{-mcbcond} when targeting a cpu that supports such
20977 instructions, such as niagara-4 and later.
20983 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
20984 population count instruction. The default is @option{-mpopc}
20985 when targeting a cpu that supports such instructions, such as Niagara-2 and
20992 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
20993 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
20994 when targeting a cpu that supports such instructions, such as Niagara-3 and
20998 @opindex mfix-at697f
20999 Enable the documented workaround for the single erratum of the Atmel AT697F
21000 processor (which corresponds to erratum #13 of the AT697E processor).
21003 @opindex mfix-ut699
21004 Enable the documented workarounds for the floating-point errata and the data
21005 cache nullify errata of the UT699 processor.
21008 These @samp{-m} options are supported in addition to the above
21009 on SPARC-V9 processors in 64-bit environments:
21016 Generate code for a 32-bit or 64-bit environment.
21017 The 32-bit environment sets int, long and pointer to 32 bits.
21018 The 64-bit environment sets int to 32 bits and long and pointer
21021 @item -mcmodel=@var{which}
21023 Set the code model to one of
21027 The Medium/Low code model: 64-bit addresses, programs
21028 must be linked in the low 32 bits of memory. Programs can be statically
21029 or dynamically linked.
21032 The Medium/Middle code model: 64-bit addresses, programs
21033 must be linked in the low 44 bits of memory, the text and data segments must
21034 be less than 2GB in size and the data segment must be located within 2GB of
21038 The Medium/Anywhere code model: 64-bit addresses, programs
21039 may be linked anywhere in memory, the text and data segments must be less
21040 than 2GB in size and the data segment must be located within 2GB of the
21044 The Medium/Anywhere code model for embedded systems:
21045 64-bit addresses, the text and data segments must be less than 2GB in
21046 size, both starting anywhere in memory (determined at link time). The
21047 global register %g4 points to the base of the data segment. Programs
21048 are statically linked and PIC is not supported.
21051 @item -mmemory-model=@var{mem-model}
21052 @opindex mmemory-model
21053 Set the memory model in force on the processor to one of
21057 The default memory model for the processor and operating system.
21060 Relaxed Memory Order
21063 Partial Store Order
21069 Sequential Consistency
21072 These memory models are formally defined in Appendix D of the Sparc V9
21073 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21076 @itemx -mno-stack-bias
21077 @opindex mstack-bias
21078 @opindex mno-stack-bias
21079 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21080 frame pointer if present, are offset by @minus{}2047 which must be added back
21081 when making stack frame references. This is the default in 64-bit mode.
21082 Otherwise, assume no such offset is present.
21086 @subsection SPU Options
21087 @cindex SPU options
21089 These @samp{-m} options are supported on the SPU:
21093 @itemx -merror-reloc
21094 @opindex mwarn-reloc
21095 @opindex merror-reloc
21097 The loader for SPU does not handle dynamic relocations. By default, GCC
21098 gives an error when it generates code that requires a dynamic
21099 relocation. @option{-mno-error-reloc} disables the error,
21100 @option{-mwarn-reloc} generates a warning instead.
21103 @itemx -munsafe-dma
21105 @opindex munsafe-dma
21107 Instructions that initiate or test completion of DMA must not be
21108 reordered with respect to loads and stores of the memory that is being
21110 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21111 memory accesses, but that can lead to inefficient code in places where the
21112 memory is known to not change. Rather than mark the memory as volatile,
21113 you can use @option{-msafe-dma} to tell the compiler to treat
21114 the DMA instructions as potentially affecting all memory.
21116 @item -mbranch-hints
21117 @opindex mbranch-hints
21119 By default, GCC generates a branch hint instruction to avoid
21120 pipeline stalls for always-taken or probably-taken branches. A hint
21121 is not generated closer than 8 instructions away from its branch.
21122 There is little reason to disable them, except for debugging purposes,
21123 or to make an object a little bit smaller.
21127 @opindex msmall-mem
21128 @opindex mlarge-mem
21130 By default, GCC generates code assuming that addresses are never larger
21131 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21132 a full 32-bit address.
21137 By default, GCC links against startup code that assumes the SPU-style
21138 main function interface (which has an unconventional parameter list).
21139 With @option{-mstdmain}, GCC links your program against startup
21140 code that assumes a C99-style interface to @code{main}, including a
21141 local copy of @code{argv} strings.
21143 @item -mfixed-range=@var{register-range}
21144 @opindex mfixed-range
21145 Generate code treating the given register range as fixed registers.
21146 A fixed register is one that the register allocator cannot use. This is
21147 useful when compiling kernel code. A register range is specified as
21148 two registers separated by a dash. Multiple register ranges can be
21149 specified separated by a comma.
21155 Compile code assuming that pointers to the PPU address space accessed
21156 via the @code{__ea} named address space qualifier are either 32 or 64
21157 bits wide. The default is 32 bits. As this is an ABI-changing option,
21158 all object code in an executable must be compiled with the same setting.
21160 @item -maddress-space-conversion
21161 @itemx -mno-address-space-conversion
21162 @opindex maddress-space-conversion
21163 @opindex mno-address-space-conversion
21164 Allow/disallow treating the @code{__ea} address space as superset
21165 of the generic address space. This enables explicit type casts
21166 between @code{__ea} and generic pointer as well as implicit
21167 conversions of generic pointers to @code{__ea} pointers. The
21168 default is to allow address space pointer conversions.
21170 @item -mcache-size=@var{cache-size}
21171 @opindex mcache-size
21172 This option controls the version of libgcc that the compiler links to an
21173 executable and selects a software-managed cache for accessing variables
21174 in the @code{__ea} address space with a particular cache size. Possible
21175 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21176 and @samp{128}. The default cache size is 64KB.
21178 @item -matomic-updates
21179 @itemx -mno-atomic-updates
21180 @opindex matomic-updates
21181 @opindex mno-atomic-updates
21182 This option controls the version of libgcc that the compiler links to an
21183 executable and selects whether atomic updates to the software-managed
21184 cache of PPU-side variables are used. If you use atomic updates, changes
21185 to a PPU variable from SPU code using the @code{__ea} named address space
21186 qualifier do not interfere with changes to other PPU variables residing
21187 in the same cache line from PPU code. If you do not use atomic updates,
21188 such interference may occur; however, writing back cache lines is
21189 more efficient. The default behavior is to use atomic updates.
21192 @itemx -mdual-nops=@var{n}
21193 @opindex mdual-nops
21194 By default, GCC inserts nops to increase dual issue when it expects
21195 it to increase performance. @var{n} can be a value from 0 to 10. A
21196 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21197 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21199 @item -mhint-max-nops=@var{n}
21200 @opindex mhint-max-nops
21201 Maximum number of nops to insert for a branch hint. A branch hint must
21202 be at least 8 instructions away from the branch it is affecting. GCC
21203 inserts up to @var{n} nops to enforce this, otherwise it does not
21204 generate the branch hint.
21206 @item -mhint-max-distance=@var{n}
21207 @opindex mhint-max-distance
21208 The encoding of the branch hint instruction limits the hint to be within
21209 256 instructions of the branch it is affecting. By default, GCC makes
21210 sure it is within 125.
21213 @opindex msafe-hints
21214 Work around a hardware bug that causes the SPU to stall indefinitely.
21215 By default, GCC inserts the @code{hbrp} instruction to make sure
21216 this stall won't happen.
21220 @node System V Options
21221 @subsection Options for System V
21223 These additional options are available on System V Release 4 for
21224 compatibility with other compilers on those systems:
21229 Create a shared object.
21230 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21234 Identify the versions of each tool used by the compiler, in a
21235 @code{.ident} assembler directive in the output.
21239 Refrain from adding @code{.ident} directives to the output file (this is
21242 @item -YP,@var{dirs}
21244 Search the directories @var{dirs}, and no others, for libraries
21245 specified with @option{-l}.
21247 @item -Ym,@var{dir}
21249 Look in the directory @var{dir} to find the M4 preprocessor.
21250 The assembler uses this option.
21251 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21252 @c the generic assembler that comes with Solaris takes just -Ym.
21255 @node TILE-Gx Options
21256 @subsection TILE-Gx Options
21257 @cindex TILE-Gx options
21259 These @samp{-m} options are supported on the TILE-Gx:
21262 @item -mcmodel=small
21263 @opindex mcmodel=small
21264 Generate code for the small model. The distance for direct calls is
21265 limited to 500M in either direction. PC-relative addresses are 32
21266 bits. Absolute addresses support the full address range.
21268 @item -mcmodel=large
21269 @opindex mcmodel=large
21270 Generate code for the large model. There is no limitation on call
21271 distance, pc-relative addresses, or absolute addresses.
21273 @item -mcpu=@var{name}
21275 Selects the type of CPU to be targeted. Currently the only supported
21276 type is @samp{tilegx}.
21282 Generate code for a 32-bit or 64-bit environment. The 32-bit
21283 environment sets int, long, and pointer to 32 bits. The 64-bit
21284 environment sets int to 32 bits and long and pointer to 64 bits.
21287 @itemx -mlittle-endian
21288 @opindex mbig-endian
21289 @opindex mlittle-endian
21290 Generate code in big/little endian mode, respectively.
21293 @node TILEPro Options
21294 @subsection TILEPro Options
21295 @cindex TILEPro options
21297 These @samp{-m} options are supported on the TILEPro:
21300 @item -mcpu=@var{name}
21302 Selects the type of CPU to be targeted. Currently the only supported
21303 type is @samp{tilepro}.
21307 Generate code for a 32-bit environment, which sets int, long, and
21308 pointer to 32 bits. This is the only supported behavior so the flag
21309 is essentially ignored.
21313 @subsection V850 Options
21314 @cindex V850 Options
21316 These @samp{-m} options are defined for V850 implementations:
21320 @itemx -mno-long-calls
21321 @opindex mlong-calls
21322 @opindex mno-long-calls
21323 Treat all calls as being far away (near). If calls are assumed to be
21324 far away, the compiler always loads the function's address into a
21325 register, and calls indirect through the pointer.
21331 Do not optimize (do optimize) basic blocks that use the same index
21332 pointer 4 or more times to copy pointer into the @code{ep} register, and
21333 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21334 option is on by default if you optimize.
21336 @item -mno-prolog-function
21337 @itemx -mprolog-function
21338 @opindex mno-prolog-function
21339 @opindex mprolog-function
21340 Do not use (do use) external functions to save and restore registers
21341 at the prologue and epilogue of a function. The external functions
21342 are slower, but use less code space if more than one function saves
21343 the same number of registers. The @option{-mprolog-function} option
21344 is on by default if you optimize.
21348 Try to make the code as small as possible. At present, this just turns
21349 on the @option{-mep} and @option{-mprolog-function} options.
21351 @item -mtda=@var{n}
21353 Put static or global variables whose size is @var{n} bytes or less into
21354 the tiny data area that register @code{ep} points to. The tiny data
21355 area can hold up to 256 bytes in total (128 bytes for byte references).
21357 @item -msda=@var{n}
21359 Put static or global variables whose size is @var{n} bytes or less into
21360 the small data area that register @code{gp} points to. The small data
21361 area can hold up to 64 kilobytes.
21363 @item -mzda=@var{n}
21365 Put static or global variables whose size is @var{n} bytes or less into
21366 the first 32 kilobytes of memory.
21370 Specify that the target processor is the V850.
21374 Specify that the target processor is the V850E3V5. The preprocessor
21375 constant @code{__v850e3v5__} is defined if this option is used.
21379 Specify that the target processor is the V850E3V5. This is an alias for
21380 the @option{-mv850e3v5} option.
21384 Specify that the target processor is the V850E2V3. The preprocessor
21385 constant @code{__v850e2v3__} is defined if this option is used.
21389 Specify that the target processor is the V850E2. The preprocessor
21390 constant @code{__v850e2__} is defined if this option is used.
21394 Specify that the target processor is the V850E1. The preprocessor
21395 constants @code{__v850e1__} and @code{__v850e__} are defined if
21396 this option is used.
21400 Specify that the target processor is the V850ES. This is an alias for
21401 the @option{-mv850e1} option.
21405 Specify that the target processor is the V850E@. The preprocessor
21406 constant @code{__v850e__} is defined if this option is used.
21408 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21409 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21410 are defined then a default target processor is chosen and the
21411 relevant @samp{__v850*__} preprocessor constant is defined.
21413 The preprocessor constants @code{__v850} and @code{__v851__} are always
21414 defined, regardless of which processor variant is the target.
21416 @item -mdisable-callt
21417 @itemx -mno-disable-callt
21418 @opindex mdisable-callt
21419 @opindex mno-disable-callt
21420 This option suppresses generation of the @code{CALLT} instruction for the
21421 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21424 This option is enabled by default when the RH850 ABI is
21425 in use (see @option{-mrh850-abi}), and disabled by default when the
21426 GCC ABI is in use. If @code{CALLT} instructions are being generated
21427 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
21433 Pass on (or do not pass on) the @option{-mrelax} command line option
21437 @itemx -mno-long-jumps
21438 @opindex mlong-jumps
21439 @opindex mno-long-jumps
21440 Disable (or re-enable) the generation of PC-relative jump instructions.
21443 @itemx -mhard-float
21444 @opindex msoft-float
21445 @opindex mhard-float
21446 Disable (or re-enable) the generation of hardware floating point
21447 instructions. This option is only significant when the target
21448 architecture is @samp{V850E2V3} or higher. If hardware floating point
21449 instructions are being generated then the C preprocessor symbol
21450 @code{__FPU_OK__} is defined, otherwise the symbol
21451 @code{__NO_FPU__} is defined.
21455 Enables the use of the e3v5 LOOP instruction. The use of this
21456 instruction is not enabled by default when the e3v5 architecture is
21457 selected because its use is still experimental.
21461 @opindex mrh850-abi
21463 Enables support for the RH850 version of the V850 ABI. This is the
21464 default. With this version of the ABI the following rules apply:
21468 Integer sized structures and unions are returned via a memory pointer
21469 rather than a register.
21472 Large structures and unions (more than 8 bytes in size) are passed by
21476 Functions are aligned to 16-bit boundaries.
21479 The @option{-m8byte-align} command line option is supported.
21482 The @option{-mdisable-callt} command line option is enabled by
21483 default. The @option{-mno-disable-callt} command line option is not
21487 When this version of the ABI is enabled the C preprocessor symbol
21488 @code{__V850_RH850_ABI__} is defined.
21492 Enables support for the old GCC version of the V850 ABI. With this
21493 version of the ABI the following rules apply:
21497 Integer sized structures and unions are returned in register @code{r10}.
21500 Large structures and unions (more than 8 bytes in size) are passed by
21504 Functions are aligned to 32-bit boundaries, unless optimizing for
21508 The @option{-m8byte-align} command line option is not supported.
21511 The @option{-mdisable-callt} command line option is supported but not
21512 enabled by default.
21515 When this version of the ABI is enabled the C preprocessor symbol
21516 @code{__V850_GCC_ABI__} is defined.
21518 @item -m8byte-align
21519 @itemx -mno-8byte-align
21520 @opindex m8byte-align
21521 @opindex mno-8byte-align
21522 Enables support for @code{double} and @code{long long} types to be
21523 aligned on 8-byte boundaries. The default is to restrict the
21524 alignment of all objects to at most 4-bytes. When
21525 @option{-m8byte-align} is in effect the C preprocessor symbol
21526 @code{__V850_8BYTE_ALIGN__} is defined.
21529 @opindex mbig-switch
21530 Generate code suitable for big switch tables. Use this option only if
21531 the assembler/linker complain about out of range branches within a switch
21536 This option causes r2 and r5 to be used in the code generated by
21537 the compiler. This setting is the default.
21539 @item -mno-app-regs
21540 @opindex mno-app-regs
21541 This option causes r2 and r5 to be treated as fixed registers.
21546 @subsection VAX Options
21547 @cindex VAX options
21549 These @samp{-m} options are defined for the VAX:
21554 Do not output certain jump instructions (@code{aobleq} and so on)
21555 that the Unix assembler for the VAX cannot handle across long
21560 Do output those jump instructions, on the assumption that the
21561 GNU assembler is being used.
21565 Output code for G-format floating-point numbers instead of D-format.
21568 @node Visium Options
21569 @subsection Visium Options
21570 @cindex Visium options
21576 A program which performs file I/O and is destined to run on an MCM target
21577 should be linked with this option. It causes the libraries libc.a and
21578 libdebug.a to be linked. The program should be run on the target under
21579 the control of the GDB remote debugging stub.
21583 A program which performs file I/O and is destined to run on the simulator
21584 should be linked with option. This causes libraries libc.a and libsim.a to
21588 @itemx -mhard-float
21590 @opindex mhard-float
21591 Generate code containing floating-point instructions. This is the
21595 @itemx -msoft-float
21597 @opindex msoft-float
21598 Generate code containing library calls for floating-point.
21600 @option{-msoft-float} changes the calling convention in the output file;
21601 therefore, it is only useful if you compile @emph{all} of a program with
21602 this option. In particular, you need to compile @file{libgcc.a}, the
21603 library that comes with GCC, with @option{-msoft-float} in order for
21606 @item -mcpu=@var{cpu_type}
21608 Set the instruction set, register set, and instruction scheduling parameters
21609 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21610 @samp{mcm}, @samp{gr5} and @samp{gr6}.
21612 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
21614 By default (unless configured otherwise), GCC generates code for the GR5
21615 variant of the Visium architecture.
21617 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
21618 architecture. The only difference from GR5 code is that the compiler will
21619 generate block move instructions.
21621 @item -mtune=@var{cpu_type}
21623 Set the instruction scheduling parameters for machine type @var{cpu_type},
21624 but do not set the instruction set or register set that the option
21625 @option{-mcpu=@var{cpu_type}} would.
21629 Generate code for the supervisor mode, where there are no restrictions on
21630 the access to general registers. This is the default.
21633 @opindex muser-mode
21634 Generate code for the user mode, where the access to some general registers
21635 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
21636 mode; on the GR6, only registers r29 to r31 are affected.
21640 @subsection VMS Options
21642 These @samp{-m} options are defined for the VMS implementations:
21645 @item -mvms-return-codes
21646 @opindex mvms-return-codes
21647 Return VMS condition codes from @code{main}. The default is to return POSIX-style
21648 condition (e.g.@ error) codes.
21650 @item -mdebug-main=@var{prefix}
21651 @opindex mdebug-main=@var{prefix}
21652 Flag the first routine whose name starts with @var{prefix} as the main
21653 routine for the debugger.
21657 Default to 64-bit memory allocation routines.
21659 @item -mpointer-size=@var{size}
21660 @opindex mpointer-size=@var{size}
21661 Set the default size of pointers. Possible options for @var{size} are
21662 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
21663 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
21664 The later option disables @code{pragma pointer_size}.
21667 @node VxWorks Options
21668 @subsection VxWorks Options
21669 @cindex VxWorks Options
21671 The options in this section are defined for all VxWorks targets.
21672 Options specific to the target hardware are listed with the other
21673 options for that target.
21678 GCC can generate code for both VxWorks kernels and real time processes
21679 (RTPs). This option switches from the former to the latter. It also
21680 defines the preprocessor macro @code{__RTP__}.
21683 @opindex non-static
21684 Link an RTP executable against shared libraries rather than static
21685 libraries. The options @option{-static} and @option{-shared} can
21686 also be used for RTPs (@pxref{Link Options}); @option{-static}
21693 These options are passed down to the linker. They are defined for
21694 compatibility with Diab.
21697 @opindex Xbind-lazy
21698 Enable lazy binding of function calls. This option is equivalent to
21699 @option{-Wl,-z,now} and is defined for compatibility with Diab.
21703 Disable lazy binding of function calls. This option is the default and
21704 is defined for compatibility with Diab.
21708 @subsection x86 Options
21709 @cindex x86 Options
21711 These @samp{-m} options are defined for the x86 family of computers.
21715 @item -march=@var{cpu-type}
21717 Generate instructions for the machine type @var{cpu-type}. In contrast to
21718 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
21719 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
21720 to generate code that may not run at all on processors other than the one
21721 indicated. Specifying @option{-march=@var{cpu-type}} implies
21722 @option{-mtune=@var{cpu-type}}.
21724 The choices for @var{cpu-type} are:
21728 This selects the CPU to generate code for at compilation time by determining
21729 the processor type of the compiling machine. Using @option{-march=native}
21730 enables all instruction subsets supported by the local machine (hence
21731 the result might not run on different machines). Using @option{-mtune=native}
21732 produces code optimized for the local machine under the constraints
21733 of the selected instruction set.
21736 Original Intel i386 CPU@.
21739 Intel i486 CPU@. (No scheduling is implemented for this chip.)
21743 Intel Pentium CPU with no MMX support.
21746 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
21749 Intel Pentium Pro CPU@.
21752 When used with @option{-march}, the Pentium Pro
21753 instruction set is used, so the code runs on all i686 family chips.
21754 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
21757 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
21762 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
21766 Intel Pentium M; low-power version of Intel Pentium III CPU
21767 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
21771 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
21774 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
21778 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
21779 SSE2 and SSE3 instruction set support.
21782 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
21783 instruction set support.
21786 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21787 SSE4.1, SSE4.2 and POPCNT instruction set support.
21790 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21791 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
21794 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21795 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
21798 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
21799 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
21800 instruction set support.
21803 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21804 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
21805 BMI, BMI2 and F16C instruction set support.
21808 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21809 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
21810 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
21813 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
21814 instruction set support.
21817 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
21818 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
21821 AMD K6 CPU with MMX instruction set support.
21825 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
21828 @itemx athlon-tbird
21829 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
21835 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
21836 instruction set support.
21842 Processors based on the AMD K8 core with x86-64 instruction set support,
21843 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
21844 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
21845 instruction set extensions.)
21848 @itemx opteron-sse3
21849 @itemx athlon64-sse3
21850 Improved versions of AMD K8 cores with SSE3 instruction set support.
21854 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
21855 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
21856 instruction set extensions.)
21859 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
21860 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
21861 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
21863 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
21864 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
21865 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
21868 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
21869 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
21870 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
21871 64-bit instruction set extensions.
21873 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
21874 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
21875 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
21876 SSE4.2, ABM and 64-bit instruction set extensions.
21879 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
21880 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
21881 instruction set extensions.)
21884 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
21885 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
21886 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
21889 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
21893 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
21894 instruction set support.
21897 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
21898 implemented for this chip.)
21901 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
21903 implemented for this chip.)
21906 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
21909 @item -mtune=@var{cpu-type}
21911 Tune to @var{cpu-type} everything applicable about the generated code, except
21912 for the ABI and the set of available instructions.
21913 While picking a specific @var{cpu-type} schedules things appropriately
21914 for that particular chip, the compiler does not generate any code that
21915 cannot run on the default machine type unless you use a
21916 @option{-march=@var{cpu-type}} option.
21917 For example, if GCC is configured for i686-pc-linux-gnu
21918 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
21919 but still runs on i686 machines.
21921 The choices for @var{cpu-type} are the same as for @option{-march}.
21922 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
21926 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
21927 If you know the CPU on which your code will run, then you should use
21928 the corresponding @option{-mtune} or @option{-march} option instead of
21929 @option{-mtune=generic}. But, if you do not know exactly what CPU users
21930 of your application will have, then you should use this option.
21932 As new processors are deployed in the marketplace, the behavior of this
21933 option will change. Therefore, if you upgrade to a newer version of
21934 GCC, code generation controlled by this option will change to reflect
21936 that are most common at the time that version of GCC is released.
21938 There is no @option{-march=generic} option because @option{-march}
21939 indicates the instruction set the compiler can use, and there is no
21940 generic instruction set applicable to all processors. In contrast,
21941 @option{-mtune} indicates the processor (or, in this case, collection of
21942 processors) for which the code is optimized.
21945 Produce code optimized for the most current Intel processors, which are
21946 Haswell and Silvermont for this version of GCC. If you know the CPU
21947 on which your code will run, then you should use the corresponding
21948 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
21949 But, if you want your application performs better on both Haswell and
21950 Silvermont, then you should use this option.
21952 As new Intel processors are deployed in the marketplace, the behavior of
21953 this option will change. Therefore, if you upgrade to a newer version of
21954 GCC, code generation controlled by this option will change to reflect
21955 the most current Intel processors at the time that version of GCC is
21958 There is no @option{-march=intel} option because @option{-march} indicates
21959 the instruction set the compiler can use, and there is no common
21960 instruction set applicable to all processors. In contrast,
21961 @option{-mtune} indicates the processor (or, in this case, collection of
21962 processors) for which the code is optimized.
21965 @item -mcpu=@var{cpu-type}
21967 A deprecated synonym for @option{-mtune}.
21969 @item -mfpmath=@var{unit}
21971 Generate floating-point arithmetic for selected unit @var{unit}. The choices
21972 for @var{unit} are:
21976 Use the standard 387 floating-point coprocessor present on the majority of chips and
21977 emulated otherwise. Code compiled with this option runs almost everywhere.
21978 The temporary results are computed in 80-bit precision instead of the precision
21979 specified by the type, resulting in slightly different results compared to most
21980 of other chips. See @option{-ffloat-store} for more detailed description.
21982 This is the default choice for x86-32 targets.
21985 Use scalar floating-point instructions present in the SSE instruction set.
21986 This instruction set is supported by Pentium III and newer chips,
21987 and in the AMD line
21988 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
21989 instruction set supports only single-precision arithmetic, thus the double and
21990 extended-precision arithmetic are still done using 387. A later version, present
21991 only in Pentium 4 and AMD x86-64 chips, supports double-precision
21994 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
21995 or @option{-msse2} switches to enable SSE extensions and make this option
21996 effective. For the x86-64 compiler, these extensions are enabled by default.
21998 The resulting code should be considerably faster in the majority of cases and avoid
21999 the numerical instability problems of 387 code, but may break some existing
22000 code that expects temporaries to be 80 bits.
22002 This is the default choice for the x86-64 compiler.
22007 Attempt to utilize both instruction sets at once. This effectively doubles the
22008 amount of available registers, and on chips with separate execution units for
22009 387 and SSE the execution resources too. Use this option with care, as it is
22010 still experimental, because the GCC register allocator does not model separate
22011 functional units well, resulting in unstable performance.
22014 @item -masm=@var{dialect}
22015 @opindex masm=@var{dialect}
22016 Output assembly instructions using selected @var{dialect}. Also affects
22017 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22018 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22019 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22020 not support @samp{intel}.
22023 @itemx -mno-ieee-fp
22025 @opindex mno-ieee-fp
22026 Control whether or not the compiler uses IEEE floating-point
22027 comparisons. These correctly handle the case where the result of a
22028 comparison is unordered.
22031 @opindex msoft-float
22032 Generate output containing library calls for floating point.
22034 @strong{Warning:} the requisite libraries are not part of GCC@.
22035 Normally the facilities of the machine's usual C compiler are used, but
22036 this can't be done directly in cross-compilation. You must make your
22037 own arrangements to provide suitable library functions for
22040 On machines where a function returns floating-point results in the 80387
22041 register stack, some floating-point opcodes may be emitted even if
22042 @option{-msoft-float} is used.
22044 @item -mno-fp-ret-in-387
22045 @opindex mno-fp-ret-in-387
22046 Do not use the FPU registers for return values of functions.
22048 The usual calling convention has functions return values of types
22049 @code{float} and @code{double} in an FPU register, even if there
22050 is no FPU@. The idea is that the operating system should emulate
22053 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22054 in ordinary CPU registers instead.
22056 @item -mno-fancy-math-387
22057 @opindex mno-fancy-math-387
22058 Some 387 emulators do not support the @code{sin}, @code{cos} and
22059 @code{sqrt} instructions for the 387. Specify this option to avoid
22060 generating those instructions. This option is the default on FreeBSD,
22061 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22062 indicates that the target CPU always has an FPU and so the
22063 instruction does not need emulation. These
22064 instructions are not generated unless you also use the
22065 @option{-funsafe-math-optimizations} switch.
22067 @item -malign-double
22068 @itemx -mno-align-double
22069 @opindex malign-double
22070 @opindex mno-align-double
22071 Control whether GCC aligns @code{double}, @code{long double}, and
22072 @code{long long} variables on a two-word boundary or a one-word
22073 boundary. Aligning @code{double} variables on a two-word boundary
22074 produces code that runs somewhat faster on a Pentium at the
22075 expense of more memory.
22077 On x86-64, @option{-malign-double} is enabled by default.
22079 @strong{Warning:} if you use the @option{-malign-double} switch,
22080 structures containing the above types are aligned differently than
22081 the published application binary interface specifications for the x86-32
22082 and are not binary compatible with structures in code compiled
22083 without that switch.
22085 @item -m96bit-long-double
22086 @itemx -m128bit-long-double
22087 @opindex m96bit-long-double
22088 @opindex m128bit-long-double
22089 These switches control the size of @code{long double} type. The x86-32
22090 application binary interface specifies the size to be 96 bits,
22091 so @option{-m96bit-long-double} is the default in 32-bit mode.
22093 Modern architectures (Pentium and newer) prefer @code{long double}
22094 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22095 conforming to the ABI, this is not possible. So specifying
22096 @option{-m128bit-long-double} aligns @code{long double}
22097 to a 16-byte boundary by padding the @code{long double} with an additional
22100 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22101 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22103 Notice that neither of these options enable any extra precision over the x87
22104 standard of 80 bits for a @code{long double}.
22106 @strong{Warning:} if you override the default value for your target ABI, this
22107 changes the size of
22108 structures and arrays containing @code{long double} variables,
22109 as well as modifying the function calling convention for functions taking
22110 @code{long double}. Hence they are not binary-compatible
22111 with code compiled without that switch.
22113 @item -mlong-double-64
22114 @itemx -mlong-double-80
22115 @itemx -mlong-double-128
22116 @opindex mlong-double-64
22117 @opindex mlong-double-80
22118 @opindex mlong-double-128
22119 These switches control the size of @code{long double} type. A size
22120 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22121 type. This is the default for 32-bit Bionic C library. A size
22122 of 128 bits makes the @code{long double} type equivalent to the
22123 @code{__float128} type. This is the default for 64-bit Bionic C library.
22125 @strong{Warning:} if you override the default value for your target ABI, this
22126 changes the size of
22127 structures and arrays containing @code{long double} variables,
22128 as well as modifying the function calling convention for functions taking
22129 @code{long double}. Hence they are not binary-compatible
22130 with code compiled without that switch.
22132 @item -malign-data=@var{type}
22133 @opindex malign-data
22134 Control how GCC aligns variables. Supported values for @var{type} are
22135 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22136 and earlier, @samp{abi} uses alignment value as specified by the
22137 psABI, and @samp{cacheline} uses increased alignment value to match
22138 the cache line size. @samp{compat} is the default.
22140 @item -mlarge-data-threshold=@var{threshold}
22141 @opindex mlarge-data-threshold
22142 When @option{-mcmodel=medium} is specified, data objects larger than
22143 @var{threshold} are placed in the large data section. This value must be the
22144 same across all objects linked into the binary, and defaults to 65535.
22148 Use a different function-calling convention, in which functions that
22149 take a fixed number of arguments return with the @code{ret @var{num}}
22150 instruction, which pops their arguments while returning. This saves one
22151 instruction in the caller since there is no need to pop the arguments
22154 You can specify that an individual function is called with this calling
22155 sequence with the function attribute @code{stdcall}. You can also
22156 override the @option{-mrtd} option by using the function attribute
22157 @code{cdecl}. @xref{Function Attributes}.
22159 @strong{Warning:} this calling convention is incompatible with the one
22160 normally used on Unix, so you cannot use it if you need to call
22161 libraries compiled with the Unix compiler.
22163 Also, you must provide function prototypes for all functions that
22164 take variable numbers of arguments (including @code{printf});
22165 otherwise incorrect code is generated for calls to those
22168 In addition, seriously incorrect code results if you call a
22169 function with too many arguments. (Normally, extra arguments are
22170 harmlessly ignored.)
22172 @item -mregparm=@var{num}
22174 Control how many registers are used to pass integer arguments. By
22175 default, no registers are used to pass arguments, and at most 3
22176 registers can be used. You can control this behavior for a specific
22177 function by using the function attribute @code{regparm}.
22178 @xref{Function Attributes}.
22180 @strong{Warning:} if you use this switch, and
22181 @var{num} is nonzero, then you must build all modules with the same
22182 value, including any libraries. This includes the system libraries and
22186 @opindex msseregparm
22187 Use SSE register passing conventions for float and double arguments
22188 and return values. You can control this behavior for a specific
22189 function by using the function attribute @code{sseregparm}.
22190 @xref{Function Attributes}.
22192 @strong{Warning:} if you use this switch then you must build all
22193 modules with the same value, including any libraries. This includes
22194 the system libraries and startup modules.
22196 @item -mvect8-ret-in-mem
22197 @opindex mvect8-ret-in-mem
22198 Return 8-byte vectors in memory instead of MMX registers. This is the
22199 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
22200 Studio compilers until version 12. Later compiler versions (starting
22201 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
22202 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
22203 you need to remain compatible with existing code produced by those
22204 previous compiler versions or older versions of GCC@.
22213 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
22214 is specified, the significands of results of floating-point operations are
22215 rounded to 24 bits (single precision); @option{-mpc64} rounds the
22216 significands of results of floating-point operations to 53 bits (double
22217 precision) and @option{-mpc80} rounds the significands of results of
22218 floating-point operations to 64 bits (extended double precision), which is
22219 the default. When this option is used, floating-point operations in higher
22220 precisions are not available to the programmer without setting the FPU
22221 control word explicitly.
22223 Setting the rounding of floating-point operations to less than the default
22224 80 bits can speed some programs by 2% or more. Note that some mathematical
22225 libraries assume that extended-precision (80-bit) floating-point operations
22226 are enabled by default; routines in such libraries could suffer significant
22227 loss of accuracy, typically through so-called ``catastrophic cancellation'',
22228 when this option is used to set the precision to less than extended precision.
22230 @item -mstackrealign
22231 @opindex mstackrealign
22232 Realign the stack at entry. On the x86, the @option{-mstackrealign}
22233 option generates an alternate prologue and epilogue that realigns the
22234 run-time stack if necessary. This supports mixing legacy codes that keep
22235 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
22236 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
22237 applicable to individual functions.
22239 @item -mpreferred-stack-boundary=@var{num}
22240 @opindex mpreferred-stack-boundary
22241 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
22242 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
22243 the default is 4 (16 bytes or 128 bits).
22245 @strong{Warning:} When generating code for the x86-64 architecture with
22246 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
22247 used to keep the stack boundary aligned to 8 byte boundary. Since
22248 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
22249 intended to be used in controlled environment where stack space is
22250 important limitation. This option leads to wrong code when functions
22251 compiled with 16 byte stack alignment (such as functions from a standard
22252 library) are called with misaligned stack. In this case, SSE
22253 instructions may lead to misaligned memory access traps. In addition,
22254 variable arguments are handled incorrectly for 16 byte aligned
22255 objects (including x87 long double and __int128), leading to wrong
22256 results. You must build all modules with
22257 @option{-mpreferred-stack-boundary=3}, including any libraries. This
22258 includes the system libraries and startup modules.
22260 @item -mincoming-stack-boundary=@var{num}
22261 @opindex mincoming-stack-boundary
22262 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
22263 boundary. If @option{-mincoming-stack-boundary} is not specified,
22264 the one specified by @option{-mpreferred-stack-boundary} is used.
22266 On Pentium and Pentium Pro, @code{double} and @code{long double} values
22267 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
22268 suffer significant run time performance penalties. On Pentium III, the
22269 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
22270 properly if it is not 16-byte aligned.
22272 To ensure proper alignment of this values on the stack, the stack boundary
22273 must be as aligned as that required by any value stored on the stack.
22274 Further, every function must be generated such that it keeps the stack
22275 aligned. Thus calling a function compiled with a higher preferred
22276 stack boundary from a function compiled with a lower preferred stack
22277 boundary most likely misaligns the stack. It is recommended that
22278 libraries that use callbacks always use the default setting.
22280 This extra alignment does consume extra stack space, and generally
22281 increases code size. Code that is sensitive to stack space usage, such
22282 as embedded systems and operating system kernels, may want to reduce the
22283 preferred alignment to @option{-mpreferred-stack-boundary=2}.
22329 @opindex mclfushopt
22347 @itemx -mprefetchwt1
22348 @opindex mprefetchwt1
22396 These switches enable the use of instructions in the MMX, SSE,
22397 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
22398 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
22399 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX or 3DNow!@:
22400 extended instruction sets. Each has a corresponding @option{-mno-} option
22401 to disable use of these instructions.
22403 These extensions are also available as built-in functions: see
22404 @ref{x86 Built-in Functions}, for details of the functions enabled and
22405 disabled by these switches.
22407 To generate SSE/SSE2 instructions automatically from floating-point
22408 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
22410 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
22411 generates new AVX instructions or AVX equivalence for all SSEx instructions
22414 These options enable GCC to use these extended instructions in
22415 generated code, even without @option{-mfpmath=sse}. Applications that
22416 perform run-time CPU detection must compile separate files for each
22417 supported architecture, using the appropriate flags. In particular,
22418 the file containing the CPU detection code should be compiled without
22421 @item -mdump-tune-features
22422 @opindex mdump-tune-features
22423 This option instructs GCC to dump the names of the x86 performance
22424 tuning features and default settings. The names can be used in
22425 @option{-mtune-ctrl=@var{feature-list}}.
22427 @item -mtune-ctrl=@var{feature-list}
22428 @opindex mtune-ctrl=@var{feature-list}
22429 This option is used to do fine grain control of x86 code generation features.
22430 @var{feature-list} is a comma separated list of @var{feature} names. See also
22431 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
22432 on if it is not preceded with @samp{^}, otherwise, it is turned off.
22433 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
22434 developers. Using it may lead to code paths not covered by testing and can
22435 potentially result in compiler ICEs or runtime errors.
22438 @opindex mno-default
22439 This option instructs GCC to turn off all tunable features. See also
22440 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
22444 This option instructs GCC to emit a @code{cld} instruction in the prologue
22445 of functions that use string instructions. String instructions depend on
22446 the DF flag to select between autoincrement or autodecrement mode. While the
22447 ABI specifies the DF flag to be cleared on function entry, some operating
22448 systems violate this specification by not clearing the DF flag in their
22449 exception dispatchers. The exception handler can be invoked with the DF flag
22450 set, which leads to wrong direction mode when string instructions are used.
22451 This option can be enabled by default on 32-bit x86 targets by configuring
22452 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
22453 instructions can be suppressed with the @option{-mno-cld} compiler option
22457 @opindex mvzeroupper
22458 This option instructs GCC to emit a @code{vzeroupper} instruction
22459 before a transfer of control flow out of the function to minimize
22460 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
22463 @item -mprefer-avx128
22464 @opindex mprefer-avx128
22465 This option instructs GCC to use 128-bit AVX instructions instead of
22466 256-bit AVX instructions in the auto-vectorizer.
22470 This option enables GCC to generate @code{CMPXCHG16B} instructions.
22471 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
22472 (or oword) data types.
22473 This is useful for high-resolution counters that can be updated
22474 by multiple processors (or cores). This instruction is generated as part of
22475 atomic built-in functions: see @ref{__sync Builtins} or
22476 @ref{__atomic Builtins} for details.
22480 This option enables generation of @code{SAHF} instructions in 64-bit code.
22481 Early Intel Pentium 4 CPUs with Intel 64 support,
22482 prior to the introduction of Pentium 4 G1 step in December 2005,
22483 lacked the @code{LAHF} and @code{SAHF} instructions
22484 which are supported by AMD64.
22485 These are load and store instructions, respectively, for certain status flags.
22486 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
22487 @code{drem}, and @code{remainder} built-in functions;
22488 see @ref{Other Builtins} for details.
22492 This option enables use of the @code{movbe} instruction to implement
22493 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
22497 This option enables built-in functions @code{__builtin_ia32_crc32qi},
22498 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
22499 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
22503 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
22504 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
22505 with an additional Newton-Raphson step
22506 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
22507 (and their vectorized
22508 variants) for single-precision floating-point arguments. These instructions
22509 are generated only when @option{-funsafe-math-optimizations} is enabled
22510 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
22511 Note that while the throughput of the sequence is higher than the throughput
22512 of the non-reciprocal instruction, the precision of the sequence can be
22513 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
22515 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
22516 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
22517 combination), and doesn't need @option{-mrecip}.
22519 Also note that GCC emits the above sequence with additional Newton-Raphson step
22520 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
22521 already with @option{-ffast-math} (or the above option combination), and
22522 doesn't need @option{-mrecip}.
22524 @item -mrecip=@var{opt}
22525 @opindex mrecip=opt
22526 This option controls which reciprocal estimate instructions
22527 may be used. @var{opt} is a comma-separated list of options, which may
22528 be preceded by a @samp{!} to invert the option:
22532 Enable all estimate instructions.
22535 Enable the default instructions, equivalent to @option{-mrecip}.
22538 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22541 Enable the approximation for scalar division.
22544 Enable the approximation for vectorized division.
22547 Enable the approximation for scalar square root.
22550 Enable the approximation for vectorized square root.
22553 So, for example, @option{-mrecip=all,!sqrt} enables
22554 all of the reciprocal approximations, except for square root.
22556 @item -mveclibabi=@var{type}
22557 @opindex mveclibabi
22558 Specifies the ABI type to use for vectorizing intrinsics using an
22559 external library. Supported values for @var{type} are @samp{svml}
22560 for the Intel short
22561 vector math library and @samp{acml} for the AMD math core library.
22562 To use this option, both @option{-ftree-vectorize} and
22563 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
22564 ABI-compatible library must be specified at link time.
22566 GCC currently emits calls to @code{vmldExp2},
22567 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
22568 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
22569 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
22570 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
22571 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
22572 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
22573 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
22574 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
22575 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
22576 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
22577 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
22578 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
22579 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
22580 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
22581 when @option{-mveclibabi=acml} is used.
22583 @item -mabi=@var{name}
22585 Generate code for the specified calling convention. Permissible values
22586 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
22587 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
22588 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
22589 You can control this behavior for specific functions by
22590 using the function attributes @code{ms_abi} and @code{sysv_abi}.
22591 @xref{Function Attributes}.
22593 @item -mtls-dialect=@var{type}
22594 @opindex mtls-dialect
22595 Generate code to access thread-local storage using the @samp{gnu} or
22596 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
22597 @samp{gnu2} is more efficient, but it may add compile- and run-time
22598 requirements that cannot be satisfied on all systems.
22601 @itemx -mno-push-args
22602 @opindex mpush-args
22603 @opindex mno-push-args
22604 Use PUSH operations to store outgoing parameters. This method is shorter
22605 and usually equally fast as method using SUB/MOV operations and is enabled
22606 by default. In some cases disabling it may improve performance because of
22607 improved scheduling and reduced dependencies.
22609 @item -maccumulate-outgoing-args
22610 @opindex maccumulate-outgoing-args
22611 If enabled, the maximum amount of space required for outgoing arguments is
22612 computed in the function prologue. This is faster on most modern CPUs
22613 because of reduced dependencies, improved scheduling and reduced stack usage
22614 when the preferred stack boundary is not equal to 2. The drawback is a notable
22615 increase in code size. This switch implies @option{-mno-push-args}.
22619 Support thread-safe exception handling on MinGW. Programs that rely
22620 on thread-safe exception handling must compile and link all code with the
22621 @option{-mthreads} option. When compiling, @option{-mthreads} defines
22622 @option{-D_MT}; when linking, it links in a special thread helper library
22623 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
22625 @item -mno-align-stringops
22626 @opindex mno-align-stringops
22627 Do not align the destination of inlined string operations. This switch reduces
22628 code size and improves performance in case the destination is already aligned,
22629 but GCC doesn't know about it.
22631 @item -minline-all-stringops
22632 @opindex minline-all-stringops
22633 By default GCC inlines string operations only when the destination is
22634 known to be aligned to least a 4-byte boundary.
22635 This enables more inlining and increases code
22636 size, but may improve performance of code that depends on fast
22637 @code{memcpy}, @code{strlen},
22638 and @code{memset} for short lengths.
22640 @item -minline-stringops-dynamically
22641 @opindex minline-stringops-dynamically
22642 For string operations of unknown size, use run-time checks with
22643 inline code for small blocks and a library call for large blocks.
22645 @item -mstringop-strategy=@var{alg}
22646 @opindex mstringop-strategy=@var{alg}
22647 Override the internal decision heuristic for the particular algorithm to use
22648 for inlining string operations. The allowed values for @var{alg} are:
22654 Expand using i386 @code{rep} prefix of the specified size.
22658 @itemx unrolled_loop
22659 Expand into an inline loop.
22662 Always use a library call.
22665 @item -mmemcpy-strategy=@var{strategy}
22666 @opindex mmemcpy-strategy=@var{strategy}
22667 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
22668 should be inlined and what inline algorithm to use when the expected size
22669 of the copy operation is known. @var{strategy}
22670 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
22671 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
22672 the max byte size with which inline algorithm @var{alg} is allowed. For the last
22673 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
22674 in the list must be specified in increasing order. The minimal byte size for
22675 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
22678 @item -mmemset-strategy=@var{strategy}
22679 @opindex mmemset-strategy=@var{strategy}
22680 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
22681 @code{__builtin_memset} expansion.
22683 @item -momit-leaf-frame-pointer
22684 @opindex momit-leaf-frame-pointer
22685 Don't keep the frame pointer in a register for leaf functions. This
22686 avoids the instructions to save, set up, and restore frame pointers and
22687 makes an extra register available in leaf functions. The option
22688 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
22689 which might make debugging harder.
22691 @item -mtls-direct-seg-refs
22692 @itemx -mno-tls-direct-seg-refs
22693 @opindex mtls-direct-seg-refs
22694 Controls whether TLS variables may be accessed with offsets from the
22695 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
22696 or whether the thread base pointer must be added. Whether or not this
22697 is valid depends on the operating system, and whether it maps the
22698 segment to cover the entire TLS area.
22700 For systems that use the GNU C Library, the default is on.
22703 @itemx -mno-sse2avx
22705 Specify that the assembler should encode SSE instructions with VEX
22706 prefix. The option @option{-mavx} turns this on by default.
22711 If profiling is active (@option{-pg}), put the profiling
22712 counter call before the prologue.
22713 Note: On x86 architectures the attribute @code{ms_hook_prologue}
22714 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
22716 @item -mrecord-mcount
22717 @itemx -mno-record-mcount
22718 @opindex mrecord-mcount
22719 If profiling is active (@option{-pg}), generate a __mcount_loc section
22720 that contains pointers to each profiling call. This is useful for
22721 automatically patching and out calls.
22724 @itemx -mno-nop-mcount
22725 @opindex mnop-mcount
22726 If profiling is active (@option{-pg}), generate the calls to
22727 the profiling functions as nops. This is useful when they
22728 should be patched in later dynamically. This is likely only
22729 useful together with @option{-mrecord-mcount}.
22731 @item -mskip-rax-setup
22732 @itemx -mno-skip-rax-setup
22733 @opindex mskip-rax-setup
22734 When generating code for the x86-64 architecture with SSE extensions
22735 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
22736 register when there are no variable arguments passed in vector registers.
22738 @strong{Warning:} Since RAX register is used to avoid unnecessarily
22739 saving vector registers on stack when passing variable arguments, the
22740 impacts of this option are callees may waste some stack space,
22741 misbehave or jump to a random location. GCC 4.4 or newer don't have
22742 those issues, regardless the RAX register value.
22745 @itemx -mno-8bit-idiv
22746 @opindex m8bit-idiv
22747 On some processors, like Intel Atom, 8-bit unsigned integer divide is
22748 much faster than 32-bit/64-bit integer divide. This option generates a
22749 run-time check. If both dividend and divisor are within range of 0
22750 to 255, 8-bit unsigned integer divide is used instead of
22751 32-bit/64-bit integer divide.
22753 @item -mavx256-split-unaligned-load
22754 @itemx -mavx256-split-unaligned-store
22755 @opindex mavx256-split-unaligned-load
22756 @opindex mavx256-split-unaligned-store
22757 Split 32-byte AVX unaligned load and store.
22759 @item -mstack-protector-guard=@var{guard}
22760 @opindex mstack-protector-guard=@var{guard}
22761 Generate stack protection code using canary at @var{guard}. Supported
22762 locations are @samp{global} for global canary or @samp{tls} for per-thread
22763 canary in the TLS block (the default). This option has effect only when
22764 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
22768 These @samp{-m} switches are supported in addition to the above
22769 on x86-64 processors in 64-bit environments.
22780 Generate code for a 16-bit, 32-bit or 64-bit environment.
22781 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
22783 generates code that runs on any i386 system.
22785 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
22786 types to 64 bits, and generates code for the x86-64 architecture.
22787 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
22788 and @option{-mdynamic-no-pic} options.
22790 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
22792 generates code for the x86-64 architecture.
22794 The @option{-m16} option is the same as @option{-m32}, except for that
22795 it outputs the @code{.code16gcc} assembly directive at the beginning of
22796 the assembly output so that the binary can run in 16-bit mode.
22798 @item -mno-red-zone
22799 @opindex mno-red-zone
22800 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
22801 by the x86-64 ABI; it is a 128-byte area beyond the location of the
22802 stack pointer that is not modified by signal or interrupt handlers
22803 and therefore can be used for temporary data without adjusting the stack
22804 pointer. The flag @option{-mno-red-zone} disables this red zone.
22806 @item -mcmodel=small
22807 @opindex mcmodel=small
22808 Generate code for the small code model: the program and its symbols must
22809 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
22810 Programs can be statically or dynamically linked. This is the default
22813 @item -mcmodel=kernel
22814 @opindex mcmodel=kernel
22815 Generate code for the kernel code model. The kernel runs in the
22816 negative 2 GB of the address space.
22817 This model has to be used for Linux kernel code.
22819 @item -mcmodel=medium
22820 @opindex mcmodel=medium
22821 Generate code for the medium model: the program is linked in the lower 2
22822 GB of the address space. Small symbols are also placed there. Symbols
22823 with sizes larger than @option{-mlarge-data-threshold} are put into
22824 large data or BSS sections and can be located above 2GB. Programs can
22825 be statically or dynamically linked.
22827 @item -mcmodel=large
22828 @opindex mcmodel=large
22829 Generate code for the large model. This model makes no assumptions
22830 about addresses and sizes of sections.
22832 @item -maddress-mode=long
22833 @opindex maddress-mode=long
22834 Generate code for long address mode. This is only supported for 64-bit
22835 and x32 environments. It is the default address mode for 64-bit
22838 @item -maddress-mode=short
22839 @opindex maddress-mode=short
22840 Generate code for short address mode. This is only supported for 32-bit
22841 and x32 environments. It is the default address mode for 32-bit and
22845 @node x86 Windows Options
22846 @subsection x86 Windows Options
22847 @cindex x86 Windows Options
22848 @cindex Windows Options for x86
22850 These additional options are available for Microsoft Windows targets:
22856 specifies that a console application is to be generated, by
22857 instructing the linker to set the PE header subsystem type
22858 required for console applications.
22859 This option is available for Cygwin and MinGW targets and is
22860 enabled by default on those targets.
22864 This option is available for Cygwin and MinGW targets. It
22865 specifies that a DLL---a dynamic link library---is to be
22866 generated, enabling the selection of the required runtime
22867 startup object and entry point.
22869 @item -mnop-fun-dllimport
22870 @opindex mnop-fun-dllimport
22871 This option is available for Cygwin and MinGW targets. It
22872 specifies that the @code{dllimport} attribute should be ignored.
22876 This option is available for MinGW targets. It specifies
22877 that MinGW-specific thread support is to be used.
22881 This option is available for MinGW-w64 targets. It causes
22882 the @code{UNICODE} preprocessor macro to be predefined, and
22883 chooses Unicode-capable runtime startup code.
22887 This option is available for Cygwin and MinGW targets. It
22888 specifies that the typical Microsoft Windows predefined macros are to
22889 be set in the pre-processor, but does not influence the choice
22890 of runtime library/startup code.
22894 This option is available for Cygwin and MinGW targets. It
22895 specifies that a GUI application is to be generated by
22896 instructing the linker to set the PE header subsystem type
22899 @item -fno-set-stack-executable
22900 @opindex fno-set-stack-executable
22901 This option is available for MinGW targets. It specifies that
22902 the executable flag for the stack used by nested functions isn't
22903 set. This is necessary for binaries running in kernel mode of
22904 Microsoft Windows, as there the User32 API, which is used to set executable
22905 privileges, isn't available.
22907 @item -fwritable-relocated-rdata
22908 @opindex fno-writable-relocated-rdata
22909 This option is available for MinGW and Cygwin targets. It specifies
22910 that relocated-data in read-only section is put into .data
22911 section. This is a necessary for older runtimes not supporting
22912 modification of .rdata sections for pseudo-relocation.
22914 @item -mpe-aligned-commons
22915 @opindex mpe-aligned-commons
22916 This option is available for Cygwin and MinGW targets. It
22917 specifies that the GNU extension to the PE file format that
22918 permits the correct alignment of COMMON variables should be
22919 used when generating code. It is enabled by default if
22920 GCC detects that the target assembler found during configuration
22921 supports the feature.
22924 See also under @ref{x86 Options} for standard options.
22926 @node Xstormy16 Options
22927 @subsection Xstormy16 Options
22928 @cindex Xstormy16 Options
22930 These options are defined for Xstormy16:
22935 Choose startup files and linker script suitable for the simulator.
22938 @node Xtensa Options
22939 @subsection Xtensa Options
22940 @cindex Xtensa Options
22942 These options are supported for Xtensa targets:
22946 @itemx -mno-const16
22948 @opindex mno-const16
22949 Enable or disable use of @code{CONST16} instructions for loading
22950 constant values. The @code{CONST16} instruction is currently not a
22951 standard option from Tensilica. When enabled, @code{CONST16}
22952 instructions are always used in place of the standard @code{L32R}
22953 instructions. The use of @code{CONST16} is enabled by default only if
22954 the @code{L32R} instruction is not available.
22957 @itemx -mno-fused-madd
22958 @opindex mfused-madd
22959 @opindex mno-fused-madd
22960 Enable or disable use of fused multiply/add and multiply/subtract
22961 instructions in the floating-point option. This has no effect if the
22962 floating-point option is not also enabled. Disabling fused multiply/add
22963 and multiply/subtract instructions forces the compiler to use separate
22964 instructions for the multiply and add/subtract operations. This may be
22965 desirable in some cases where strict IEEE 754-compliant results are
22966 required: the fused multiply add/subtract instructions do not round the
22967 intermediate result, thereby producing results with @emph{more} bits of
22968 precision than specified by the IEEE standard. Disabling fused multiply
22969 add/subtract instructions also ensures that the program output is not
22970 sensitive to the compiler's ability to combine multiply and add/subtract
22973 @item -mserialize-volatile
22974 @itemx -mno-serialize-volatile
22975 @opindex mserialize-volatile
22976 @opindex mno-serialize-volatile
22977 When this option is enabled, GCC inserts @code{MEMW} instructions before
22978 @code{volatile} memory references to guarantee sequential consistency.
22979 The default is @option{-mserialize-volatile}. Use
22980 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
22982 @item -mforce-no-pic
22983 @opindex mforce-no-pic
22984 For targets, like GNU/Linux, where all user-mode Xtensa code must be
22985 position-independent code (PIC), this option disables PIC for compiling
22988 @item -mtext-section-literals
22989 @itemx -mno-text-section-literals
22990 @opindex mtext-section-literals
22991 @opindex mno-text-section-literals
22992 These options control the treatment of literal pools. The default is
22993 @option{-mno-text-section-literals}, which places literals in a separate
22994 section in the output file. This allows the literal pool to be placed
22995 in a data RAM/ROM, and it also allows the linker to combine literal
22996 pools from separate object files to remove redundant literals and
22997 improve code size. With @option{-mtext-section-literals}, the literals
22998 are interspersed in the text section in order to keep them as close as
22999 possible to their references. This may be necessary for large assembly
23002 @item -mtarget-align
23003 @itemx -mno-target-align
23004 @opindex mtarget-align
23005 @opindex mno-target-align
23006 When this option is enabled, GCC instructs the assembler to
23007 automatically align instructions to reduce branch penalties at the
23008 expense of some code density. The assembler attempts to widen density
23009 instructions to align branch targets and the instructions following call
23010 instructions. If there are not enough preceding safe density
23011 instructions to align a target, no widening is performed. The
23012 default is @option{-mtarget-align}. These options do not affect the
23013 treatment of auto-aligned instructions like @code{LOOP}, which the
23014 assembler always aligns, either by widening density instructions or
23015 by inserting NOP instructions.
23018 @itemx -mno-longcalls
23019 @opindex mlongcalls
23020 @opindex mno-longcalls
23021 When this option is enabled, GCC instructs the assembler to translate
23022 direct calls to indirect calls unless it can determine that the target
23023 of a direct call is in the range allowed by the call instruction. This
23024 translation typically occurs for calls to functions in other source
23025 files. Specifically, the assembler translates a direct @code{CALL}
23026 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
23027 The default is @option{-mno-longcalls}. This option should be used in
23028 programs where the call target can potentially be out of range. This
23029 option is implemented in the assembler, not the compiler, so the
23030 assembly code generated by GCC still shows direct call
23031 instructions---look at the disassembled object code to see the actual
23032 instructions. Note that the assembler uses an indirect call for
23033 every cross-file call, not just those that really are out of range.
23036 @node zSeries Options
23037 @subsection zSeries Options
23038 @cindex zSeries options
23040 These are listed under @xref{S/390 and zSeries Options}.
23042 @node Code Gen Options
23043 @section Options for Code Generation Conventions
23044 @cindex code generation conventions
23045 @cindex options, code generation
23046 @cindex run-time options
23048 These machine-independent options control the interface conventions
23049 used in code generation.
23051 Most of them have both positive and negative forms; the negative form
23052 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
23053 one of the forms is listed---the one that is not the default. You
23054 can figure out the other form by either removing @samp{no-} or adding
23058 @item -fbounds-check
23059 @opindex fbounds-check
23060 For front ends that support it, generate additional code to check that
23061 indices used to access arrays are within the declared range. This is
23062 currently only supported by the Java and Fortran front ends, where
23063 this option defaults to true and false respectively.
23065 @item -fstack-reuse=@var{reuse-level}
23066 @opindex fstack_reuse
23067 This option controls stack space reuse for user declared local/auto variables
23068 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
23069 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
23070 local variables and temporaries, @samp{named_vars} enables the reuse only for
23071 user defined local variables with names, and @samp{none} disables stack reuse
23072 completely. The default value is @samp{all}. The option is needed when the
23073 program extends the lifetime of a scoped local variable or a compiler generated
23074 temporary beyond the end point defined by the language. When a lifetime of
23075 a variable ends, and if the variable lives in memory, the optimizing compiler
23076 has the freedom to reuse its stack space with other temporaries or scoped
23077 local variables whose live range does not overlap with it. Legacy code extending
23078 local lifetime is likely to break with the stack reuse optimization.
23097 if (*p == 10) // out of scope use of local1
23108 A(int k) : i(k), j(k) @{ @}
23115 void foo(const A& ar)
23122 foo(A(10)); // temp object's lifetime ends when foo returns
23128 ap->i+= 10; // ap references out of scope temp whose space
23129 // is reused with a. What is the value of ap->i?
23134 The lifetime of a compiler generated temporary is well defined by the C++
23135 standard. When a lifetime of a temporary ends, and if the temporary lives
23136 in memory, the optimizing compiler has the freedom to reuse its stack
23137 space with other temporaries or scoped local variables whose live range
23138 does not overlap with it. However some of the legacy code relies on
23139 the behavior of older compilers in which temporaries' stack space is
23140 not reused, the aggressive stack reuse can lead to runtime errors. This
23141 option is used to control the temporary stack reuse optimization.
23145 This option generates traps for signed overflow on addition, subtraction,
23146 multiplication operations.
23150 This option instructs the compiler to assume that signed arithmetic
23151 overflow of addition, subtraction and multiplication wraps around
23152 using twos-complement representation. This flag enables some optimizations
23153 and disables others. This option is enabled by default for the Java
23154 front end, as required by the Java language specification.
23157 @opindex fexceptions
23158 Enable exception handling. Generates extra code needed to propagate
23159 exceptions. For some targets, this implies GCC generates frame
23160 unwind information for all functions, which can produce significant data
23161 size overhead, although it does not affect execution. If you do not
23162 specify this option, GCC enables it by default for languages like
23163 C++ that normally require exception handling, and disables it for
23164 languages like C that do not normally require it. However, you may need
23165 to enable this option when compiling C code that needs to interoperate
23166 properly with exception handlers written in C++. You may also wish to
23167 disable this option if you are compiling older C++ programs that don't
23168 use exception handling.
23170 @item -fnon-call-exceptions
23171 @opindex fnon-call-exceptions
23172 Generate code that allows trapping instructions to throw exceptions.
23173 Note that this requires platform-specific runtime support that does
23174 not exist everywhere. Moreover, it only allows @emph{trapping}
23175 instructions to throw exceptions, i.e.@: memory references or floating-point
23176 instructions. It does not allow exceptions to be thrown from
23177 arbitrary signal handlers such as @code{SIGALRM}.
23179 @item -fdelete-dead-exceptions
23180 @opindex fdelete-dead-exceptions
23181 Consider that instructions that may throw exceptions but don't otherwise
23182 contribute to the execution of the program can be optimized away.
23183 This option is enabled by default for the Ada front end, as permitted by
23184 the Ada language specification.
23185 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23187 @item -funwind-tables
23188 @opindex funwind-tables
23189 Similar to @option{-fexceptions}, except that it just generates any needed
23190 static data, but does not affect the generated code in any other way.
23191 You normally do not need to enable this option; instead, a language processor
23192 that needs this handling enables it on your behalf.
23194 @item -fasynchronous-unwind-tables
23195 @opindex fasynchronous-unwind-tables
23196 Generate unwind table in DWARF 2 format, if supported by target machine. The
23197 table is exact at each instruction boundary, so it can be used for stack
23198 unwinding from asynchronous events (such as debugger or garbage collector).
23200 @item -fno-gnu-unique
23201 @opindex fno-gnu-unique
23202 On systems with recent GNU assembler and C library, the C++ compiler
23203 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23204 of template static data members and static local variables in inline
23205 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23206 is necessary to avoid problems with a library used by two different
23207 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23208 therefore disagreeing with the other one about the binding of the
23209 symbol. But this causes @code{dlclose} to be ignored for affected
23210 DSOs; if your program relies on reinitialization of a DSO via
23211 @code{dlclose} and @code{dlopen}, you can use
23212 @option{-fno-gnu-unique}.
23214 @item -fpcc-struct-return
23215 @opindex fpcc-struct-return
23216 Return ``short'' @code{struct} and @code{union} values in memory like
23217 longer ones, rather than in registers. This convention is less
23218 efficient, but it has the advantage of allowing intercallability between
23219 GCC-compiled files and files compiled with other compilers, particularly
23220 the Portable C Compiler (pcc).
23222 The precise convention for returning structures in memory depends
23223 on the target configuration macros.
23225 Short structures and unions are those whose size and alignment match
23226 that of some integer type.
23228 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23229 switch is not binary compatible with code compiled with the
23230 @option{-freg-struct-return} switch.
23231 Use it to conform to a non-default application binary interface.
23233 @item -freg-struct-return
23234 @opindex freg-struct-return
23235 Return @code{struct} and @code{union} values in registers when possible.
23236 This is more efficient for small structures than
23237 @option{-fpcc-struct-return}.
23239 If you specify neither @option{-fpcc-struct-return} nor
23240 @option{-freg-struct-return}, GCC defaults to whichever convention is
23241 standard for the target. If there is no standard convention, GCC
23242 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23243 the principal compiler. In those cases, we can choose the standard, and
23244 we chose the more efficient register return alternative.
23246 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23247 switch is not binary compatible with code compiled with the
23248 @option{-fpcc-struct-return} switch.
23249 Use it to conform to a non-default application binary interface.
23251 @item -fshort-enums
23252 @opindex fshort-enums
23253 Allocate to an @code{enum} type only as many bytes as it needs for the
23254 declared range of possible values. Specifically, the @code{enum} type
23255 is equivalent to the smallest integer type that has enough room.
23257 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23258 code that is not binary compatible with code generated without that switch.
23259 Use it to conform to a non-default application binary interface.
23261 @item -fshort-double
23262 @opindex fshort-double
23263 Use the same size for @code{double} as for @code{float}.
23265 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23266 code that is not binary compatible with code generated without that switch.
23267 Use it to conform to a non-default application binary interface.
23269 @item -fshort-wchar
23270 @opindex fshort-wchar
23271 Override the underlying type for @code{wchar_t} to be @code{short
23272 unsigned int} instead of the default for the target. This option is
23273 useful for building programs to run under WINE@.
23275 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23276 code that is not binary compatible with code generated without that switch.
23277 Use it to conform to a non-default application binary interface.
23280 @opindex fno-common
23281 In C code, controls the placement of uninitialized global variables.
23282 Unix C compilers have traditionally permitted multiple definitions of
23283 such variables in different compilation units by placing the variables
23285 This is the behavior specified by @option{-fcommon}, and is the default
23286 for GCC on most targets.
23287 On the other hand, this behavior is not required by ISO C, and on some
23288 targets may carry a speed or code size penalty on variable references.
23289 The @option{-fno-common} option specifies that the compiler should place
23290 uninitialized global variables in the data section of the object file,
23291 rather than generating them as common blocks.
23292 This has the effect that if the same variable is declared
23293 (without @code{extern}) in two different compilations,
23294 you get a multiple-definition error when you link them.
23295 In this case, you must compile with @option{-fcommon} instead.
23296 Compiling with @option{-fno-common} is useful on targets for which
23297 it provides better performance, or if you wish to verify that the
23298 program will work on other systems that always treat uninitialized
23299 variable declarations this way.
23303 Ignore the @code{#ident} directive.
23305 @item -finhibit-size-directive
23306 @opindex finhibit-size-directive
23307 Don't output a @code{.size} assembler directive, or anything else that
23308 would cause trouble if the function is split in the middle, and the
23309 two halves are placed at locations far apart in memory. This option is
23310 used when compiling @file{crtstuff.c}; you should not need to use it
23313 @item -fverbose-asm
23314 @opindex fverbose-asm
23315 Put extra commentary information in the generated assembly code to
23316 make it more readable. This option is generally only of use to those
23317 who actually need to read the generated assembly code (perhaps while
23318 debugging the compiler itself).
23320 @option{-fno-verbose-asm}, the default, causes the
23321 extra information to be omitted and is useful when comparing two assembler
23324 @item -frecord-gcc-switches
23325 @opindex frecord-gcc-switches
23326 This switch causes the command line used to invoke the
23327 compiler to be recorded into the object file that is being created.
23328 This switch is only implemented on some targets and the exact format
23329 of the recording is target and binary file format dependent, but it
23330 usually takes the form of a section containing ASCII text. This
23331 switch is related to the @option{-fverbose-asm} switch, but that
23332 switch only records information in the assembler output file as
23333 comments, so it never reaches the object file.
23334 See also @option{-grecord-gcc-switches} for another
23335 way of storing compiler options into the object file.
23339 @cindex global offset table
23341 Generate position-independent code (PIC) suitable for use in a shared
23342 library, if supported for the target machine. Such code accesses all
23343 constant addresses through a global offset table (GOT)@. The dynamic
23344 loader resolves the GOT entries when the program starts (the dynamic
23345 loader is not part of GCC; it is part of the operating system). If
23346 the GOT size for the linked executable exceeds a machine-specific
23347 maximum size, you get an error message from the linker indicating that
23348 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23349 instead. (These maximums are 8k on the SPARC and 32k
23350 on the m68k and RS/6000. The x86 has no such limit.)
23352 Position-independent code requires special support, and therefore works
23353 only on certain machines. For the x86, GCC supports PIC for System V
23354 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23355 position-independent.
23357 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23362 If supported for the target machine, emit position-independent code,
23363 suitable for dynamic linking and avoiding any limit on the size of the
23364 global offset table. This option makes a difference on the m68k,
23365 PowerPC and SPARC@.
23367 Position-independent code requires special support, and therefore works
23368 only on certain machines.
23370 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23377 These options are similar to @option{-fpic} and @option{-fPIC}, but
23378 generated position independent code can be only linked into executables.
23379 Usually these options are used when @option{-pie} GCC option is
23380 used during linking.
23382 @option{-fpie} and @option{-fPIE} both define the macros
23383 @code{__pie__} and @code{__PIE__}. The macros have the value 1
23384 for @option{-fpie} and 2 for @option{-fPIE}.
23386 @item -fno-jump-tables
23387 @opindex fno-jump-tables
23388 Do not use jump tables for switch statements even where it would be
23389 more efficient than other code generation strategies. This option is
23390 of use in conjunction with @option{-fpic} or @option{-fPIC} for
23391 building code that forms part of a dynamic linker and cannot
23392 reference the address of a jump table. On some targets, jump tables
23393 do not require a GOT and this option is not needed.
23395 @item -ffixed-@var{reg}
23397 Treat the register named @var{reg} as a fixed register; generated code
23398 should never refer to it (except perhaps as a stack pointer, frame
23399 pointer or in some other fixed role).
23401 @var{reg} must be the name of a register. The register names accepted
23402 are machine-specific and are defined in the @code{REGISTER_NAMES}
23403 macro in the machine description macro file.
23405 This flag does not have a negative form, because it specifies a
23408 @item -fcall-used-@var{reg}
23409 @opindex fcall-used
23410 Treat the register named @var{reg} as an allocable register that is
23411 clobbered by function calls. It may be allocated for temporaries or
23412 variables that do not live across a call. Functions compiled this way
23413 do not save and restore the register @var{reg}.
23415 It is an error to use this flag with the frame pointer or stack pointer.
23416 Use of this flag for other registers that have fixed pervasive roles in
23417 the machine's execution model produces disastrous results.
23419 This flag does not have a negative form, because it specifies a
23422 @item -fcall-saved-@var{reg}
23423 @opindex fcall-saved
23424 Treat the register named @var{reg} as an allocable register saved by
23425 functions. It may be allocated even for temporaries or variables that
23426 live across a call. Functions compiled this way save and restore
23427 the register @var{reg} if they use it.
23429 It is an error to use this flag with the frame pointer or stack pointer.
23430 Use of this flag for other registers that have fixed pervasive roles in
23431 the machine's execution model produces disastrous results.
23433 A different sort of disaster results from the use of this flag for
23434 a register in which function values may be returned.
23436 This flag does not have a negative form, because it specifies a
23439 @item -fpack-struct[=@var{n}]
23440 @opindex fpack-struct
23441 Without a value specified, pack all structure members together without
23442 holes. When a value is specified (which must be a small power of two), pack
23443 structure members according to this value, representing the maximum
23444 alignment (that is, objects with default alignment requirements larger than
23445 this are output potentially unaligned at the next fitting location.
23447 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
23448 code that is not binary compatible with code generated without that switch.
23449 Additionally, it makes the code suboptimal.
23450 Use it to conform to a non-default application binary interface.
23452 @item -finstrument-functions
23453 @opindex finstrument-functions
23454 Generate instrumentation calls for entry and exit to functions. Just
23455 after function entry and just before function exit, the following
23456 profiling functions are called with the address of the current
23457 function and its call site. (On some platforms,
23458 @code{__builtin_return_address} does not work beyond the current
23459 function, so the call site information may not be available to the
23460 profiling functions otherwise.)
23463 void __cyg_profile_func_enter (void *this_fn,
23465 void __cyg_profile_func_exit (void *this_fn,
23469 The first argument is the address of the start of the current function,
23470 which may be looked up exactly in the symbol table.
23472 This instrumentation is also done for functions expanded inline in other
23473 functions. The profiling calls indicate where, conceptually, the
23474 inline function is entered and exited. This means that addressable
23475 versions of such functions must be available. If all your uses of a
23476 function are expanded inline, this may mean an additional expansion of
23477 code size. If you use @code{extern inline} in your C code, an
23478 addressable version of such functions must be provided. (This is
23479 normally the case anyway, but if you get lucky and the optimizer always
23480 expands the functions inline, you might have gotten away without
23481 providing static copies.)
23483 A function may be given the attribute @code{no_instrument_function}, in
23484 which case this instrumentation is not done. This can be used, for
23485 example, for the profiling functions listed above, high-priority
23486 interrupt routines, and any functions from which the profiling functions
23487 cannot safely be called (perhaps signal handlers, if the profiling
23488 routines generate output or allocate memory).
23490 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23491 @opindex finstrument-functions-exclude-file-list
23493 Set the list of functions that are excluded from instrumentation (see
23494 the description of @option{-finstrument-functions}). If the file that
23495 contains a function definition matches with one of @var{file}, then
23496 that function is not instrumented. The match is done on substrings:
23497 if the @var{file} parameter is a substring of the file name, it is
23498 considered to be a match.
23503 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
23507 excludes any inline function defined in files whose pathnames
23508 contain @file{/bits/stl} or @file{include/sys}.
23510 If, for some reason, you want to include letter @samp{,} in one of
23511 @var{sym}, write @samp{\,}. For example,
23512 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
23513 (note the single quote surrounding the option).
23515 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
23516 @opindex finstrument-functions-exclude-function-list
23518 This is similar to @option{-finstrument-functions-exclude-file-list},
23519 but this option sets the list of function names to be excluded from
23520 instrumentation. The function name to be matched is its user-visible
23521 name, such as @code{vector<int> blah(const vector<int> &)}, not the
23522 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
23523 match is done on substrings: if the @var{sym} parameter is a substring
23524 of the function name, it is considered to be a match. For C99 and C++
23525 extended identifiers, the function name must be given in UTF-8, not
23526 using universal character names.
23528 @item -fstack-check
23529 @opindex fstack-check
23530 Generate code to verify that you do not go beyond the boundary of the
23531 stack. You should specify this flag if you are running in an
23532 environment with multiple threads, but you only rarely need to specify it in
23533 a single-threaded environment since stack overflow is automatically
23534 detected on nearly all systems if there is only one stack.
23536 Note that this switch does not actually cause checking to be done; the
23537 operating system or the language runtime must do that. The switch causes
23538 generation of code to ensure that they see the stack being extended.
23540 You can additionally specify a string parameter: @samp{no} means no
23541 checking, @samp{generic} means force the use of old-style checking,
23542 @samp{specific} means use the best checking method and is equivalent
23543 to bare @option{-fstack-check}.
23545 Old-style checking is a generic mechanism that requires no specific
23546 target support in the compiler but comes with the following drawbacks:
23550 Modified allocation strategy for large objects: they are always
23551 allocated dynamically if their size exceeds a fixed threshold.
23554 Fixed limit on the size of the static frame of functions: when it is
23555 topped by a particular function, stack checking is not reliable and
23556 a warning is issued by the compiler.
23559 Inefficiency: because of both the modified allocation strategy and the
23560 generic implementation, code performance is hampered.
23563 Note that old-style stack checking is also the fallback method for
23564 @samp{specific} if no target support has been added in the compiler.
23566 @item -fstack-limit-register=@var{reg}
23567 @itemx -fstack-limit-symbol=@var{sym}
23568 @itemx -fno-stack-limit
23569 @opindex fstack-limit-register
23570 @opindex fstack-limit-symbol
23571 @opindex fno-stack-limit
23572 Generate code to ensure that the stack does not grow beyond a certain value,
23573 either the value of a register or the address of a symbol. If a larger
23574 stack is required, a signal is raised at run time. For most targets,
23575 the signal is raised before the stack overruns the boundary, so
23576 it is possible to catch the signal without taking special precautions.
23578 For instance, if the stack starts at absolute address @samp{0x80000000}
23579 and grows downwards, you can use the flags
23580 @option{-fstack-limit-symbol=__stack_limit} and
23581 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
23582 of 128KB@. Note that this may only work with the GNU linker.
23584 @item -fsplit-stack
23585 @opindex fsplit-stack
23586 Generate code to automatically split the stack before it overflows.
23587 The resulting program has a discontiguous stack which can only
23588 overflow if the program is unable to allocate any more memory. This
23589 is most useful when running threaded programs, as it is no longer
23590 necessary to calculate a good stack size to use for each thread. This
23591 is currently only implemented for the x86 targets running
23594 When code compiled with @option{-fsplit-stack} calls code compiled
23595 without @option{-fsplit-stack}, there may not be much stack space
23596 available for the latter code to run. If compiling all code,
23597 including library code, with @option{-fsplit-stack} is not an option,
23598 then the linker can fix up these calls so that the code compiled
23599 without @option{-fsplit-stack} always has a large stack. Support for
23600 this is implemented in the gold linker in GNU binutils release 2.21
23603 @item -fleading-underscore
23604 @opindex fleading-underscore
23605 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
23606 change the way C symbols are represented in the object file. One use
23607 is to help link with legacy assembly code.
23609 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
23610 generate code that is not binary compatible with code generated without that
23611 switch. Use it to conform to a non-default application binary interface.
23612 Not all targets provide complete support for this switch.
23614 @item -ftls-model=@var{model}
23615 @opindex ftls-model
23616 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
23617 The @var{model} argument should be one of @samp{global-dynamic},
23618 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
23619 Note that the choice is subject to optimization: the compiler may use
23620 a more efficient model for symbols not visible outside of the translation
23621 unit, or if @option{-fpic} is not given on the command line.
23623 The default without @option{-fpic} is @samp{initial-exec}; with
23624 @option{-fpic} the default is @samp{global-dynamic}.
23626 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
23627 @opindex fvisibility
23628 Set the default ELF image symbol visibility to the specified option---all
23629 symbols are marked with this unless overridden within the code.
23630 Using this feature can very substantially improve linking and
23631 load times of shared object libraries, produce more optimized
23632 code, provide near-perfect API export and prevent symbol clashes.
23633 It is @strong{strongly} recommended that you use this in any shared objects
23636 Despite the nomenclature, @samp{default} always means public; i.e.,
23637 available to be linked against from outside the shared object.
23638 @samp{protected} and @samp{internal} are pretty useless in real-world
23639 usage so the only other commonly used option is @samp{hidden}.
23640 The default if @option{-fvisibility} isn't specified is
23641 @samp{default}, i.e., make every
23642 symbol public---this causes the same behavior as previous versions of
23645 A good explanation of the benefits offered by ensuring ELF
23646 symbols have the correct visibility is given by ``How To Write
23647 Shared Libraries'' by Ulrich Drepper (which can be found at
23648 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
23649 solution made possible by this option to marking things hidden when
23650 the default is public is to make the default hidden and mark things
23651 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
23652 and @code{__attribute__ ((visibility("default")))} instead of
23653 @code{__declspec(dllexport)} you get almost identical semantics with
23654 identical syntax. This is a great boon to those working with
23655 cross-platform projects.
23657 For those adding visibility support to existing code, you may find
23658 @code{#pragma GCC visibility} of use. This works by you enclosing
23659 the declarations you wish to set visibility for with (for example)
23660 @code{#pragma GCC visibility push(hidden)} and
23661 @code{#pragma GCC visibility pop}.
23662 Bear in mind that symbol visibility should be viewed @strong{as
23663 part of the API interface contract} and thus all new code should
23664 always specify visibility when it is not the default; i.e., declarations
23665 only for use within the local DSO should @strong{always} be marked explicitly
23666 as hidden as so to avoid PLT indirection overheads---making this
23667 abundantly clear also aids readability and self-documentation of the code.
23668 Note that due to ISO C++ specification requirements, @code{operator new} and
23669 @code{operator delete} must always be of default visibility.
23671 Be aware that headers from outside your project, in particular system
23672 headers and headers from any other library you use, may not be
23673 expecting to be compiled with visibility other than the default. You
23674 may need to explicitly say @code{#pragma GCC visibility push(default)}
23675 before including any such headers.
23677 @code{extern} declarations are not affected by @option{-fvisibility}, so
23678 a lot of code can be recompiled with @option{-fvisibility=hidden} with
23679 no modifications. However, this means that calls to @code{extern}
23680 functions with no explicit visibility use the PLT, so it is more
23681 effective to use @code{__attribute ((visibility))} and/or
23682 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
23683 declarations should be treated as hidden.
23685 Note that @option{-fvisibility} does affect C++ vague linkage
23686 entities. This means that, for instance, an exception class that is
23687 be thrown between DSOs must be explicitly marked with default
23688 visibility so that the @samp{type_info} nodes are unified between
23691 An overview of these techniques, their benefits and how to use them
23692 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
23694 @item -fstrict-volatile-bitfields
23695 @opindex fstrict-volatile-bitfields
23696 This option should be used if accesses to volatile bit-fields (or other
23697 structure fields, although the compiler usually honors those types
23698 anyway) should use a single access of the width of the
23699 field's type, aligned to a natural alignment if possible. For
23700 example, targets with memory-mapped peripheral registers might require
23701 all such accesses to be 16 bits wide; with this flag you can
23702 declare all peripheral bit-fields as @code{unsigned short} (assuming short
23703 is 16 bits on these targets) to force GCC to use 16-bit accesses
23704 instead of, perhaps, a more efficient 32-bit access.
23706 If this option is disabled, the compiler uses the most efficient
23707 instruction. In the previous example, that might be a 32-bit load
23708 instruction, even though that accesses bytes that do not contain
23709 any portion of the bit-field, or memory-mapped registers unrelated to
23710 the one being updated.
23712 In some cases, such as when the @code{packed} attribute is applied to a
23713 structure field, it may not be possible to access the field with a single
23714 read or write that is correctly aligned for the target machine. In this
23715 case GCC falls back to generating multiple accesses rather than code that
23716 will fault or truncate the result at run time.
23718 Note: Due to restrictions of the C/C++11 memory model, write accesses are
23719 not allowed to touch non bit-field members. It is therefore recommended
23720 to define all bits of the field's type as bit-field members.
23722 The default value of this option is determined by the application binary
23723 interface for the target processor.
23725 @item -fsync-libcalls
23726 @opindex fsync-libcalls
23727 This option controls whether any out-of-line instance of the @code{__sync}
23728 family of functions may be used to implement the C++11 @code{__atomic}
23729 family of functions.
23731 The default value of this option is enabled, thus the only useful form
23732 of the option is @option{-fno-sync-libcalls}. This option is used in
23733 the implementation of the @file{libatomic} runtime library.
23739 @node Environment Variables
23740 @section Environment Variables Affecting GCC
23741 @cindex environment variables
23743 @c man begin ENVIRONMENT
23744 This section describes several environment variables that affect how GCC
23745 operates. Some of them work by specifying directories or prefixes to use
23746 when searching for various kinds of files. Some are used to specify other
23747 aspects of the compilation environment.
23749 Note that you can also specify places to search using options such as
23750 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
23751 take precedence over places specified using environment variables, which
23752 in turn take precedence over those specified by the configuration of GCC@.
23753 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
23754 GNU Compiler Collection (GCC) Internals}.
23759 @c @itemx LC_COLLATE
23761 @c @itemx LC_MONETARY
23762 @c @itemx LC_NUMERIC
23767 @c @findex LC_COLLATE
23768 @findex LC_MESSAGES
23769 @c @findex LC_MONETARY
23770 @c @findex LC_NUMERIC
23774 These environment variables control the way that GCC uses
23775 localization information which allows GCC to work with different
23776 national conventions. GCC inspects the locale categories
23777 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
23778 so. These locale categories can be set to any value supported by your
23779 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
23780 Kingdom encoded in UTF-8.
23782 The @env{LC_CTYPE} environment variable specifies character
23783 classification. GCC uses it to determine the character boundaries in
23784 a string; this is needed for some multibyte encodings that contain quote
23785 and escape characters that are otherwise interpreted as a string
23788 The @env{LC_MESSAGES} environment variable specifies the language to
23789 use in diagnostic messages.
23791 If the @env{LC_ALL} environment variable is set, it overrides the value
23792 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
23793 and @env{LC_MESSAGES} default to the value of the @env{LANG}
23794 environment variable. If none of these variables are set, GCC
23795 defaults to traditional C English behavior.
23799 If @env{TMPDIR} is set, it specifies the directory to use for temporary
23800 files. GCC uses temporary files to hold the output of one stage of
23801 compilation which is to be used as input to the next stage: for example,
23802 the output of the preprocessor, which is the input to the compiler
23805 @item GCC_COMPARE_DEBUG
23806 @findex GCC_COMPARE_DEBUG
23807 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
23808 @option{-fcompare-debug} to the compiler driver. See the documentation
23809 of this option for more details.
23811 @item GCC_EXEC_PREFIX
23812 @findex GCC_EXEC_PREFIX
23813 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
23814 names of the subprograms executed by the compiler. No slash is added
23815 when this prefix is combined with the name of a subprogram, but you can
23816 specify a prefix that ends with a slash if you wish.
23818 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
23819 an appropriate prefix to use based on the pathname it is invoked with.
23821 If GCC cannot find the subprogram using the specified prefix, it
23822 tries looking in the usual places for the subprogram.
23824 The default value of @env{GCC_EXEC_PREFIX} is
23825 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
23826 the installed compiler. In many cases @var{prefix} is the value
23827 of @code{prefix} when you ran the @file{configure} script.
23829 Other prefixes specified with @option{-B} take precedence over this prefix.
23831 This prefix is also used for finding files such as @file{crt0.o} that are
23834 In addition, the prefix is used in an unusual way in finding the
23835 directories to search for header files. For each of the standard
23836 directories whose name normally begins with @samp{/usr/local/lib/gcc}
23837 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
23838 replacing that beginning with the specified prefix to produce an
23839 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
23840 @file{foo/bar} just before it searches the standard directory
23841 @file{/usr/local/lib/bar}.
23842 If a standard directory begins with the configured
23843 @var{prefix} then the value of @var{prefix} is replaced by
23844 @env{GCC_EXEC_PREFIX} when looking for header files.
23846 @item COMPILER_PATH
23847 @findex COMPILER_PATH
23848 The value of @env{COMPILER_PATH} is a colon-separated list of
23849 directories, much like @env{PATH}. GCC tries the directories thus
23850 specified when searching for subprograms, if it can't find the
23851 subprograms using @env{GCC_EXEC_PREFIX}.
23854 @findex LIBRARY_PATH
23855 The value of @env{LIBRARY_PATH} is a colon-separated list of
23856 directories, much like @env{PATH}. When configured as a native compiler,
23857 GCC tries the directories thus specified when searching for special
23858 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
23859 using GCC also uses these directories when searching for ordinary
23860 libraries for the @option{-l} option (but directories specified with
23861 @option{-L} come first).
23865 @cindex locale definition
23866 This variable is used to pass locale information to the compiler. One way in
23867 which this information is used is to determine the character set to be used
23868 when character literals, string literals and comments are parsed in C and C++.
23869 When the compiler is configured to allow multibyte characters,
23870 the following values for @env{LANG} are recognized:
23874 Recognize JIS characters.
23876 Recognize SJIS characters.
23878 Recognize EUCJP characters.
23881 If @env{LANG} is not defined, or if it has some other value, then the
23882 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
23883 recognize and translate multibyte characters.
23887 Some additional environment variables affect the behavior of the
23890 @include cppenv.texi
23894 @node Precompiled Headers
23895 @section Using Precompiled Headers
23896 @cindex precompiled headers
23897 @cindex speed of compilation
23899 Often large projects have many header files that are included in every
23900 source file. The time the compiler takes to process these header files
23901 over and over again can account for nearly all of the time required to
23902 build the project. To make builds faster, GCC allows you to
23903 @dfn{precompile} a header file.
23905 To create a precompiled header file, simply compile it as you would any
23906 other file, if necessary using the @option{-x} option to make the driver
23907 treat it as a C or C++ header file. You may want to use a
23908 tool like @command{make} to keep the precompiled header up-to-date when
23909 the headers it contains change.
23911 A precompiled header file is searched for when @code{#include} is
23912 seen in the compilation. As it searches for the included file
23913 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
23914 compiler looks for a precompiled header in each directory just before it
23915 looks for the include file in that directory. The name searched for is
23916 the name specified in the @code{#include} with @samp{.gch} appended. If
23917 the precompiled header file can't be used, it is ignored.
23919 For instance, if you have @code{#include "all.h"}, and you have
23920 @file{all.h.gch} in the same directory as @file{all.h}, then the
23921 precompiled header file is used if possible, and the original
23922 header is used otherwise.
23924 Alternatively, you might decide to put the precompiled header file in a
23925 directory and use @option{-I} to ensure that directory is searched
23926 before (or instead of) the directory containing the original header.
23927 Then, if you want to check that the precompiled header file is always
23928 used, you can put a file of the same name as the original header in this
23929 directory containing an @code{#error} command.
23931 This also works with @option{-include}. So yet another way to use
23932 precompiled headers, good for projects not designed with precompiled
23933 header files in mind, is to simply take most of the header files used by
23934 a project, include them from another header file, precompile that header
23935 file, and @option{-include} the precompiled header. If the header files
23936 have guards against multiple inclusion, they are skipped because
23937 they've already been included (in the precompiled header).
23939 If you need to precompile the same header file for different
23940 languages, targets, or compiler options, you can instead make a
23941 @emph{directory} named like @file{all.h.gch}, and put each precompiled
23942 header in the directory, perhaps using @option{-o}. It doesn't matter
23943 what you call the files in the directory; every precompiled header in
23944 the directory is considered. The first precompiled header
23945 encountered in the directory that is valid for this compilation is
23946 used; they're searched in no particular order.
23948 There are many other possibilities, limited only by your imagination,
23949 good sense, and the constraints of your build system.
23951 A precompiled header file can be used only when these conditions apply:
23955 Only one precompiled header can be used in a particular compilation.
23958 A precompiled header can't be used once the first C token is seen. You
23959 can have preprocessor directives before a precompiled header; you cannot
23960 include a precompiled header from inside another header.
23963 The precompiled header file must be produced for the same language as
23964 the current compilation. You can't use a C precompiled header for a C++
23968 The precompiled header file must have been produced by the same compiler
23969 binary as the current compilation is using.
23972 Any macros defined before the precompiled header is included must
23973 either be defined in the same way as when the precompiled header was
23974 generated, or must not affect the precompiled header, which usually
23975 means that they don't appear in the precompiled header at all.
23977 The @option{-D} option is one way to define a macro before a
23978 precompiled header is included; using a @code{#define} can also do it.
23979 There are also some options that define macros implicitly, like
23980 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
23983 @item If debugging information is output when using the precompiled
23984 header, using @option{-g} or similar, the same kind of debugging information
23985 must have been output when building the precompiled header. However,
23986 a precompiled header built using @option{-g} can be used in a compilation
23987 when no debugging information is being output.
23989 @item The same @option{-m} options must generally be used when building
23990 and using the precompiled header. @xref{Submodel Options},
23991 for any cases where this rule is relaxed.
23993 @item Each of the following options must be the same when building and using
23994 the precompiled header:
23996 @gccoptlist{-fexceptions}
23999 Some other command-line options starting with @option{-f},
24000 @option{-p}, or @option{-O} must be defined in the same way as when
24001 the precompiled header was generated. At present, it's not clear
24002 which options are safe to change and which are not; the safest choice
24003 is to use exactly the same options when generating and using the
24004 precompiled header. The following are known to be safe:
24006 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
24007 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
24008 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
24013 For all of these except the last, the compiler automatically
24014 ignores the precompiled header if the conditions aren't met. If you
24015 find an option combination that doesn't work and doesn't cause the
24016 precompiled header to be ignored, please consider filing a bug report,
24019 If you do use differing options when generating and using the
24020 precompiled header, the actual behavior is a mixture of the
24021 behavior for the options. For instance, if you use @option{-g} to
24022 generate the precompiled header but not when using it, you may or may
24023 not get debugging information for routines in the precompiled header.